ARTICLE X. DATA

Section 279. Data, AI, Cybersecurity, and Verifiable Compute Governance Purpose

279.1 Data Governance Purpose.

279.1.1 Data governance shall protect the Corporation’s public-benefit purposes, research integrity, evidence integrity, methods integrity, public-good technical asset stewardship, public authority trust, community safeguards, privacy, cybersecurity, protected knowledge, legal compliance, correctionability, and institutional continuity by governing the collection, receipt, creation, classification, access, processing, use, sharing, publication, retention, deletion, correction, restriction, and archival of data.

279.1.2 Data governance shall apply to all data received, generated, observed, inferred, derived, classified, transformed, stored, accessed, transmitted, published, restricted, or archived by or for the Corporation, including research data, evidence data, observability data, public authority data, personal information, health-sensitive data, cyber-sensitive data, infrastructure-sensitive data, community-protected data, Tribal / Indigenous data, local and territorial knowledge, cultural and environmental knowledge, protected knowledge, model outputs, inference records, telemetry, logs, datasets, documents, maps, dashboards, repositories, software artifacts, and technical baselines.

279.1.3 Data governance shall be interpreted as a fiduciary-grade institutional control function and not as a technical convenience, file-management practice, informal research preference, platform default, vendor feature, or optional administrative procedure.

279.2 Privacy Purpose.

279.2.1 Privacy governance shall protect persons, communities, rights-bearing data, personal information, sensitive personal information, health-sensitive data, public authority data, children’s or youth data where applicable, education data where applicable, community-protected data, and other privacy-relevant information from unauthorized collection, use, disclosure, retention, linkage, profiling, inference, publication, AI processing, re-identification, or misuse.

279.2.2 Privacy governance shall support lawfulness, fairness, transparency, purpose limitation, data minimization, access control, security, accuracy, correction, deletion where applicable, complaint pathways, non-retaliation, public-safe publication, and rights-respecting research.

279.2.3 Privacy protection shall not be subordinated to institutional convenience, sponsor preference, provider capability, AI-tool availability, publication pressure, public authority interest, capital-reader interest, media interest, or technical ambition.

279.3 AI Governance Purpose.

279.3.1 AI governance shall control the acquisition, development, deployment, configuration, access, use, monitoring, review, and retirement of AI systems used by the Corporation, including generative AI, agentic systems, retrieval systems, classifiers, summarizers, transcription tools, translation tools, embedding systems, model-evaluation tools, anomaly detectors, routing tools, decision-support systems, and AI-assisted research workflows.

279.3.2 AI governance shall preserve human responsibility, source-grounded evidence, model transparency where appropriate, no-training restrictions, prompt and inference controls, bias review, hallucination review, data leakage prevention, protected knowledge protection, public authority boundary discipline, public-safe publication, correctionability, and non-execution.

279.3.3 AI systems shall not be treated as directors, officers, public authorities, certifiers, recognizers, finance-readiness authorities, procurement authorities, legal advisors, auditors, emergency commanders, public warning bodies, or final arbiters of technical truth.

279.4 Cybersecurity Purpose.

279.4.1 Cybersecurity governance shall protect the Corporation’s systems, records, repositories, credentials, keys, tokens, datasets, models, inference records, controlled rooms, public-good software, technical baselines, dashboards, maps, communications, financial systems, and institutional operations from unauthorized access, misuse, compromise, disclosure, manipulation, destruction, disruption, or exfiltration.

279.4.2 Cybersecurity shall support confidentiality, integrity, availability, resilience, least privilege, secure development, secure collaboration, vulnerability management, incident response, business continuity, evidence preservation, public authority trust, protected knowledge protection, and public-safe publication.

279.4.3 Cybersecurity governance shall not convert the Corporation into a managed security services provider, incident commander, regulator, cyber-certification body, public authority cyber command body, insurance underwriter, or cyber-compliance approver.

279.5 Verifiable Compute Purpose.

279.5.1 Verifiable compute governance shall support trustworthy, traceable, reproducible where appropriate, auditable where lawful and safe, and correctionable computation used in research, evidence, methods, observability, AI evaluation, public-good software, technical baselines, dashboards, simulations, digital twins, and public-safe publications.

279.5.2 Verifiable compute may include compute provenance, environment records, code versioning, dependency records, execution logs, workflow receipts, cryptographic receipts where appropriate, container records, hardware or cloud environment records, model evaluation records, reproducible computation methods, secure execution records, and controlled compute access.

279.5.3 Verifiable compute shall support evidence integrity and technical memory but shall not constitute certification, legal compliance approval, public authority approval, finance-readiness, procurement approval, security guarantee, rating, or enterprise execution.

279.6 Verifiable Intelligence Purpose.

279.6.1 Verifiable intelligence governance shall support source-grounded, evidence-linked, reviewable, auditable where appropriate, human-governed, and correctionable use of AI-assisted analysis, machine inference, knowledge retrieval, classification, summarization, translation, anomaly detection, and decision-support outputs.

279.6.2 Verifiable intelligence shall require linkage between AI-assisted outputs and source records, methods, model identity, version, inference context, human review, limitations, uncertainty, access class, public-safe status, and correction path where material.

279.6.3 Verifiable intelligence shall not mean that AI outputs are inherently true. It means that intelligence outputs are traceable, reviewable, bounded, documented, and capable of correction.

279.7 Controlled-Room Purpose.

279.7.1 Controlled rooms shall provide secure, access-controlled, purpose-limited environments for reviewing restricted data, public authority data, cyber-sensitive data, infrastructure-sensitive data, health-sensitive data, community-protected data, protected knowledge, confidential research materials, controlled technology, privileged materials, or other sensitive records.

279.7.2 Controlled rooms may include clean rooms, data rooms, no-download rooms, restricted repositories, public authority rooms, regulator-listening rooms, public finance reader rooms, safeguards rooms, cyber rooms, protected knowledge rooms, and other controlled environments approved by competent authority.

279.7.3 Controlled-room access shall be based on eligibility, authority, need, confidentiality obligations, data / AI / cyber controls, conflict review, public authority capacity, safeguards, and access records, and shall not be granted as a donor, sponsor, provider, investor, public authority, or relationship benefit.

279.8 Public Authority Data Protection Purpose.

279.8.1 Public authority data protection shall preserve the confidentiality, integrity, lawful use, public records status, restricted-use status, public-sector ethics status, security, privacy, and public-safe meaning of data received from or relating to public authorities.

279.8.2 Public authority data shall be handled according to applicable law, contract, public records requirements, FOIA and state sunshine considerations where applicable, public authority restrictions, confidentiality terms, procurement rules, grant rules, government ethics rules, security requirements, and public authority capacity records.

279.8.3 Receipt or handling of public authority data shall not create public authority delegation, official adoption, public warning authority, emergency command, procurement approval, regulatory approval, public finance approval, grant approval, or sovereign decision-making authority.

279.9 Rights-Bearing Data Protection Purpose.

279.9.1 Rights-bearing data protection shall govern data that relates to persons, communities, groups, identities, protected classes, rights, benefits, access, participation, vulnerability, civil rights, accessibility, public health, employment, education, housing, public safety, emergency exposure, environmental justice, or other interests capable of affecting rights, dignity, safety, or opportunity.

279.9.2 Rights-bearing data shall be governed by heightened safeguards, including purpose limitation, minimization, fairness, bias review, civil rights review, accessibility review, non-discrimination, non-retaliation, privacy, security, consent or lawful basis where required, public-safe review, and correction pathways.

279.10 Health-Sensitive Data Protection Purpose.

279.10.1 Health-sensitive data protection shall govern health, public health, clinical, behavioral, epidemiological, disability-related, emergency medical, health-system, vulnerable-population, youth, or health-adjacent information handled by the Corporation.

279.10.2 Health-sensitive data shall be reviewed for applicable privacy obligations, ethical review, consent, de-identification, re-identification risk, public health authority restrictions, vulnerable-population safeguards, AI-use restrictions, publication risk, and public-safe handling.

279.10.3 The Corporation shall not use health-sensitive data to provide diagnosis, clinical advice, public health orders, emergency medical direction, health certification, public warning, or public authority action.

279.11 Cyber-Sensitive and Infrastructure-Sensitive Data Protection Purpose.

279.11.1 Cyber-sensitive and infrastructure-sensitive data protection shall govern data that could expose vulnerabilities, credentials, keys, tokens, architecture diagrams, network configurations, incident records, telemetry, security logs, critical infrastructure locations, operational dependencies, emergency systems, public works systems, utilities, telecom systems, ports, transportation, water, energy, food, health systems, or other sensitive infrastructure.

279.11.2 Such data shall be handled through access controls, need-to-know review, public-safe publication review, redaction, aggregation, controlled-room use, cyber review, incident-response coordination, and correctionability.

279.11.3 The Corporation shall not publish or route cyber-sensitive or infrastructure-sensitive data in a manner that increases harm, enables misuse, creates public warning confusion, or compromises public authority trust.

279.12 Community-Protected, Tribal / Indigenous, Local, Territorial, Cultural, Environmental, and Protected Knowledge Protection Purpose.

279.12.1 Community-protected, Tribal / Indigenous, local, territorial, cultural, environmental, sacred, ecological, vulnerability-related, and protected knowledge shall be handled with heightened respect, restriction, protocol sensitivity, consent awareness, attribution discipline, public-safe mapping controls, AI-use restrictions, non-extraction discipline, non-retaliation, and correction pathways.

279.12.2 Protected knowledge shall not be collected, translated, generalized, embedded, trained on, mapped, published, transferred, commercialized, or used for sponsor, provider, public authority, finance-facing, or enterprise purposes without lawful authority, appropriate permission, protocol compliance where applicable, and safeguards review.

279.13 Public-Safe Publication Support Purpose.

279.13.1 Data, AI, cybersecurity, and compute governance shall support public-safe publication by ensuring that outputs released by the Corporation are accurate, lawful, evidence-supported, privacy-aware, cyber-safe, infrastructure-safe, protected-knowledge-safe, public authority-boundary compliant, finance-boundary compliant, certification-boundary compliant, procurement-neutral, recognition-boundary compliant, and correctionable.

279.13.2 Public-safe publication controls may require redaction, aggregation, masking, delay, controlled annexes, restricted circulation, no-download access, non-attribution, limitation language, correction notices, or withdrawal.

279.14 Evidence Integrity Support Purpose.

279.14.1 Data, AI, cybersecurity, and compute governance shall support evidence integrity by preserving source lineage, provenance, custody, timestamps, permissions, data rights, transformation history, model identity, inference records, compute records, access logs, correction history, and auditability where appropriate.

279.14.2 No data, AI output, cyber log, compute result, telemetry stream, model output, or dashboard element shall be treated as evidence unless reviewed, classified, permissioned, contextualized, and correctionable under the applicable evidence controls.

279.15 Research Integrity Support Purpose.

279.15.1 Data, AI, cybersecurity, and compute governance shall support research integrity by ensuring lawful collection, ethical use, secure storage, accurate analysis, appropriate AI assistance, reproducible or reviewable computation where appropriate, limitation disclosure, uncertainty disclosure, bias review, conflict management, and correction.

279.15.2 Research integrity shall not be compromised by undocumented data transformations, unapproved AI tools, insecure repositories, hidden model training, missing inference records, uncontrolled access, provider dependency, sponsor pressure, or public authority ambiguity.

279.16 Public-Good Technical Asset Support Purpose.

279.16.1 Data, AI, cybersecurity, and compute governance shall support public-good technical assets, including public-good software, open technical baselines, schemas, APIs, SDKs, dashboards, maps, model cards, system cards, benchmark cards, proof receipts, observability tools, methods repositories, and technical documentation.

279.16.2 Public-good technical assets shall be secured, versioned, licensed, documented, access-controlled where appropriate, dependency-reviewed, vulnerability-reviewed, AI-use-reviewed where applicable, protected from private enclosure, and correctionable.

279.17 Incident Response Purpose.

279.17.1 Incident response shall ensure that suspected or confirmed data, AI, cybersecurity, privacy, compute, controlled-room, protected knowledge, public authority data, repository, model, or publication incidents are identified, contained, investigated, corrected, reported where required, and learned from.

279.17.2 Incident response shall preserve evidence, logs, privilege, confidentiality, public authority trust, non-retaliation, protected knowledge, affected-person protection, public-safe communication, and recurrence prevention.

279.18 Non-Execution Boundary for Data, AI, Cybersecurity, and Compute Outputs.

279.18.1 Data, AI, cybersecurity, verifiable compute, verifiable intelligence, controlled-room outputs, dashboards, logs, model outputs, benchmark outputs, public-good software outputs, technical baselines, and public-safe reports shall not constitute or be represented as:

279.18.1(a) public authority action, public warning, emergency command, public safety order, regulatory approval, procurement approval, grant approval, public finance approval, or sovereign decision; 279.18.1(b) securities offering, investment advice, broker-dealer activity, capital placement, lending, insurance, underwriting, rating, guarantee, bankability, investability, finance-readiness, or public finance execution; 279.18.1(c) certification, accreditation, legal compliance approval, cybersecurity compliance approval, privacy compliance approval, AI safety approval, professional advice, provider qualification, or procurement qualification; or 279.18.1(d) provider preference, enterprise deployment approval, infrastructure operation, national company execution, Project SPV execution, managed security service, commercial AI service, or operational command.

279.18.2 Where outputs may be misread as execution, the Corporation shall apply limitation language, restrict access, re-scope the output, correct public claims, or decline publication.

279.19 Data, AI, Cybersecurity, and Verifiable Compute Governance Records.

279.19.1 The Corporation shall maintain Data, AI, Cybersecurity, and Verifiable Compute Governance Records, including data governance records, privacy records, AI governance records, cybersecurity records, verifiable compute records, verifiable intelligence records, controlled-room records, public authority data protection records, rights-bearing data records, health-sensitive data records, cyber-sensitive and infrastructure-sensitive data records, protected knowledge records, public-safe publication support records, evidence integrity support records, research integrity support records, public-good technical asset records, incident response records, non-execution boundary records, corrections, supersessions, withdrawals, and archive records.

---

Section 280. Data Governance Program

280.1 Data Governance Program Requirement.

280.1.1 The Corporation shall maintain a Data Governance Program proportionate to its size, risk profile, data classes, research activities, public authority interfaces, public-good technical assets, AI use, cyber exposure, controlled rooms, publications, cross-border activities, community safeguards, protected knowledge obligations, and institutional maturity.

280.1.2 The Data Governance Program shall govern data inventory, data classification, data stewardship roles, lawful basis, authority, permission, consent, notice, access control, use limitation, data quality, minimization, retention, deletion, sharing, publication, AI use, security, incident response, correction, and records.

280.1.3 The Data Governance Program shall apply to all directors, officers, employees, contractors, fellows, advisors, volunteers, committee members, researchers, technical contributors, reviewers, collaborators, vendors, providers, sponsors, public authority participants, and other persons with access to Corporation data.

280.2 Board Oversight.

280.2.1 The Board shall oversee the Data Governance Program at the level of mission alignment, legal compliance, risk appetite, public-benefit purpose, fiscal capacity, public authority trust, privacy, cybersecurity, AI governance, protected knowledge, safeguards, and institutional accountability.

280.2.2 Board oversight may include approval of data governance policies, risk thresholds, sensitive-data categories, AI-use restrictions, controlled-room rules, incident escalation thresholds, public-safe publication principles, and records-retention frameworks.

280.2.3 The Board may delegate implementation oversight to a committee, officer, data governance body, security body, research integrity body, or other competent authority, provided that delegation is recorded and does not eliminate Board accountability for material governance failures.

280.3 Officer Responsibility.

280.3.1 Officers designated by the Board or policy shall be responsible for implementing the Data Governance Program within delegated authority.

280.3.2 Officer responsibilities may include maintaining inventories, approving data classifications, overseeing access controls, coordinating privacy review, coordinating AI governance, coordinating cybersecurity, maintaining controlled rooms, ensuring records retention, managing incidents, and reporting material risks to the Board.

280.3.3 Officers shall not authorize data use, AI processing, access, publication, sharing, deletion, or transfer inconsistent with law, restriction, public authority obligations, protected knowledge duties, safeguards, or this Bylaw.

280.4 Data Governance Policy.

280.4.1 The Corporation shall adopt and maintain one or more data governance policies or equivalent recorded controls addressing data lifecycle, classification, access, stewardship, lawful basis, permissions, consent, notice, sharing, publication, AI use, retention, deletion, correction, incident response, and controlled-room use.

280.4.2 Data governance policies shall be reviewed periodically and updated when law, technology, data classes, public authority interfaces, AI tools, cybersecurity risk, research activities, safeguards, or institutional operations materially change.

280.4.3 Policy exceptions shall require recorded approval, risk assessment, mitigation, expiry or review date, and correction path.

280.5 Data Inventory.

280.5.1 The Corporation shall maintain a Data Inventory identifying material datasets, records collections, evidence repositories, research datasets, public authority data, personal information, health-sensitive data, cyber-sensitive data, infrastructure-sensitive data, protected knowledge, model records, inference records, telemetry streams, logs, dashboards, maps, software repositories, and technical asset data.

280.5.2 The Data Inventory shall identify, where applicable, source, owner, custodian, purpose, data class, location, system, format, access class, lawful basis, permissions, restrictions, retention, deletion, AI-use status, public-safe status, and correction path.

280.6 Data Register.

280.6.1 The Corporation shall maintain a Data Register or equivalent structured record for material data assets.

280.6.2 The Data Register shall support governance, auditability, access control, public-safe publication, research integrity, privacy compliance, AI governance, cybersecurity, data rights, protected knowledge handling, and correctionability.

280.6.3 Data Register entries shall be updated when data is received, created, transformed, reclassified, shared, published, restricted, corrected, superseded, archived, deleted, or affected by an incident.

280.7 Processing Register.

280.7.1 The Corporation shall maintain a Processing Register or equivalent record identifying material processing activities, including collection, receipt, storage, analysis, transformation, AI processing, sharing, publication, retention, deletion, transfer, and archival.

280.7.2 The Processing Register shall identify purpose, lawful basis or authority, data categories, data subjects or rights-bearing persons where applicable, recipients, systems, processors, cross-border transfers, AI tools, security controls, retention, public-safe status, and risk assessments where applicable.

280.8 Data Stewardship Roles.

280.8.1 The Corporation shall define data stewardship roles sufficient to manage data responsibly throughout the data lifecycle.

280.8.2 Data stewardship roles may include Data Owner, Data Custodian, Data Processor, Data User, Data Contributor, reviewer, approver, privacy lead, security lead, AI governance lead, public authority data lead, protected knowledge lead, and controlled-room custodian.

280.8.3 Data stewardship roles shall be recorded and shall not create authority beyond the applicable delegation, policy, contract, access control, or law.

280.9 Data Owner.

280.9.1 A Data Owner is the authorized person, role, committee, or body responsible for determining the business, research, evidence, public-good, or institutional purpose of a data asset and approving its use within applicable authority.

280.9.2 The Data Owner shall ensure that data use aligns with lawful basis, permission, purpose limitation, classification, access controls, retention, public-safe publication status, and correction obligations.

280.9.3 Data Owner status does not create personal ownership, private control, publication veto, public authority authority, finance-readiness authority, certification authority, recognition authority, or procurement authority.

280.10 Data Custodian.

280.10.1 A Data Custodian is the authorized person, role, system owner, or function responsible for operational custody, storage, security, access control, backup, retention, deletion, logging, archival, and technical integrity of a data asset.

280.10.2 The Data Custodian shall maintain access records, security controls, system records, version records, correction records, and incident records as appropriate.

280.10.3 Custody shall not confer authority to use, publish, share, train on, commercialize, or disclose data beyond approved purpose and permission.

280.11 Data Processor.

280.11.1 A Data Processor is a person or entity that processes data on behalf of the Corporation or under the Corporation’s instructions, including vendors, service providers, cloud providers, AI providers, software providers, contractors, consultants, and technical processors.

280.11.2 Data Processors shall be governed by written or recorded terms appropriate to data class and risk, including confidentiality, security, permitted use, prohibited use, sub-processing, breach notice, retention, deletion, audit, no-sale, no-unauthorized-use, no-unauthorized-AI-training, and return or deletion obligations.

280.11.3 No Data Processor shall use Corporation data for its own purposes unless separately authorized by lawful agreement, data rights, public-safe review, and competent approval.

280.12 Data User.

280.12.1 A Data User is any person authorized to access or use data for a defined purpose.

280.12.2 Data Users shall comply with access limits, purpose limits, confidentiality, privacy, AI-use restrictions, public authority restrictions, protected knowledge restrictions, cyber controls, publication controls, and correction obligations.

280.12.3 Data Users shall not copy, export, upload, train on, disclose, publish, share, or reuse data outside authorized purpose.

280.13 Data Contributor.

280.13.1 A Data Contributor is any person, entity, system, public authority, community participant, provider, sponsor, researcher, collaborator, sensor, repository, model, or other source contributing data to the Corporation.

280.13.2 Data Contributor records shall identify contributor identity where lawful and appropriate, contributor capacity, authority, permissions, restrictions, source context, data rights, attribution status, public-safe status, and correction path.

280.13.3 Contribution of data shall not itself create ownership transfer, publication permission, AI-training permission, public authority action, recognition, finance-readiness, certification, procurement approval, provider preference, or endorsement.

280.14 Data Subject or Rights-Bearing Person Where Applicable.

280.14.1 A Data Subject or rights-bearing person includes a person whose personal information, sensitive personal information, health-sensitive data, education data, youth data, biometric data, geolocation data, public authority-related data, civil rights-related data, or other rights-bearing data is collected, used, stored, analyzed, inferred, shared, published, or otherwise processed by the Corporation.

280.14.2 Where applicable law or policy gives such persons rights of access, correction, deletion, restriction, objection, portability, appeal, complaint, consent withdrawal, or other rights, the Corporation shall maintain pathways to receive and process such requests.

280.15 Data Source Authority.

280.15.1 Data Source Authority means the legal, contractual, institutional, public authority, community, Tribal / Indigenous, contributor, license, consent, research, or other authority under which data is provided to or used by the Corporation.

280.15.2 Data Source Authority shall be recorded for material datasets and sensitive data classes.

280.15.3 Where source authority is uncertain, the Corporation shall restrict, hold, quarantine, clarify, delete, return, or decline use pending review.

280.16 Data Lifecycle Governance.

280.16.1 Data Lifecycle Governance shall govern collection, receipt, creation, ingestion, classification, storage, transformation, analysis, AI processing, review, sharing, publication, retention, correction, restriction, archival, deletion, and secure disposal.

280.16.2 Each lifecycle stage shall be governed by purpose limitation, data minimization, access control, security, permission, data quality, public-safe status, safeguards, and correctionability.

280.17 Data Quality Governance.

280.17.1 Data Quality Governance shall address completeness, accuracy, timeliness, consistency, validity, reliability, representativeness, bias, missingness, duplication, transformation history, source authority, and fitness for purpose.

280.17.2 Data of insufficient quality shall be labeled, corrected, supplemented, restricted, excluded, or used only with limitations appropriate to the intended purpose.

280.18 Data Minimization.

280.18.1 The Corporation shall collect, receive, process, retain, publish, and share only data reasonably necessary for lawful, public-benefit, mission-aligned, and approved purposes.

280.18.2 Data minimization shall apply to research design, evidence intake, AI processing, public authority learning, controlled-room access, public-safe publication, datasets, dashboards, maps, model prompts, inference records, and technical baselines.

280.18.3 Data shall not be retained merely because storage is inexpensive, AI systems may later find it useful, sponsors or providers request it, public authority participants show interest, or future unspecified research might arise.

280.19 Purpose Limitation.

280.19.1 Data shall be used only for the purpose for which it was collected, received, generated, authorized, or otherwise lawfully approved, unless reuse is reviewed and authorized under this Bylaw.

280.19.2 Purpose limitation shall prevent unauthorized reuse for AI training, model improvement, embeddings, public-safe publication, finance-readiness inputs, provider development, sponsor benefit, public authority use beyond capacity, commercial use, or external sharing.

280.20 Data Governance Records.

280.20.1 The Corporation shall maintain Data Governance Records, including data governance policies, Board oversight records, officer responsibility records, Data Inventory entries, Data Register entries, Processing Register entries, stewardship role records, Data Owner records, Data Custodian records, Data Processor records, Data User records, Data Contributor records, Data Subject or rights-bearing person records where applicable, Data Source Authority records, lifecycle governance records, data quality records, minimization records, purpose limitation records, corrections, deletions, restrictions, and archive records.

---

Section 281. Lawful Basis, Authority, Permission, Consent, Notice, and Use Limitation

281.1 Lawful Basis Requirement.

281.1.1 The Corporation shall not collect, receive, store, process, analyze, transform, classify, publish, share, license, train on, embed, archive, delete, or otherwise use data unless a lawful basis, authority, permission, consent where required, notice where required, or other competent basis supports the activity.

281.1.2 Lawful basis shall be determined in relation to the data class, jurisdiction, source, data subject, contributor, public authority context, research purpose, contract, consent, license, legal obligation, legitimate institutional function, public-benefit purpose, and applicable law.

281.1.3 Absence of objection, technical access, platform availability, public website availability, sponsor submission, provider submission, public authority presence, community participation, or prior informal use shall not by itself establish lawful basis.

281.2 Authority Record Requirement.

281.2.1 Material data activities shall have authority records identifying the institutional, legal, contractual, research, public authority, community, Tribal / Indigenous, license, consent, or contributor authority under which the activity occurs.

281.2.2 Authority records shall be required for public authority data, rights-bearing data, health-sensitive data, children’s or youth data where applicable, education data where applicable, cyber-sensitive data, infrastructure-sensitive data, controlled technology data, export-controlled or sanctions-sensitive data, community-protected data, and protected knowledge.

281.3 Permission Record Requirement.

281.3.1 Permission records shall identify whether the Corporation may receive, access, store, process, analyze, transform, summarize, classify, publish, share, license, embed, train on, archive, delete, or otherwise use the data.

281.3.2 Permission records shall include permitted uses, prohibited uses, restrictions, attribution terms, non-attribution terms, retention terms, deletion terms, sharing terms, publication terms, AI-use terms, no-training terms, and correction terms where applicable.

281.4 Consent Where Required.

281.4.1 Consent shall be obtained where required by law, ethics review, research protocol, contract, community safeguards, Tribal / Indigenous protocol, public authority restriction, data contributor requirement, or institutional policy.

281.4.2 Consent shall be specific enough for the data activity, informed to the extent required, voluntary, recorded, revocable where applicable, and not obtained through coercion, retaliation threat, public authority pressure, sponsor pressure, provider pressure, scholarship pressure, employment pressure, or community pressure.

281.4.3 Consent for one purpose shall not automatically permit reuse for AI training, publication, mapping, finance-readiness inputs, provider development, sponsor benefit, public authority sharing, external transfer, or commercial use.

281.5 Notice Where Required.

281.5.1 Notice shall be provided where required by law, policy, research protocol, consent process, public authority requirement, platform terms, or institutional practice.

281.5.2 Notice shall describe, as applicable, the Corporation’s identity, purpose of processing, data categories, sources, uses, recipients, AI use, retention, rights, complaint pathways, public-safe publication possibility, restrictions, and contact pathway.

281.5.3 Notice shall be accurate and shall not conceal AI processing, public authority data use, sponsor or provider involvement where material, protected knowledge implications, or publication risks.

281.6 Contractual Authority.

281.6.1 Contractual authority may support data processing where a valid contract, data-sharing agreement, research agreement, grant agreement, sponsorship agreement, vendor agreement, public authority agreement, contributor agreement, license, or other instrument grants or restricts the activity.

281.6.2 Contractual authority shall be reviewed for scope, term, permitted use, prohibited use, confidentiality, privacy, public authority restrictions, data rights, AI-use restrictions, sub-processing, cross-border transfer, publication, retention, deletion, audit, correction, and termination obligations.

281.7 Public Authority Permission.

281.7.1 Public authority permission shall be recorded where data is received from, provided to, generated with, or used in relation to public authorities.

281.7.2 Public authority permission records shall identify the public authority entity, contributor capacity, official or non-official status, data restrictions, public records implications, public-sector ethics implications, confidentiality, security, permitted uses, publication limits, and whether any public authority approval or action exists.

281.7.3 Public authority permission to receive or use data shall not imply official adoption, public warning authority, emergency command, procurement approval, regulatory approval, public finance approval, grant approval, or sovereign obligation.

281.8 Research Authority.

281.8.1 Research authority may arise from approved research protocol, ethics determination, IRB or equivalent review, public-benefit purpose, data contributor agreement, license, consent, law, or other competent record.

281.8.2 Research authority shall identify scope, data classes, participant protections, privacy protections, AI-use limits, publication limits, retention, deletion, safeguards, and correction path.

281.8.3 Research authority shall not be used to bypass privacy, public authority, protected knowledge, cyber, export-control, sanctions, or publication restrictions.

281.9 Community Permission Where Applicable.

281.9.1 Community permission shall be obtained or respected where required by law, ethics review, community safeguards, research protocol, local practice, protected knowledge requirements, or public-safe mapping review.

281.9.2 Community permission may include consent, non-consent, attribution rules, non-attribution rules, withdrawal pathways, restriction terms, grievance pathways, public-safe publication conditions, and correction rights.

281.9.3 Community participation shall not be treated as blanket permission for publication, mapping, AI processing, sponsor use, provider use, or public authority sharing.

281.10 Tribal / Indigenous Protocol Permission Where Applicable.

281.10.1 Tribal / Indigenous protocol permission shall be required where applicable for Tribal or Indigenous data, Indigenous knowledge, cultural knowledge, environmental knowledge, sacred knowledge, land-based knowledge, local knowledge, community-protected information, public-safe mapping, or other protected knowledge.

281.10.2 The Corporation shall not assume that public availability, academic publication, government possession, platform access, individual disclosure, sponsor possession, or provider access creates Tribal / Indigenous permission.

281.10.3 Where permission is uncertain, the Corporation shall apply restrictive handling until clarified by competent record.

281.11 Data Contributor Permission.

281.11.1 Data Contributor permission shall identify what data is contributed, by whom or what source, in what capacity, under what rights, for what purposes, under what restrictions, and with what correction or withdrawal pathways.

281.11.2 Data Contributor permission shall not be overread to include broader reuse, publication, AI training, public authority sharing, finance-facing use, provider development, sponsor benefit, or external sharing without review.

281.12 Dataset License Authority.

281.12.1 Dataset license authority shall be reviewed before using licensed datasets, open datasets, proprietary datasets, research datasets, public authority datasets, commercial datasets, geospatial datasets, model datasets, or community datasets.

281.12.2 License review shall assess permitted use, attribution, redistribution, derivative works, commercial use, public authority use, AI training, embeddings, model improvement, publication, restrictions, termination, and audit obligations.

281.13 Open Data Authority.

281.13.1 Open data may be used only according to its actual license, public release terms, public authority restrictions, data rights, privacy constraints, attribution requirements, and public-safe status.

281.13.2 Open availability shall not eliminate obligations relating to privacy, re-identification, protected knowledge, public authority sensitivity, cyber sensitivity, infrastructure sensitivity, export-control, sanctions, or public-safe publication.

281.14 Restricted Data Authority.

281.14.1 Restricted data shall be used only under the restrictions that govern it and only by authorized persons, for approved purposes, in approved systems, with approved safeguards.

281.14.2 Restricted data shall not be copied, exported, summarized, embedded, trained on, published, mapped, disclosed, or externally shared except as expressly authorized.

281.15 Use Limitation.

281.15.1 Data use shall be limited to approved purposes, approved users, approved systems, approved time periods, approved outputs, approved AI tools, approved publication pathways, and approved sharing pathways.

281.15.2 Use limitation shall be enforced through access controls, policy controls, workflow controls, contracts, technical controls, training, monitoring, and corrective action.

281.16 Reuse Review.

281.16.1 Reuse of data for a purpose different from the original purpose shall require review for lawful basis, permission, consent, notice, contract restrictions, data rights, public authority restrictions, privacy, AI-use restrictions, cyber risk, protected knowledge, community safeguards, ethics, public-safe publication, and correctionability.

281.16.2 Reuse review shall be documented before reuse occurs, except in emergency protective action where immediate action is necessary and subsequent review is recorded promptly.

281.17 No Unauthorized Reuse for AI Training, Fine-Tuning, Embeddings, Publication, Finance-Readiness Inputs, Provider Development, Sponsor Benefit, or External Sharing.

281.17.1 No data shall be reused for AI training, fine-tuning, embeddings, model improvement, retrieval systems, publication, finance-readiness inputs, GRA-facing inputs, provider development, sponsor benefit, donor benefit, public authority sharing, commercial use, external sharing, or enterprise-stack activity unless lawful basis, permission, data rights, public-safe status, safeguards, and competent approval support such use.

281.17.2 Unauthorized reuse shall trigger hold, deletion where appropriate, access restriction, incident review, correction of affected outputs, notification where required, and enforcement.

281.18 Lawful Basis, Permission, Consent, Notice, and Use Records.

281.18.1 The Corporation shall maintain Lawful Basis, Permission, Consent, Notice, and Use Records, including lawful basis determinations, authority records, permission records, consent records, consent withdrawal records, notice records, contractual authority records, public authority permission records, research authority records, community permission records, Tribal / Indigenous protocol permission records, Data Contributor permission records, dataset license records, open data authority records, restricted data authority records, use limitation records, reuse reviews, unauthorized reuse records, corrections, deletions, restrictions, and archive records.

---

Section 282. Data Classification System

282.1 Data Classification Purpose.

282.1.1 The Corporation shall maintain a Data Classification System to classify data according to public-safe status, confidentiality, rights impact, privacy sensitivity, public authority restrictions, cyber sensitivity, infrastructure sensitivity, financial sensitivity, commercial sensitivity, research sensitivity, community safeguards, protected knowledge, controlled technology, export-control, sanctions sensitivity, children’s or youth status where applicable, and permitted use.

282.1.2 Data classification shall determine access controls, storage requirements, AI-use rules, sharing limits, publication pathways, controlled-room requirements, retention, deletion, public-safe review, incident escalation, and correction obligations.

282.1.3 Classification shall be reviewed and updated when facts, law, permissions, restrictions, public authority status, data rights, sensitivity, technology, public-safe status, or safeguards change.

282.2 Public Data.

282.2.1 Public Data means data lawfully available for public access and approved for public use by the Corporation without additional access restriction.

282.2.2 Public Data shall still be reviewed for accuracy, source, license, attribution, re-identification risk, protected knowledge risk, public authority sensitivity, cyber sensitivity, infrastructure sensitivity, and public-safe context before reliance or publication.

282.2.3 Public availability shall not by itself authorize AI training, redistribution, derivative use, commercial use, or publication in a new public-safe context.

282.3 Public-Safe Data.

282.3.1 Public-Safe Data means data approved for public release or public-facing use after review for privacy, confidentiality, public authority restrictions, protected knowledge, cyber sensitivity, infrastructure sensitivity, community harm, public-safe mapping, boundary risks, and misuse risk.

282.3.2 Public-Safe Data may include redacted, aggregated, masked, delayed, summarized, generalized, or otherwise controlled data.

282.3.3 Public-Safe Data classification shall identify limitations, permitted uses, public-safe status, correction path, and whether downstream users may rely on, reproduce, or redistribute the data.

282.4 Internal Data.

282.4.1 Internal Data means data intended for ordinary internal Corporation use by authorized persons for approved purposes and not approved for public release.

282.4.2 Internal Data shall be protected from unauthorized external sharing and shall not be uploaded to unapproved AI systems, public repositories, external platforms, or public communication channels.

282.5 Confidential Data.

282.5.1 Confidential Data means data subject to confidentiality obligations, including contracts, grants, donor restrictions, sponsorship terms, employment records, personnel records, unpublished research, internal deliberations, Board materials, non-public financial information, non-public partner information, non-public public authority materials, and controlled communications.

282.5.2 Confidential Data shall be access-controlled, stored securely, shared only with authorized persons, and reviewed before AI processing, publication, external sharing, or deletion.

282.6 Restricted Data.

282.6.1 Restricted Data means data subject to heightened access, use, sharing, publication, processing, AI-use, retention, or deletion limitations due to law, contract, public authority restriction, privacy, security, protected knowledge, export-control, sanctions, ethics review, or safeguards.

282.6.2 Restricted Data may require controlled-room handling, no-download access, encryption, segmentation, logging, approval workflows, and special incident escalation.

282.7 Rights-Bearing Data.

282.7.1 Rights-Bearing Data means data whose collection, classification, inference, use, sharing, publication, or automated processing may affect the rights, access, dignity, safety, opportunity, benefits, vulnerability, reputation, participation, or legal interests of persons or communities.

282.7.2 Rights-Bearing Data shall receive heightened review for fairness, bias, civil rights, accessibility, consent, notice, privacy, non-retaliation, public-safe publication, and correction.

282.8 Personal Information.

282.8.1 Personal Information means information that identifies, relates to, describes, is reasonably capable of being associated with, or could reasonably be linked with a particular person or household, subject to applicable law.

282.8.2 Personal Information shall be handled according to applicable privacy obligations, notices, consents where required, access controls, retention limits, security controls, rights pathways, and breach response rules.

282.9 Sensitive Personal Information.

282.9.1 Sensitive Personal Information includes personal information that receives heightened legal, ethical, or institutional protection, including government identifiers, financial account information, precise geolocation, biometric information, health information, disability information, racial or ethnic information, religious information, union status, sexual orientation, citizenship or immigration information, children’s data, education data, vulnerability-related data, or other sensitive categories under applicable law.

282.9.2 Sensitive Personal Information shall be collected and used only where necessary, lawful, protected, purpose-limited, access-controlled, and reviewed for publication and AI-use risk.

282.10 Health-Sensitive Data.

282.10.1 Health-Sensitive Data includes health, public health, medical, clinical, behavioral health, disability, epidemiological, health-system, emergency medical, health-services, or health-adjacent information capable of affecting privacy, stigma, rights, safety, or public interpretation.

282.10.2 Health-Sensitive Data shall be handled through heightened privacy, ethics, security, consent, de-identification, public-safe publication, and public authority review where applicable.

282.11 Public Authority Data.

282.11.1 Public Authority Data means data received from, created with, relating to, or restricted by a public authority, including data from agencies, public officials, public employees, public institutions, public infrastructure operators, public health bodies, emergency management bodies, public safety bodies, utilities, public works bodies, regulators, grantors, procurement bodies, and public finance actors.

282.11.2 Public Authority Data shall be classified according to capacity, authority, public records implications, confidentiality, security, official status, permitted use, public-safe status, and publication restrictions.

282.12 Cyber-Sensitive Data.

282.12.1 Cyber-Sensitive Data includes credentials, keys, tokens, secrets, authentication records, access logs, endpoint logs, network logs, cloud logs, repository logs, incident records, vulnerability details, exploit information, security architecture, threat indicators, forensic records, and cyber-risk information.

282.12.2 Cyber-Sensitive Data shall be access-controlled, secured, logged, restricted from public release unless public-safe, and handled under incident and vulnerability protocols where applicable.

282.13 Infrastructure-Sensitive Data.

282.13.1 Infrastructure-Sensitive Data includes data relating to the location, design, operation, vulnerability, dependency, performance, outage, recovery, security, or resilience of critical infrastructure, public works, utilities, telecom, ports, transportation, energy, water, food, health, emergency, cyber-physical, or industrial systems.

282.13.2 Infrastructure-Sensitive Data shall be reviewed for public-safe mapping, national security sensitivity, cyber risk, public authority restrictions, community harm, insurance sensitivity, finance-boundary risk, and misuse risk.

282.14 Finance-Sensitive Data.

282.14.1 Finance-Sensitive Data includes non-public financial information, donor information, sponsor information, funder information, grant budgets, restricted-fund data, bank data, payment data, payroll data, reimbursement data, reserve data, contract pricing, capital-reader materials, finance-readiness-facing inputs, and financial boundary-sensitive materials.

282.14.2 Finance-Sensitive Data shall not be used to imply investment advice, finance-readiness, underwriting approval, rating, public finance approval, bankability, insurance-readiness, creditworthiness, or capital execution.

282.15 Commercially Sensitive Data.

282.15.1 Commercially Sensitive Data includes trade secrets, non-public business information, vendor information, provider documentation, pricing, roadmap information, customer information, procurement-sensitive materials, proprietary datasets, technical documentation, product information, and confidential collaboration materials.

282.15.2 Commercially Sensitive Data shall be handled according to contract, confidentiality, IP, public-safe, competition, procurement-neutrality, provider-neutrality, and conflict controls.

282.16 Research-Sensitive Data.

282.16.1 Research-Sensitive Data includes unpublished research, preliminary findings, reviewer notes, peer review records, ethics records, participant information, raw field notes, protected research materials, experimental results, incomplete datasets, disputed evidence, and materials not ready for publication.

282.16.2 Research-Sensitive Data shall be protected to preserve research integrity, confidentiality, participant protection, public-safe framing, and correctionability.

282.17 Community-Protected Data.

282.17.1 Community-Protected Data includes data provided by or relating to communities where disclosure, misuse, mapping, attribution, AI processing, re-identification, public authority sharing, sponsor use, provider use, or external sharing could create harm, stigma, retaliation, extraction, vulnerability, or loss of trust.

282.17.2 Community-Protected Data shall be governed by safeguards, consent or non-consent, attribution or non-attribution, public-safe mapping, grievance pathways, restriction, correction, and non-retaliation.

282.18 Tribal / Indigenous, Local, Territorial, Cultural, Environmental, and Protected Knowledge Data.

282.18.1 Tribal / Indigenous, local, territorial, cultural, environmental, sacred, ecological, land-based, vulnerability-related, and Protected Knowledge Data shall be treated as highly sensitive unless competent record determines otherwise.

282.18.2 Such data shall be governed by protocol respect, authority records, permission records, protected knowledge review, public-safe mapping controls, AI-use restrictions, attribution discipline, withdrawal or restriction pathways, and correction.

282.19 Controlled Technology Data.

282.19.1 Controlled Technology Data includes technical data, software, designs, specifications, methods, encryption information, AI models, cybersecurity tools, telecom methods, AI-RAN / O-RAN materials, DePIN or DLT materials, geospatial systems, digital twin systems, advanced manufacturing information, semiconductor information, controlled research, or other materials subject to controlled-technology review.

282.19.2 Controlled Technology Data shall be reviewed for access, nationality where lawful and required, export-control, sanctions, licensing, public release, controlled-room requirements, and technical security.

282.20 Export-Controlled or Sanctions-Sensitive Data.

282.20.1 Export-Controlled or Sanctions-Sensitive Data includes data, software, technical information, controlled technology, transaction information, end-user information, restricted-party information, jurisdiction-sensitive information, or other materials that may implicate sanctions, export-control, anti-terrorism, or national security-sensitive obligations.

282.20.2 Such data shall be held, restricted, licensed where required, denied, deleted, returned, quarantined, or otherwise controlled pending competent review.

282.21 Children’s or Youth Data Where Applicable.

282.21.1 Children’s or Youth Data includes data concerning minors, students, youth participants, school participants, youth program participants, or other protected young persons.

282.21.2 Children’s or Youth Data shall receive heightened privacy, consent, notice, ethics, safety, access, retention, deletion, publication, AI-use, and safeguards review where applicable.

282.22 Data Classification Review and Reclassification.

282.22.1 Data classification shall be reviewed at intake, before material processing, before AI use, before sharing, before publication, before controlled-room access, before cross-border transfer, before retention expiry, after correction, after challenge, after incident, and when law, facts, permissions, sensitivity, or public-safe status changes.

282.22.2 Reclassification shall be recorded, including reason, authority, prior classification, new classification, affected users, affected outputs, access changes, publication effects, and correction obligations.

282.23 Data Classification Records.

282.23.1 The Corporation shall maintain Data Classification Records, including classification rules, classification decisions, public data records, public-safe data records, internal data records, confidential data records, restricted data records, rights-bearing data records, personal information records, sensitive personal information records, health-sensitive data records, public authority data records, cyber-sensitive data records, infrastructure-sensitive data records, finance-sensitive data records, commercially sensitive data records, research-sensitive data records, community-protected data records, Tribal / Indigenous and protected knowledge data records, controlled technology records, export-control and sanctions-sensitive records, children’s or youth data records, classification reviews, reclassifications, corrections, restrictions, and archive records.

---

Section 283. Privacy Compliance and Rights-Bearing Data

283.1 Privacy Compliance Purpose.

283.1.1 Privacy compliance shall ensure that the Corporation collects, receives, processes, stores, analyzes, shares, publishes, deletes, restricts, and archives personal information, sensitive personal information, rights-bearing data, health-sensitive data, children’s or youth data, education data, public authority data, and community-protected data lawfully, fairly, securely, transparently, purpose-limited, and correctionably.

283.1.2 Privacy compliance shall be interpreted broadly to include legal compliance, ethical data stewardship, civil rights, accessibility, data minimization, privacy by design, privacy impact assessment, public-safe publication, AI-use controls, cross-border transfer controls, complaint pathways, and incident response.

283.2 Federal Privacy Obligations Where Applicable.

283.2.1 The Corporation shall comply with applicable federal privacy, consumer protection, health, education, communications, cybersecurity, civil rights, public authority, employment, tax, nonprofit, and sectoral requirements where they apply to the Corporation’s data activities.

283.2.2 Federal privacy review shall be conducted where data relates to health, children, education, telecommunications, financial information, biometric information, consumer protection, public authority data, grants, public-sector programs, research activities, or federally funded activity.

283.3 State Privacy Obligations Where Applicable.

283.3.1 The Corporation shall comply with applicable state privacy, consumer data protection, biometric, health data, breach notification, data security, children’s privacy, education privacy, employment privacy, charitable solicitation, public records, cybersecurity, and civil rights obligations where they apply.

283.3.2 State privacy obligations shall be reviewed in relation to the Corporation’s all-states-and-territories operating posture, events, research, public authority learning, donor and sponsor activity, contractors, employees, datasets, publications, websites, platforms, subscriptions, Academy programs, and AI use.

283.4 Territorial Privacy Obligations Where Applicable.

283.4.1 The Corporation shall review and comply with applicable privacy, data protection, public authority, health, youth, employment, breach notification, civil rights, and local legal obligations in United States territorial and insular contexts where they apply.

283.4.2 Territorial privacy review shall account for language access, infrastructure constraints, disaster contexts, public health contexts, public authority capacity, local law, federal-territorial interfaces, and public-safe publication.

283.5 Sectoral Privacy Obligations Where Applicable.

283.5.1 The Corporation shall identify and comply with sectoral privacy obligations where data relates to health, education, communications, employment, financial information, biometric information, geolocation, public authority data, public safety, emergency management, utilities, critical infrastructure, research, or other regulated sectors.

283.5.2 Sectoral privacy obligations shall be addressed before data is collected, received, processed, shared, published, used in AI systems, or transferred across jurisdictions.

283.6 Health Privacy Obligations Where Applicable.

283.6.1 Health privacy obligations shall be reviewed wherever the Corporation handles health-sensitive data, public health data, clinical data, health-system data, disability-related data, health-adjacent data, emergency medical information, epidemiological data, or vulnerable-population health information.

283.6.2 Health-sensitive processing may require consent, authorization, de-identification, ethical review, public health authority review, business associate or equivalent terms where applicable, security controls, publication restrictions, and heightened incident response.

283.6.3 Health privacy governance shall preserve the Corporation’s non-clinical, non-diagnostic, non-public-health-order, non-emergency-command role.

283.7 Children’s Data Obligations Where Applicable.

283.7.1 Children’s or youth data shall be processed only where lawful, necessary, safeguarded, consented where required, noticed where required, ethically reviewed where appropriate, and protected against profiling, exposure, retaliation, inappropriate AI use, public-safe mapping risk, and publication harm.

283.7.2 Children’s or youth data shall not be processed through AI systems, published, shared, mapped, or retained beyond approved purpose without heightened review.

283.8 Education Data Obligations Where Applicable.

283.8.1 Education data shall be governed where the Corporation works with students, fellows, schools, universities, training programs, Academy programs, workforce programs, public authority learning, or research involving education records.

283.8.2 Education data processing shall be reviewed for applicable law, institutional policy, consent, school or university restrictions, student privacy, youth safeguards, accessibility, civil rights, AI-use limits, publication limits, and retention.

283.9 Public Authority Data Privacy Obligations.

283.9.1 Public authority data involving personal information, public health data, emergency management data, public safety data, benefits data, utility data, infrastructure data, procurement data, grant data, regulatory data, or public finance data shall be processed according to public authority restrictions, applicable privacy law, public records requirements, confidentiality, security, and capacity records.

283.9.2 Public authority data shall not be repurposed for sponsor benefit, provider development, AI training, finance-readiness inputs, public release, or external sharing without competent public authority permission and Corporation review.

283.10 Cross-Border Privacy Obligations.

283.10.1 Cross-border data processing shall be reviewed for privacy, data protection, data localization, transfer restrictions, contractual safeguards, public authority restrictions, Indigenous data governance, protected knowledge, cloud location, AI processing, cybersecurity, sanctions, export-control, and public-safe publication.

283.10.2 Cross-border privacy review shall preserve the Corporation’s United States legal-seat discipline, legal separateness from GCRI Canada and other entities, no shared treasury, no agency, no partnership, no joint venture, and no uncontrolled data transfer.

283.11 Notice and Transparency.

283.11.1 The Corporation shall provide privacy notices, research notices, public authority data notices, participant notices, website notices, AI-use notices, controlled-room notices, or other notices where required or appropriate.

283.11.2 Notices shall be accurate, accessible, understandable for the intended audience, updated when material practices change, and aligned with actual data use.

283.11.3 Notice shall not be used as a substitute for consent, legal authority, ethics review, public authority permission, protected knowledge permission, or safeguards where those are required.

283.12 Consent and Choice Where Required.

283.12.1 The Corporation shall provide consent, choice, opt-in, opt-out, withdrawal, restriction, non-attribution, or similar mechanisms where required by law, ethics review, protocol, contract, public authority terms, community safeguards, or institutional policy.

283.12.2 Consent and choice mechanisms shall be accessible, not coercive, records-supported, and capable of honoring withdrawal or restriction where applicable.

283.13 Access, Correction, Deletion, Portability, Restriction, and Objection Where Applicable.

283.13.1 Where applicable law, policy, contract, research protocol, public authority requirement, or institutional commitment provides rights to access, correction, deletion, portability, restriction, objection, withdrawal, or similar rights, the Corporation shall maintain processes to receive, verify, evaluate, respond to, record, and implement such requests.

283.13.2 Rights requests may be limited where necessary to protect other persons, public authority restrictions, legal obligations, research integrity, evidence integrity, protected knowledge, cybersecurity, privilege, retention obligations, or public-safe archive obligations.

283.13.3 Denials, partial denials, delays, or limitations shall be recorded with reason and appeal or complaint pathway where applicable.

283.14 Complaint and Appeal Pathways Where Applicable.

283.14.1 The Corporation shall maintain complaint and appeal pathways where required or appropriate for privacy concerns, rights-bearing data concerns, data misuse, AI misuse, protected knowledge misuse, public authority data concerns, rights requests, access denials, correction denials, deletion denials, publication concerns, and safeguards concerns.

283.14.2 Complaint pathways shall protect good-faith complainants from retaliation and shall provide accessible channels appropriate to the affected population.

283.15 Privacy Impact Assessment Where Required or Appropriate.

283.15.1 Privacy Impact Assessment shall be conducted where required or appropriate for material processing of personal information, sensitive personal information, rights-bearing data, health-sensitive data, children’s or youth data, public authority data, community-protected data, AI-assisted processing, cross-border transfer, public-safe publication, dashboards, maps, controlled rooms, or new systems.

283.15.2 Privacy Impact Assessment shall identify purpose, lawful basis, data categories, affected persons, risks, safeguards, retention, deletion, sharing, AI use, public-safe publication, incident response, rights pathways, and residual risk.

283.16 Data Protection Impact Assessment Where Required or Appropriate.

283.16.1 Data Protection Impact Assessment or equivalent heightened assessment shall be conducted where processing presents heightened risk to rights, freedoms, dignity, safety, civil rights, accessibility, protected participation, public authority trust, community safeguards, or protected knowledge.

283.16.2 Heightened risk processing may include large-scale processing, sensitive data processing, automated processing, profiling, AI-assisted classification, public-safe mapping, cyber-sensitive processing, infrastructure-sensitive processing, health-sensitive processing, children’s data processing, cross-border transfer, or processing involving vulnerable communities.

283.16.3 Processing shall be modified, restricted, delayed, refused, or escalated where assessment identifies unacceptable residual risk.

283.17 Privacy by Design.

283.17.1 The Corporation shall apply privacy by design and privacy by default principles to systems, repositories, dashboards, maps, AI tools, controlled rooms, research protocols, public authority learning materials, publications, software, technical baselines, and data-sharing arrangements.

283.17.2 Privacy by design may include minimization, pseudonymization, aggregation, access controls, encryption, logging, deletion schedules, no-training defaults, restricted embeddings, differential access, public-safe review, role separation, and human review.

283.17.3 Privacy controls shall be integrated at design stage rather than retrofitted after collection, processing, AI use, or publication where reasonably feasible.

283.18 Privacy Compliance Records.

283.18.1 The Corporation shall maintain Privacy Compliance Records, including federal privacy reviews, state privacy reviews, territorial privacy reviews, sectoral privacy reviews, health privacy reviews, children’s data reviews, education data reviews, public authority data privacy reviews, cross-border privacy reviews, notices, consent records, choice records, access requests, correction requests, deletion requests, portability requests, restriction requests, objection requests, complaint records, appeal records, privacy impact assessments, data protection impact assessments, privacy by design records, incident records, corrections, restrictions, deletions, and archive records.

Section 284. Public Authority Data

284.1 Public Authority Data Definition.

284.1.1 Public Authority Data means any data, record, document, dataset, telemetry, image, map, dashboard material, operational information, public-sector communication, public records material, confidential public-sector material, public health material, emergency-management material, public safety material, public works material, utility material, infrastructure material, regulatory material, grant material, procurement material, public finance material, cyber material, environmental material, geospatial material, or other information received from, created with, held for, generated by, derived from, or relating to a public authority.

284.1.2 Public Authority Data includes data associated with federal, state, District of Columbia, territorial, Tribal, Indigenous governmental, local, county, municipal, metropolitan, utility, port, public health, emergency management, public safety, public works, telecom, energy, water, food, cyber, infrastructure, regulatory, grant, procurement, and public finance contexts.

284.1.3 Public Authority Data shall be treated according to its source, authority, legal status, public records status, confidentiality status, capacity record, data class, permitted use, public-safe status, publication restrictions, transfer restrictions, AI-use restrictions, and correction path.

284.2 Public Authority Data Contribution Purpose.

284.2.1 Public Authority Data may be contributed to the Corporation only for lawful public-benefit purposes, including research, evidence development, methods development, observability methods, ontology, public authority learning, public-good software, technical baselines, public-safe publication, safeguards, and Nexus-compatible evidence infrastructure.

284.2.2 Public Authority Data contribution shall not create public authority delegation, official adoption, public warning authority, emergency command, procurement approval, grant approval, regulatory approval, public finance approval, sovereign obligation, public-private partnership, recognition, finance-readiness, certification, or enterprise execution.

284.2.3 The Corporation shall not accept, use, publish, transfer, or process Public Authority Data where the authority, permission, capacity, restriction, confidentiality, or public-safe status is unclear and material to lawful handling.

284.3 Federal Public Authority Data.

284.3.1 Federal Public Authority Data shall be reviewed for applicable federal restrictions, public records considerations, confidentiality terms, grant conditions, procurement rules, government ethics rules, privacy, security, controlled unclassified information where applicable, export-control, sanctions, public health rules, cyber rules, infrastructure sensitivity, and public-safe publication constraints.

284.3.2 Federal participation or federal data contribution shall be capacity-classified and shall not be represented as federal endorsement, official adoption, regulatory approval, procurement approval, grant approval, public finance approval, public warning, emergency command, or federal decision unless supported by a separate competent federal record.

284.4 State Public Authority Data.

284.4.1 State Public Authority Data shall be reviewed according to the applicable state’s law, public records obligations, open meetings or sunshine rules where applicable, procurement rules, grant rules, privacy rules, cybersecurity rules, public health rules, emergency management rules, public-sector ethics rules, and confidentiality obligations.

284.4.2 State Public Authority Data shall not be generalized across states where state-specific law, agency structure, public records status, privacy obligation, public authority capacity, or publication restriction materially differs.

284.5 District of Columbia and Territorial Public Authority Data.

284.5.1 District of Columbia and territorial Public Authority Data shall be reviewed according to applicable District, territorial, federal-territorial, local, public authority, public records, procurement, grant, privacy, cybersecurity, public health, emergency management, civil rights, language-access, and public-safe publication obligations.

284.5.2 Territorial Public Authority Data shall be handled with attention to island, insular, disaster, infrastructure, public health, public safety, supply chain, language access, accessibility, and community context.

284.6 Tribal and Indigenous Government Data Where Lawfully and Respectfully Handled.

284.6.1 Tribal and Indigenous government data shall be handled only through lawful, respectful, authority-sensitive, protocol-aware, consent-aware, restriction-aware, and safeguards-based pathways.

284.6.2 The Corporation shall not assume that federal, state, local, academic, public authority, nonprofit, sponsor, provider, or individual possession of Tribal or Indigenous government data authorizes use, disclosure, AI processing, mapping, translation, publication, transfer, or reuse.

284.6.3 Where Tribal or Indigenous authority, consent, restriction, attribution, non-attribution, publication permission, mapping permission, AI-use permission, or withdrawal right is uncertain, the Corporation shall apply restrictive handling pending competent clarification.

284.7 Local, County, Municipal, Metropolitan, Utility, Port, Public Health, Emergency Management, Public Safety, Public Works, Telecom, Energy, Water, Food, Cyber, and Infrastructure Data.

284.7.1 Local and sectoral Public Authority Data shall be reviewed for source authority, operational sensitivity, public records status, public-safe mapping, infrastructure sensitivity, cyber sensitivity, privacy, community exposure, procurement sensitivity, utility restrictions, emergency-management restrictions, public health restrictions, and public safety implications.

284.7.2 Utility, port, telecom, energy, water, food, cyber, public works, transportation, public health, emergency management, and infrastructure data shall be classified for misuse risk, operational exposure, cyber risk, public authority restriction, and public-safe publication status before use in reports, dashboards, maps, technical baselines, public authority learning materials, or Nexus-facing materials.

284.8 Capacity Record Requirement.

284.8.1 Public Authority Data shall be accompanied by a Capacity Record where public authority involvement may affect meaning, access, restriction, publication, reliance, public records status, or public-safe interpretation.

284.8.2 Capacity Records shall identify whether the contributor or participant acts in official capacity, observer capacity, learning capacity, regulator-listening capacity, public finance reader capacity, technical staff capacity, procurement capacity, grant capacity, emergency-management capacity, public health capacity, personal capacity, advisory capacity, or another recorded capacity.

284.8.3 Absence of a Capacity Record shall require hold, restriction, clarification, non-public handling, or limitation language where public authority meaning is material.

284.9 Lawful Authority Requirement.

284.9.1 The Corporation shall not receive, process, analyze, publish, share, transfer, embed, train on, or archive Public Authority Data unless lawful authority, permission, contract, grant term, public authority instruction, research authority, data-sharing agreement, license, or other competent basis supports the activity.

284.9.2 Lawful authority shall be specific enough for the proposed use and shall not be inferred from meeting attendance, public authority interest, informal email exchange, verbal encouragement, technical access, platform access, sponsor involvement, provider involvement, or prior collaboration.

284.10 Permitted Use Record.

284.10.1 Each material Public Authority Data activity shall identify permitted uses, including research, evidence classification, methods development, observability analysis, public authority learning, public-safe publication, technical baseline support, controlled-room review, dashboard use, map use, software use, AI-assisted processing, internal review, external sharing, retention, archival, deletion, and correction.

284.10.2 Permitted use records shall be interpreted narrowly where the data is confidential, restricted, rights-bearing, cyber-sensitive, infrastructure-sensitive, health-sensitive, protected knowledge-related, procurement-sensitive, grant-sensitive, or public finance-sensitive.

284.11 Prohibited Use Record.

284.11.1 Public Authority Data records shall identify prohibited uses where applicable, including unauthorized AI training, fine-tuning, embeddings, model improvement, publication, mapping, transfer, external sharing, sponsor benefit, provider development, finance-readiness input, public authority use beyond capacity, commercial reuse, public warning, emergency command, procurement advice, investment advice, recognition, certification, or enterprise execution.

284.11.2 Prohibited uses shall be enforced through access control, contractual restrictions, technical controls, controlled-room rules, review workflows, monitoring, training, and corrective action.

284.12 Public Authority Data Classification.

284.12.1 Public Authority Data shall be classified according to confidentiality, public records status, official status, source authority, sensitivity, privacy, public authority restriction, cyber sensitivity, infrastructure sensitivity, procurement sensitivity, grant sensitivity, public finance sensitivity, health sensitivity, public-safe status, AI-use status, transfer status, retention status, and correction path.

284.12.2 Classification shall be reviewed at intake, before AI processing, before sharing, before publication, before cross-border transfer, after public authority clarification, after correction, after challenge, and after incident.

284.13 Public Authority AI-Use Restrictions.

284.13.1 Public Authority Data shall not be processed through AI systems unless the AI system is approved for the applicable data class, permitted use is recorded, no-training or equivalent restrictions are in place where required, public authority restrictions are satisfied, and human review is required for material outputs.

284.13.2 AI-use restrictions may prohibit prompt submission, model training, fine-tuning, embeddings, retrieval indexing, summarization, translation, classification, anomaly detection, external API processing, cross-border processing, or vendor model improvement.

284.13.3 AI-assisted outputs derived from Public Authority Data shall not imply official public authority interpretation, decision, approval, warning, command, procurement action, or public finance action.

284.14 Public Authority Publication Restrictions.

284.14.1 Public Authority Data shall not be published, summarized, mapped, quoted, visualized, included in dashboards, included in technical baselines, included in public-safe reports, or externally circulated unless publication authority, public-safe status, confidentiality status, public records status, privacy status, and public authority restrictions support the release.

284.14.2 Publication may require aggregation, redaction, masking, delay, controlled annex, non-attribution, limitation language, public authority capacity statement, or non-public handling.

284.15 Public Authority Transfer Restrictions.

284.15.1 Public Authority Data shall not be transferred to third parties, cross-border collaborators, vendors, AI providers, cloud providers, researchers, sponsors, providers, public authorities other than the source authority, national companies, Project SPVs, enterprise actors, GRF-facing recipients, GRA-facing recipients, or Nexus-facing recipients unless lawful authority, permitted use, data rights, security, and public-safe controls support the transfer.

284.15.2 Transfers shall be logged, access-controlled, purpose-limited, and subject to confidentiality, security, deletion, correction, and onward-transfer restrictions.

284.16 Public Records, FOIA, Sunshine, Open Meetings, Public Procurement, Grant, and Confidentiality Considerations Where Applicable.

284.16.1 Public Authority Data shall be reviewed for public records, FOIA, sunshine law, open meeting, procurement, grant, government ethics, lobbying, gifts, confidentiality, public-sector cybersecurity, public-sector privacy, public authority retention, and official-capacity implications where applicable.

284.16.2 The Corporation shall not structure records, meetings, controlled rooms, or data handling to evade lawful public-sector requirements, nor shall it expose restricted public authority information without lawful authority.

284.16.3 Public records considerations may affect recordkeeping, redaction, meeting design, controlled-room procedures, public-safe publication, public authority notices, and retention.

284.17 Public Authority Data Correction, Withdrawal, and Closeout.

284.17.1 Public Authority Data shall be corrected, restricted, superseded, withdrawn, deleted where lawful, or archived where source authority, public authority clarification, corrected records, changed restrictions, privacy issues, cyber issues, public-safe defects, or data quality defects require action.

284.17.2 Closeout shall verify final status, continuing restrictions, access revocation, retention, deletion, publication status, correction status, downstream dependency notification, and public authority notice where required.

284.18 Public Authority Data Records.

284.18.1 The Corporation shall maintain Public Authority Data Records, including definition records, contribution records, federal records, state records, District of Columbia and territorial records, Tribal and Indigenous government data records, local and sectoral data records, Capacity Records, lawful authority records, permitted use records, prohibited use records, classification records, AI-use restriction records, publication restriction records, transfer restriction records, public records / FOIA / sunshine / open meetings / procurement / grant / confidentiality reviews, correction records, withdrawal records, closeout records, and archive records.

---

Section 285. Health-Sensitive, Public Health, Biosecurity, Youth, and Vulnerable-Population Data

285.1 Health-Sensitive Data Purpose.

285.1.1 Health-Sensitive Data shall be governed to protect privacy, dignity, safety, public health trust, research integrity, rights-bearing persons, youth, vulnerable populations, communities, public authorities, health systems, protected knowledge, and public-safe publication.

285.1.2 Health-Sensitive Data includes health, medical, clinical, behavioral health, disability-related, public health, epidemiological, emergency medical, health-system, biosecurity, youth, vulnerable-population, health-adjacent, and health-infrastructure information capable of affecting privacy, rights, stigma, access, safety, or public interpretation.

285.2 Public Health Data.

285.2.1 Public Health Data shall be handled according to applicable public health law, privacy obligations, public authority restrictions, research ethics, de-identification requirements, public-safe publication limits, and community safeguards.

285.2.2 Public Health Data may support public-benefit research, evidence, methods, observability, public authority learning, public-good software, and technical baselines, but shall not be used by the Corporation to issue public health orders, public warnings, emergency commands, medical advice, regulatory determinations, or clinical guidance.

285.3 Biosecurity Data.

285.3.1 Biosecurity Data shall be classified for misuse risk, dual-use risk, public health sensitivity, public authority restriction, national security sensitivity, controlled technology implications, publication risk, and public-safe status.

285.3.2 Biosecurity Data shall not be published, transferred, AI-processed, or included in public-facing technical materials where doing so may increase misuse, expose vulnerability, enable harmful action, or violate law, public authority restriction, research ethics, or safeguards.

285.4 Health System Data.

285.4.1 Health System Data includes information about hospitals, clinics, public health agencies, health infrastructure, medical supply chains, workforce capacity, emergency medical systems, health-system cyber exposure, health-system resilience, and health-service continuity.

285.4.2 Health System Data shall be reviewed for privacy, infrastructure sensitivity, public authority restrictions, operational sensitivity, cyber sensitivity, public-safe mapping, emergency-management implications, and community harm.

285.5 Clinical or Medical Data Where Applicable.

285.5.1 Clinical or Medical Data shall be handled only where lawful authority, consent or authorization where required, ethics review where applicable, privacy protections, security controls, and data minimization support the activity.

285.5.2 The Corporation shall not use Clinical or Medical Data to diagnose, treat, prescribe, provide medical advice, issue clinical recommendations, determine patient care, or substitute for licensed health professionals.

285.6 Youth and Children’s Data.

285.6.1 Youth and Children’s Data shall be collected, used, AI-processed, shared, published, or retained only where lawful, necessary, consented or authorized where required, ethically reviewed where appropriate, and protected through heightened safeguards.

285.6.2 Youth and Children’s Data shall not be included in public-safe maps, dashboards, public reports, AI training, external sharing, or sponsor or provider materials without heightened review and documented public-safe justification.

285.7 Vulnerable-Population Data.

285.7.1 Vulnerable-Population Data includes data concerning persons or communities facing heightened risk due to age, disability, health status, economic insecurity, housing insecurity, immigration status, disaster exposure, environmental exposure, public safety exposure, public health exposure, discrimination, marginalization, dependency, institutionalization, or retaliation risk.

285.7.2 Vulnerable-Population Data shall receive heightened review for privacy, civil rights, accessibility, consent, public-safe publication, AI bias, stigmatization, retaliation, and community safeguards.

285.8 Public Health Emergency Context Data.

285.8.1 Public Health Emergency Context Data shall be handled with urgency-aware safeguards that preserve privacy, public authority boundaries, public-safe communication, security, and correctionability.

285.8.2 Emergency context shall not eliminate lawful basis, consent where required, data minimization, public authority restrictions, research ethics, cyber controls, protected knowledge protections, or public-safe publication review.

285.8.3 The Corporation shall not use emergency context data to issue public health orders, emergency commands, public warnings, or operational directives.

285.9 Health Privacy Review.

285.9.1 Health Privacy Review shall be required for Health-Sensitive Data, Public Health Data, Clinical or Medical Data, youth health data, vulnerable-population health data, and health-system data where privacy or re-identification risk exists.

285.9.2 Health Privacy Review shall assess lawful basis, consent or authorization where required, notice, de-identification, minimum necessary use, access controls, AI-use restrictions, sharing restrictions, retention, deletion, publication, incident response, and rights pathways.

285.10 Research Ethics Review.

285.10.1 Health-sensitive, public health, biosecurity, youth, and vulnerable-population data shall receive research ethics review where required or appropriate.

285.10.2 Ethics review shall assess risk, benefit, consent, coercion, vulnerability, privacy, stigma, discrimination, public authority pressure, emergency context, publication risk, AI use, and correction pathways.

285.11 Minimization and De-Identification.

285.11.1 Health-sensitive, youth, vulnerable-population, and public health data shall be minimized and de-identified, pseudonymized, aggregated, masked, or otherwise protected where feasible and appropriate.

285.11.2 De-identification shall be reviewed for re-identification risk, linkage risk, small-cell risk, geospatial risk, temporal risk, public authority context, community context, and AI inference risk.

285.11.3 De-identified data shall not be treated as risk-free where re-identification, linkage, stigmatization, or community exposure remains plausible.

285.12 Access Restriction.

285.12.1 Access to health-sensitive, public health, biosecurity, youth, and vulnerable-population data shall be limited to authorized persons with a documented need to know.

285.12.2 Access may require controlled-room handling, no-download access, logging, special approvals, confidentiality commitments, ethics training, security review, and periodic access review.

285.13 AI-Use Restriction.

285.13.1 Health-sensitive, youth, vulnerable-population, public health, clinical, medical, and biosecurity data shall not be processed through AI systems unless the system is approved for the data class, AI use is lawful and permitted, no-training restrictions are in place where required, and human review is required for material outputs.

285.13.2 AI use shall be reviewed for bias, hallucination, re-identification, unauthorized inference, health misinformation, clinical overclaim, public authority overclaim, and publication risk.

285.14 Publication Restriction.

285.14.1 Health-sensitive, public health, biosecurity, youth, and vulnerable-population data shall not be published except in public-safe form approved through privacy, ethics, public authority, safeguards, and publication review.

285.14.2 Publication shall not reveal personal information, small-cell information, sensitive locations, clinical information, vulnerable-population identity, biosecurity vulnerability, emergency-system weakness, or stigmatizing community information beyond lawful and approved scope.

285.15 Public-Safe Aggregation.

285.15.1 Public-safe aggregation shall be used where publication, dashboarding, mapping, or reporting of health-sensitive or vulnerable-population data could create identification, stigma, retaliation, panic, public authority confusion, or misuse risk.

285.15.2 Aggregation shall be designed to preserve meaning without exposing protected persons, communities, or sensitive systems.

285.16 No Clinical Advice, Public Health Order, Emergency Command, or Public Warning by Data Use.

285.16.1 No use of health-sensitive, public health, biosecurity, youth, vulnerable-population, clinical, medical, or health-system data shall be represented as clinical advice, medical diagnosis, public health order, emergency command, official public warning, regulatory determination, treatment instruction, public safety order, or substitute for licensed professional or public authority action.

285.16.2 Outputs shall include limitation language where necessary to prevent such reliance.

285.17 Health-Sensitive Data Incident Escalation.

285.17.1 Suspected or confirmed incidents involving health-sensitive, public health, biosecurity, youth, vulnerable-population, clinical, medical, or health-system data shall be escalated promptly.

285.17.2 Incident escalation may require access restriction, system isolation, evidence preservation, privacy review, legal review, public authority notice where required, affected-person notice where required, funder notice where required, public-safe correction, data deletion where appropriate, and recurrence prevention.

285.18 Health-Sensitive Data Records.

285.18.1 The Corporation shall maintain Health-Sensitive Data Records, including public health data records, biosecurity data records, health-system data records, clinical or medical data records where applicable, youth and children’s data records, vulnerable-population data records, public health emergency context data records, health privacy reviews, research ethics reviews, minimization records, de-identification records, access restriction records, AI-use restriction records, publication restriction records, public-safe aggregation records, non-execution limitation records, incident escalation records, corrections, deletions, restrictions, and archive records.

---

Section 286. Cyber-Sensitive, Infrastructure-Sensitive, Security-Sensitive, and Controlled Technology Data

286.1 Cyber-Sensitive Data Purpose.

286.1.1 Cyber-Sensitive Data shall be governed to protect systems, credentials, keys, tokens, secrets, repositories, logs, vulnerabilities, incident records, threat intelligence, AI systems, cloud environments, networks, public authority data, public-good software, technical baselines, and institutional operations from misuse, disclosure, manipulation, compromise, or exploitation.

286.1.2 Cyber-Sensitive Data shall be classified, access-controlled, logged, minimized, secured, publication-reviewed, incident-reviewed, and correctionable.

286.2 Infrastructure-Sensitive Data Purpose.

286.2.1 Infrastructure-Sensitive Data shall be governed to protect critical infrastructure, public works, utilities, telecom, energy, water, food, ports, transportation, public health systems, emergency systems, cyber-physical systems, industrial systems, satellite systems, geospatial assets, AI-RAN / O-RAN systems, digital twins, and other mission-critical systems from harm, exposure, misuse, or exploitation.

286.2.2 Infrastructure-Sensitive Data shall be reviewed for public-safe mapping, operational exposure, cyber risk, physical security risk, public authority restrictions, insurance sensitivity, finance-boundary risk, community exposure, and national security sensitivity.

286.3 Security-Sensitive Data Purpose.

286.3.1 Security-Sensitive Data includes information that could compromise confidentiality, integrity, availability, safety, resilience, access control, physical security, cyber defense, operational continuity, or public authority trust if disclosed or misused.

286.3.2 Security-Sensitive Data shall be handled under need-to-know, access restriction, secure storage, secure transmission, logging, monitoring, redaction, and incident-response controls.

286.4 Controlled Technology Data Purpose.

286.4.1 Controlled Technology Data shall be governed to comply with export-control, sanctions, controlled-technology, national security sensitivity, contractual, IP, public authority, cyber, research ethics, and public-safe publication constraints.

286.4.2 Controlled Technology Data may include technical data, software, encryption information, AI models, cybersecurity methods, telecom methods, AI-RAN / O-RAN materials, DePIN or DLT materials, geospatial systems, digital twin systems, robotics, drones, advanced manufacturing, semiconductors, industrial control systems, and other controlled or potentially controlled materials.

286.5 Vulnerability Data.

286.5.1 Vulnerability Data includes information about software, hardware, model, data, network, repository, cloud, identity, AI, cyber-physical, infrastructure, telecom, or operational weaknesses.

286.5.2 Vulnerability Data shall be handled through restricted access, secure storage, need-to-know review, coordinated disclosure where applicable, public-safe redaction, remediation tracking, and incident escalation where needed.

286.6 Exploit, Indicator, Threat-Intelligence, and Incident Data.

286.6.1 Exploit, indicator, threat-intelligence, and incident data shall be classified according to sensitivity, source, permission, reliability, operational use, public authority restrictions, legal privilege, confidentiality, and public-safe status.

286.6.2 Such data shall not be publicly released or externally transferred where release would enable harm, expose defenders, compromise investigations, violate restrictions, or weaken public authority trust.

286.7 Critical Infrastructure Location, Dependency, Configuration, and Weakness Data.

286.7.1 Data identifying critical infrastructure location, dependency, configuration, operational weakness, outage pattern, recovery capacity, degraded-mode exposure, redundancy gap, supply-chain dependency, or security gap shall receive heightened review before processing, mapping, dashboarding, publication, transfer, or AI use.

286.7.2 Public-safe treatment may require aggregation, masking, coordinate reduction, time delay, omission of technical detail, controlled annexes, no-download rooms, or non-public handling.

286.8 AI-RAN, O-RAN, Telecom, Energy, Water, Port, Utility, Public Works, Transportation, Cyber-Physical, Sensor, Robotics, Drone, Satellite, Geospatial, and Digital Twin Sensitive Data.

286.8.1 Sensitive data relating to AI-RAN, O-RAN, telecom, energy, water, ports, utilities, public works, transportation, cyber-physical systems, sensors, robotics, drones, satellites, geospatial systems, and digital twins shall be reviewed for operational exposure, cyber exposure, public authority restrictions, public-safe mapping, export-control, sanctions, provider sensitivity, community exposure, and misuse risk.

286.8.2 Such data shall not be used to certify systems, approve procurement, direct operations, issue public warnings, command emergency response, guarantee resilience, or provide regulated telecom, utility, cyber, or infrastructure determinations.

286.9 Export-Control and Sanctions Review.

286.9.1 Export-Control and Sanctions Review shall be required where data, software, technical information, models, encryption, cybersecurity methods, AI-RAN / O-RAN materials, telecom materials, geospatial materials, digital twin materials, controlled technologies, restricted parties, restricted jurisdictions, prohibited end users, or prohibited end uses may be involved.

286.9.2 Data subject to unresolved export-control or sanctions concern shall be held, restricted, quarantined, denied, deleted where appropriate, returned, licensed where required, or escalated before access, transfer, publication, AI processing, or repository release.

286.10 National Security Sensitivity Review.

286.10.1 National Security Sensitivity Review shall assess whether data may implicate defense, critical infrastructure, cyber operations, controlled technology, geospatial sensitivity, public authority security, foreign access, public safety, or other national security-sensitive considerations.

286.10.2 National Security Sensitivity Review may require counsel review, access restriction, publication delay, controlled-room handling, public authority consultation where appropriate, export-control review, sanctions review, or refusal.

286.11 Access Restriction.

286.11.1 Cyber-sensitive, infrastructure-sensitive, security-sensitive, and controlled technology data shall be access-restricted according to sensitivity, lawful authority, need-to-know, role, jurisdiction, export-control status, public authority restriction, contractual obligation, and public-safe status.

286.11.2 Access may require multi-factor authentication, privileged access controls, no-download access, secure workstation requirements, controlled-room access, logging, monitoring, and periodic review.

286.12 Need-to-Know Handling.

286.12.1 Need-to-know handling shall limit access to persons whose role, task, authority, training, and approved purpose require access.

286.12.2 Curiosity, seniority, sponsor relationship, provider relationship, public authority proximity, Board status without need, technical interest, or publication role shall not by itself establish need to know.

286.13 Secure Storage and Transmission.

286.13.1 Cyber-sensitive, infrastructure-sensitive, security-sensitive, and controlled technology data shall be stored and transmitted through secure systems appropriate to classification and risk.

286.13.2 Secure storage and transmission may require encryption, access logging, secure file transfer, repository controls, key management, segmentation, backup controls, data loss prevention, endpoint controls, and prohibition on unapproved email, messaging, personal devices, public links, or unapproved AI tools.

286.14 Public-Safe Redaction.

286.14.1 Public-safe redaction shall remove or obscure details that could enable harm, compromise systems, expose protected knowledge, reveal sensitive locations, disclose credentials, identify vulnerable persons, reveal operational weaknesses, or create public authority confusion.

286.14.2 Redaction shall preserve integrity and shall not create misleading summaries, false assurance, or false absence of risk.

286.15 Coordinated Vulnerability Disclosure Where Applicable.

286.15.1 Where the Corporation identifies or receives Vulnerability Data, it may follow coordinated vulnerability disclosure practices appropriate to the source, affected system, legal constraints, public authority context, researcher safety, confidentiality, exploitability, severity, and public-safe communication.

286.15.2 Coordinated disclosure shall not be represented as certification, managed security service, public authority cyber command, emergency command, cyber insurance determination, or legal compliance approval.

286.16 No Public Release of Sensitive Operational Weakness Without Review.

286.16.1 The Corporation shall not publicly release sensitive operational weakness, exploit detail, infrastructure dependency, cyber vulnerability, system configuration, critical location, emergency-system weakness, or other sensitive operational data without competent public-safe, legal, cyber, public authority, safeguards, and publication review.

286.16.2 Where public interest supports disclosure, the Corporation shall use the least harmful lawful form, including aggregation, redaction, controlled notice, delayed disclosure, or referral.

286.17 Cyber-Sensitive and Infrastructure-Sensitive Data Records.

286.17.1 The Corporation shall maintain Cyber-Sensitive and Infrastructure-Sensitive Data Records, including cyber-sensitive data records, infrastructure-sensitive data records, security-sensitive data records, controlled technology data records, vulnerability data records, exploit / indicator / threat-intelligence / incident data records, critical infrastructure location / dependency / configuration / weakness records, AI-RAN / O-RAN / telecom / energy / water / port / utility / public works / transportation / cyber-physical / sensor / robotics / drone / satellite / geospatial / digital twin sensitive data records, export-control reviews, sanctions reviews, national security sensitivity reviews, access restriction records, need-to-know records, secure storage and transmission records, redaction records, coordinated vulnerability disclosure records, public-release review records, corrections, restrictions, incidents, and archive records.

---

Section 287. Community-Protected Data, Tribal / Indigenous Knowledge, Local Knowledge, Territorial Knowledge, Cultural Knowledge, Environmental Knowledge, and Protected Knowledge

287.1 Community-Protected Data Purpose.

287.1.1 Community-Protected Data shall be governed to prevent extraction, exposure, stigmatization, retaliation, misattribution, cultural harm, environmental harm, public authority misuse, sponsor misuse, provider misuse, AI misuse, public-safe mapping harm, and loss of trust.

287.1.2 Community-Protected Data may include community observations, local resilience knowledge, vulnerability information, lived experience, grievance information, public health concerns, environmental information, cultural information, territorial knowledge, protected participation information, and other data whose misuse could harm a community or rights-bearing group.

287.2 Tribal Sovereignty Respect.

287.2.1 The Corporation shall respect Tribal sovereignty and shall not treat Tribal-related data or knowledge as ordinary public information merely because it is technically accessible, publicly visible, held by another government, included in an academic source, or disclosed by an individual participant.

287.2.2 Tribal-related data shall be handled through authority-sensitive, protocol-aware, permission-based, and safeguards-based pathways where applicable.

287.3 Indigenous Data Safeguards.

287.3.1 Indigenous Data shall be governed with heightened safeguards reflecting Indigenous governance, collective rights, cultural context, territorial context, knowledge stewardship, consent, non-consent, attribution, non-attribution, restriction, withdrawal, correction, and benefit considerations where applicable.

287.3.2 Indigenous Data shall not be reused for AI training, public mapping, finance-facing materials, sponsor benefit, provider development, public authority sharing, or external transfer without lawful authority and appropriate safeguards review.

287.4 Indigenous Knowledge Safeguards.

287.4.1 Indigenous Knowledge shall not be extracted, generalized, translated, summarized, embedded, trained on, mapped, published, commercialized, or transferred without lawful authority, respectful permission, protocol compliance where applicable, public-safe review, and protected knowledge safeguards.

287.4.2 Indigenous Knowledge safeguards shall apply even where the knowledge has been partially published, informally shared, observed, or referenced in public authority materials.

287.5 Local Knowledge Safeguards.

287.5.1 Local Knowledge shall be handled with respect for context, attribution, non-attribution, consent, public-safe mapping, local vulnerability, community exposure, and correction.

287.5.2 Local Knowledge shall not be generalized into universal claims, used to stigmatize communities, or converted into public authority, finance, procurement, sponsor, provider, or enterprise claims without proper review and limitation.

287.6 Territorial Knowledge Safeguards.

287.6.1 Territorial Knowledge shall be handled with attention to island, insular, border, coastal, Arctic, rural, remote, and jurisdiction-specific contexts, including language access, infrastructure dependence, environmental exposure, disaster exposure, local law, and community authority.

287.6.2 Territorial Knowledge shall not be published or mapped in a manner that exposes vulnerability, infrastructure dependency, cultural sites, ecological assets, or community risk beyond public-safe scope.

287.7 Cultural Knowledge Safeguards.

287.7.1 Cultural Knowledge shall be protected against unauthorized extraction, misattribution, decontextualization, translation harm, commercialization, AI training, public mapping, and publication without appropriate authority and safeguards.

287.7.2 Cultural Knowledge restrictions may require non-attribution, restricted access, sealed records, controlled-room review, publication refusal, or withdrawal.

287.8 Environmental Knowledge Safeguards.

287.8.1 Environmental Knowledge, including ecological, biodiversity, habitat, water, land, species, climate, and environmental vulnerability information, shall be reviewed for protected knowledge, sensitive location, community, Indigenous, public authority, environmental harm, and public-safe mapping concerns.

287.8.2 Environmental Knowledge shall not be released in a manner that enables exploitation, poaching, resource extraction, ecological harm, cultural harm, or community exposure.

287.9 Protected Knowledge Definition.

287.9.1 Protected Knowledge means knowledge, data, information, observations, practices, locations, relationships, records, or meaning that require heightened protection due to Tribal, Indigenous, local, territorial, cultural, environmental, ecological, sacred, vulnerability-related, community-protected, security, privacy, dignity, rights, or public-safe considerations.

287.9.2 Protected Knowledge may be individual, collective, institutional, community-held, place-based, relational, unpublished, partially published, orally shared, inferred, mapped, observed, or embedded in data.

287.10 Community Protocols.

287.10.1 The Corporation shall respect applicable community protocols governing consent, participation, attribution, non-attribution, mapping, publication, AI use, data sharing, correction, withdrawal, grievance, and remedy.

287.10.2 Community protocols shall be recorded where appropriate and shall not be overridden by sponsor preference, provider convenience, public authority interest, publication deadline, AI-tool capability, or institutional convenience.

287.11 Consent, Non-Consent, Attribution, Withdrawal, Restriction, and Correction Pathways.

287.11.1 Community-Protected and Protected Knowledge Data shall include pathways for consent, non-consent, attribution, non-attribution, withdrawal, restriction, correction, grievance, and remedy where applicable.

287.11.2 Withdrawal or restriction requests shall receive good-faith review and may require immediate hold, access restriction, publication withdrawal, public-safe correction, or controlled correction.

287.12 Access Restriction.

287.12.1 Community-Protected and Protected Knowledge Data shall be access-restricted according to authority, consent, sensitivity, public-safe status, safeguards, role, need-to-know, and correction requirements.

287.12.2 Access may require controlled-room handling, non-download access, non-attribution, redaction, special reviewer approval, community or Tribal review where applicable, and logging.

287.13 AI-Use Restriction.

287.13.1 Community-Protected, Tribal / Indigenous, local, territorial, cultural, environmental, and Protected Knowledge Data shall not be processed through AI systems unless lawful authority, protocol permission where applicable, consent or permission where required, protected knowledge review, data rights, public-safe status, and no-training restrictions support the use.

287.13.2 AI use shall be prohibited where it risks unauthorized extraction, semantic drift, decontextualization, re-identification, sensitive mapping, cultural harm, ecological harm, or unauthorized generalization.

287.14 Public-Safe Mapping Restriction.

287.14.1 Public-safe mapping of Community-Protected or Protected Knowledge Data shall require safeguards review before release.

287.14.2 Mapping controls may include aggregation, coordinate reduction, masking, blurring, omission, delay, non-attribution, controlled legends, no-download access, or non-public handling.

287.15 No Extraction for Sponsor, Provider, Donor, Funder, Host, Public Authority, or Institutional Convenience.

287.15.1 Community-Protected and Protected Knowledge Data shall not be collected, processed, translated, summarized, mapped, AI-processed, published, shared, or transferred for sponsor storytelling, provider development, donor reporting, funder pressure, host convenience, public authority persuasion, institutional visibility, capital-reader interest, or enterprise advantage.

287.15.2 The Corporation shall reject or re-scope activities where protected knowledge would be used as reputational, financial, procurement, public authority, or commercial leverage.

287.16 Safeguards Review Before Data Sharing, Publication, Mapping, AI Use, or External Transfer.

287.16.1 Safeguards review shall be required before sharing, publishing, mapping, AI processing, embedding, external transfer, controlled-room release, public authority sharing, sponsor sharing, provider sharing, or Nexus-facing routing of Community-Protected or Protected Knowledge Data.

287.16.2 Safeguards review shall assess authority, consent, non-consent, attribution, restriction, withdrawal, community impact, retaliation risk, public-safe status, AI risk, mapping risk, public authority risk, and correction path.

287.17 Grievance and Remedy Pathway.

287.17.1 The Corporation shall maintain grievance and remedy pathways for concerns relating to Community-Protected Data, Indigenous Data, Indigenous Knowledge, local knowledge, territorial knowledge, cultural knowledge, environmental knowledge, public-safe mapping, protected knowledge, consent, attribution, withdrawal, restriction, AI use, publication, or transfer.

287.17.2 Remedies may include access restriction, correction, withdrawal, deletion where lawful, non-attribution, apology where appropriate, public-safe clarification, controlled clarification, community review, Tribal or Indigenous review where applicable, contract remedy, staff training, or recurrence prevention.

287.18 Community-Protected and Protected Knowledge Records.

287.18.1 The Corporation shall maintain Community-Protected and Protected Knowledge Records, including purpose records, Tribal sovereignty respect records, Indigenous data safeguard records, Indigenous knowledge safeguard records, local knowledge safeguard records, territorial knowledge safeguard records, cultural knowledge safeguard records, environmental knowledge safeguard records, Protected Knowledge classifications, community protocol records, consent records, non-consent records, attribution records, withdrawal records, restriction records, correction records, access restriction records, AI-use restriction records, public-safe mapping restriction records, no-extraction records, safeguards reviews, grievance records, remedy records, deletions, withdrawals, and archive records.

---

Section 288. Cross-Border Data Transfers, Sovereign Data Zones, Compute-to-Data, and North America Data Interfaces

288.1 Cross-Border Transfer Purpose.

288.1.1 Cross-border transfer controls shall ensure that data, records, evidence, datasets, model outputs, inference records, compute records, public authority data, personal information, health-sensitive data, cyber-sensitive data, infrastructure-sensitive data, controlled technology, and protected knowledge are transferred, accessed, processed, or stored across jurisdictions only where lawful, authorized, secure, necessary, public-benefit aligned, and safeguarded.

288.1.2 Cross-border transfer includes physical transfer, electronic transfer, cloud storage, remote access, API access, AI processing, inference logging, embedding, repository access, controlled-room access, publication, dashboarding, model transfer, compute transfer, and personnel access from another jurisdiction.

288.2 United States Data Localization Review.

288.2.1 United States Data Localization Review shall determine whether data must or should remain within United States systems, United States legal control, United States cloud regions, United States controlled rooms, United States repositories, or United States personnel access boundaries.

288.2.2 Localization review shall consider law, contract, public authority restrictions, privacy, cyber risk, national security sensitivity, controlled technology, public-safe publication, donor or funder restrictions, data subject rights, and institutional resilience.

288.3 State and Territorial Data Localization Review.

288.3.1 State and territorial data localization review shall determine whether data is subject to state-specific, territory-specific, local, public authority, health, education, employment, public records, procurement, cybersecurity, privacy, grant, or contractual restrictions affecting location, access, storage, processing, or transfer.

288.3.2 State and territorial restrictions shall not be overridden by national convenience, cloud defaults, AI provider settings, collaborator preference, or cross-border program design.

288.4 Tribal and Indigenous Data Sovereignty Review.

288.4.1 Tribal and Indigenous Data Sovereignty Review shall be required where data concerns Tribal Nations, Indigenous governments, Indigenous communities, Indigenous Data, Indigenous Knowledge, protected knowledge, land-based knowledge, cultural knowledge, environmental knowledge, sacred knowledge, or community-protected information.

288.4.2 Review shall assess authority, protocol, consent, non-consent, attribution, restriction, localization, access, publication, AI use, mapping, transfer, withdrawal, correction, grievance, and remedy.

288.5 Canadian Interface Review.

288.5.1 Canadian Interface Review shall be required where data is transferred to, received from, accessed by, processed in, stored in, or coordinated with GCRI Canada, Canadian public authorities, Canadian institutions, Canadian communities, Canadian providers, Canadian infrastructure actors, or Canadian collaborators.

288.5.2 Canadian Interface Review shall preserve legal separateness, no shared treasury, no agency, no partnership, no joint venture, separate records, separate authority, separate liabilities, privacy compliance, public authority restrictions, Indigenous data safeguards, cross-border transfer controls, and public-safe publication limitations.

288.6 Mexico, Caribbean, Arctic, and North America Interface Review.

288.6.1 Mexico, Caribbean, Arctic, and North America Interface Review shall be required where data is transferred to, received from, accessed by, processed in, stored in, or coordinated across those interfaces.

288.6.2 Review shall address applicable law, public authority capacity, privacy, cybersecurity, controlled technology, sanctions, export-control, public health, infrastructure sensitivity, Indigenous rights, protected knowledge, language access, disaster context, environmental context, and public-safe publication.

288.7 Public Authority Cross-Border Data Review.

288.7.1 Public Authority Data shall not be transferred across borders unless the source authority, applicable law, contract, public authority permission, confidentiality, security, public records implications, and public-safe publication status permit the transfer.

288.7.2 Public authority cross-border transfer shall not imply public authority delegation, treaty action, sovereign decision, public finance approval, procurement approval, emergency command, public warning, or official adoption.

288.8 Personal Information Cross-Border Review.

288.8.1 Personal information and rights-bearing data shall receive cross-border privacy review before transfer, remote access, cloud storage, AI processing, publication, dashboarding, or external sharing across jurisdictions.

288.8.2 Review shall assess notice, consent where required, contractual safeguards, data subject rights, security, processor terms, onward transfer, localization, breach obligations, retention, deletion, and public-safe publication.

288.9 Health, Cyber, Infrastructure, Controlled Technology, and Protected Knowledge Cross-Border Review.

288.9.1 Health-sensitive, cyber-sensitive, infrastructure-sensitive, controlled technology, export-controlled, sanctions-sensitive, and protected knowledge data shall receive heightened cross-border review.

288.9.2 Cross-border transfer may be denied, restricted, localized, aggregated, redacted, processed through compute-to-data, or handled only in a sovereign data zone or restricted processing environment.

288.10 Sanctions, Export-Control, and National Security Sensitivity Review.

288.10.1 Cross-border data activity shall be screened where required or prudent for sanctions, export-control, controlled technology, restricted parties, restricted jurisdictions, prohibited end users, prohibited end uses, anti-terrorism, national security sensitivity, and public authority restrictions.

288.10.2 Unresolved sanctions, export-control, or national security sensitivity shall require hold, denial, restriction, counsel review, public authority consultation where appropriate, or termination.

288.11 Sovereign Data Zone.

288.11.1 A Sovereign Data Zone means a legally, technically, and operationally bounded environment designed to keep data within defined jurisdictional, institutional, cloud, access, processing, or control boundaries.

288.11.2 Sovereign Data Zones may be used for public authority data, sensitive personal information, health-sensitive data, cyber-sensitive data, infrastructure-sensitive data, controlled technology, protected knowledge, or cross-border data requiring localization or heightened control.

288.11.3 Use of the term Sovereign Data Zone shall not imply sovereign authority, public authority delegation, public finance approval, public warning authority, emergency command, certification, or legal compliance guarantee.

288.12 Compute-to-Data.

288.12.1 Compute-to-Data means a controlled method in which approved computation is brought to data within a restricted environment, rather than transferring the underlying data to the compute user or external system.

288.12.2 Compute-to-Data may be used to preserve privacy, localization, public authority restrictions, protected knowledge, cyber security, infrastructure sensitivity, and data rights.

288.12.3 Compute-to-Data outputs shall be reviewed before release to ensure they do not reveal restricted data, enable re-identification, expose protected knowledge, or violate export-control, sanctions, public authority, or public-safe restrictions.

288.13 No-Download Room.

288.13.1 A No-Download Room means a controlled environment in which authorized users may view, query, analyze, or review data without downloading, copying, exporting, embedding, photographing, scraping, or transferring the underlying material.

288.13.2 No-Download Rooms shall include access controls, logging, monitoring, user obligations, output review, incident reporting, and access revocation rules.

288.14 Restricted Processing Environment.

288.14.1 A Restricted Processing Environment means an approved technical, contractual, and governance environment for processing data subject to heightened restrictions.

288.14.2 Restricted Processing Environments shall define permitted users, permitted data, permitted tools, AI-use restrictions, export restrictions, network restrictions, output review, logging, monitoring, retention, deletion, and incident response.

288.15 Contractual Transfer Controls.

288.15.1 Cross-border data transfers shall be governed by contractual controls where appropriate, including confidentiality, permitted use, prohibited use, data security, AI-use restrictions, sub-processing, onward transfer, audit, breach notice, localization, deletion, return, correction, public authority restrictions, protected knowledge restrictions, and termination.

288.15.2 Contractual controls shall not be used to authorize transfer where law, public authority restriction, protected knowledge protocol, export-control, sanctions, or safeguards prohibit transfer.

288.16 Cross-Border Transfer Denial, Restriction, Redaction, Aggregation, or Re-Scoping.

288.16.1 The Corporation shall deny, restrict, redact, aggregate, localize, anonymize, pseudonymize, summarize, re-scope, or terminate cross-border transfer where lawful authority is absent, permission is unclear, residual risk is unacceptable, protected knowledge is at risk, public authority restrictions prohibit transfer, privacy risk is excessive, cyber risk is unresolved, or export-control / sanctions concerns remain.

288.16.2 Transfer denial or restriction shall be recorded with rationale and correction pathway where applicable.

288.17 Cross-Border Transfer Records.

288.17.1 The Corporation shall maintain Cross-Border Transfer Records, including transfer purpose records, United States localization reviews, state and territorial localization reviews, Tribal and Indigenous data sovereignty reviews, Canadian Interface Reviews, Mexico / Caribbean / Arctic / North America Interface Reviews, public authority cross-border reviews, personal information cross-border reviews, health / cyber / infrastructure / controlled technology / protected knowledge cross-border reviews, sanctions reviews, export-control reviews, national security sensitivity reviews, Sovereign Data Zone records, Compute-to-Data records, No-Download Room records, Restricted Processing Environment records, contractual transfer controls, transfer denials, restrictions, redactions, aggregations, re-scoping records, corrections, deletions, and archive records.

---

Section 289. Data Access Controls, Identity, Least Privilege, Logging, Monitoring, and Access Reviews

289.1 Access Control Purpose.

289.1.1 Access controls shall protect the Corporation’s data, records, systems, repositories, controlled rooms, AI tools, models, inference records, public-good software, technical baselines, dashboards, maps, financial systems, public authority data, rights-bearing data, health-sensitive data, cyber-sensitive data, infrastructure-sensitive data, community-protected data, protected knowledge, and confidential materials.

289.1.2 Access controls shall preserve confidentiality, integrity, availability, privacy, public authority trust, research integrity, evidence integrity, cyber resilience, least privilege, segregation of duties, public-safe publication, and correctionability.

289.2 Identity Verification.

289.2.1 Access shall be granted only to persons, roles, service accounts, systems, or processors whose identity has been verified through procedures appropriate to risk.

289.2.2 Identity verification shall be required for employees, contractors, fellows, advisors, volunteers, directors, officers, researchers, reviewers, collaborators, public authority participants, providers, vendors, controlled-room users, data processors, and system administrators.

289.2.3 Shared accounts, unidentified accounts, informal access, unverified external accounts, and personal credentials shall be prohibited for material systems unless a documented exception is approved with compensating controls.

289.3 Role-Based Access Control.

289.3.1 The Corporation shall use role-based access control where appropriate to assign access according to approved roles, responsibilities, authority, training, data class, system need, and purpose.

289.3.2 Role definitions shall be reviewed periodically to ensure that access remains aligned with current duties and does not create excessive privilege, conflict, data exposure, or segregation-of-duties failure.

289.4 Attribute-Based Access Control Where Appropriate.

289.4.1 Attribute-based access control may be used where access depends on data classification, jurisdiction, public authority capacity, export-control status, project, role, time, location, device posture, training status, conflict status, confidentiality agreement, or controlled-room eligibility.

289.4.2 Attribute-based rules shall be documented, tested, reviewed, and corrected where they create improper access or improper denial.

289.5 Least Privilege.

289.5.1 Least privilege shall require that users receive only the minimum access necessary for approved purpose, task, role, duration, and data class.

289.5.2 Least privilege shall apply to repositories, datasets, dashboards, maps, AI tools, cloud systems, financial systems, collaboration tools, controlled rooms, public authority materials, protected knowledge, and administrative systems.

289.6 Need-to-Know Access.

289.6.1 Need-to-know access shall be required for restricted, confidential, cyber-sensitive, infrastructure-sensitive, health-sensitive, rights-bearing, public authority, controlled technology, community-protected, and protected knowledge data.

289.6.2 Need to know shall be determined by approved task and authority, not curiosity, seniority, sponsor status, provider status, public authority proximity, donor status, Board status without need, or general institutional interest.

289.7 Time-Limited Access.

289.7.1 Access shall be time-limited where access is temporary, project-based, review-based, incident-based, emergency-based, controlled-room-based, contractor-based, public authority-based, or otherwise not continuing.

289.7.2 Time-limited access shall expire automatically where feasible or be reviewed before renewal.

289.8 Conditional Access.

289.8.1 Conditional access may require approved device, network, location, authentication method, training completion, confidentiality agreement, conflict disclosure, export-control clearance where applicable, controlled-room authorization, or security posture before access is granted.

289.8.2 Conditional access may be tightened or suspended during incidents, elevated threat conditions, suspected compromise, policy violations, or sensitive review periods.

289.9 Multi-Factor Authentication.

289.9.1 Multi-factor authentication shall be required for material systems, privileged accounts, repositories, cloud systems, financial systems, AI systems, controlled rooms, public authority data environments, sensitive data environments, and remote access unless a documented exception with compensating controls is approved.

289.9.2 Weak, shared, reused, or unmanaged authentication methods shall be replaced where risk warrants.

289.10 Privileged Access Management.

289.10.1 Privileged access shall be limited, documented, approved, monitored, reviewed, and revoked when no longer required.

289.10.2 Privileged users shall not use administrative privileges for ordinary work where separate non-privileged access is feasible.

289.10.3 Privileged access to public authority data, protected knowledge, cyber-sensitive data, infrastructure-sensitive data, financial systems, repositories, models, inference records, or controlled rooms shall receive heightened review.

289.11 Segregation of Duties.

289.11.1 Access controls shall preserve segregation of duties in research review, evidence classification, publication approval, financial systems, procurement, contracting, payment authorization, repository release, software deployment, model approval, data deletion, incident response, and controlled-room administration.

289.11.2 No person should have unchecked authority to create, approve, publish, delete, pay, release, classify, correct, or close records where such combination creates material integrity risk.

289.12 Access Logging.

289.12.1 Access to material systems and sensitive data shall be logged to the extent technically feasible and proportionate to risk.

289.12.2 Logs may include user identity, date, time, system, data accessed, action taken, export, download, change, deletion, administrative action, AI processing, controlled-room access, and failed access attempt.

289.12.3 Logs shall be protected against unauthorized alteration, deletion, and excessive access.

289.13 Monitoring.

289.13.1 Access activity shall be monitored proportionate to risk, including unusual access, excessive download, after-hours access, failed login patterns, privileged action, repository change, dataset export, AI-tool use, controlled-room activity, and access to protected knowledge.

289.13.2 Monitoring shall be used for security, compliance, incident detection, and integrity protection and shall not be used for retaliation, improper surveillance, discrimination, or unauthorized personnel action.

289.14 Periodic Access Review.

289.14.1 The Corporation shall conduct periodic access reviews for material systems and sensitive data environments.

289.14.2 Access reviews shall verify that users, roles, privileges, service accounts, external collaborators, public authority participants, vendors, providers, contractors, and privileged accounts remain authorized and necessary.

289.14.3 Access reviews shall result in continuation, modification, reduction, suspension, or revocation of access.

289.15 Access Revocation.

289.15.1 Access shall be revoked promptly when a person’s role ends, contract ends, project ends, authorization expires, conflict requires restriction, incident requires suspension, legal restriction applies, public authority permission ends, consent is withdrawn where applicable, or access is no longer needed.

289.15.2 Revocation shall include user accounts, service accounts, repository permissions, cloud access, AI-tool access, controlled-room access, financial system access, shared drives, collaboration tools, devices, tokens, keys, certificates, and physical access where applicable.

289.16 Emergency Access.

289.16.1 Emergency access may be granted where necessary to protect data, systems, public authority trust, public safety learning materials, cyber security, protected knowledge, financial systems, evidence integrity, or institutional continuity.

289.16.2 Emergency access shall be time-limited, logged, reviewed after use, and revoked when emergency need ends.

289.17 Break-Glass Access.

289.17.1 Break-glass access means extraordinary privileged access used only when ordinary access processes are unavailable or insufficient to address urgent risk.

289.17.2 Break-glass access shall require the highest feasible authentication, logging, notice to competent authority, post-use review, justification, and correction of any excessive access.

289.17.3 Break-glass access shall not be used to bypass lawful restrictions, access protected knowledge unnecessarily, suppress evidence, alter records, approve payments, publish materials, or avoid oversight.

289.18 Access Control Records.

289.18.1 The Corporation shall maintain Access Control Records, including access control policies, identity verification records, role-based access records, attribute-based access rules, least-privilege records, need-to-know determinations, time-limited access records, conditional access rules, multi-factor authentication records, privileged access records, segregation-of-duties records, access logs, monitoring records, periodic access reviews, access revocations, emergency access records, break-glass access records, incidents, corrections, and archive records.

Section 290. Data Quality, Accuracy, Integrity, Provenance, Versioning, and Correction

290.1 Data Quality Purpose.

290.1.1 Data quality governance shall ensure that data used, received, generated, derived, processed, classified, analyzed, published, routed, retained, deleted, archived, or relied upon by the Corporation is sufficiently accurate, complete, timely, relevant, reliable, traceable, permissioned, versioned, integrity-protected, limitation-aware, and correctionable for the purpose for which it is used.

290.1.2 Data quality shall support research integrity, evidence integrity, methods integrity, observability integrity, AI governance, verifiable compute, verifiable intelligence, public authority trust, public-safe publication, public-good technical asset stewardship, community safeguards, protected knowledge protection, and validity-by-record.

290.1.3 Data quality controls shall apply to raw data, processed data, derived data, synthetic data, datasets, documents, telemetry, logs, sensor readings, AI-RAN and O-RAN signals, DePIN records, blockchain / DLT records, cyber logs, geospatial data, Earth observation data, digital twin outputs, public authority data, community-protected data, health-sensitive data, rights-bearing data, protected knowledge, model outputs, embeddings, inference records, dashboards, maps, repositories, technical baselines, and public-safe reports.

290.2 Accuracy Review.

290.2.1 Accuracy Review shall determine whether data accurately reflects the source record, observation, measurement, statement, signal, event, system state, computation, model output, contributor input, public authority record, community input, or protected knowledge input from which it derives.

290.2.2 Accuracy Review shall include review for transcription error, extraction error, translation error, unit error, coordinate error, timestamp error, version error, metadata error, identity error, source error, classification error, calculation error, AI summarization error, model inference error, transformation error, and context error.

290.2.3 Data that is inaccurate, materially misleading, unsupported, or unverifiable for its intended use shall be corrected, restricted, labeled, excluded, superseded, withdrawn, or treated as exploratory only.

290.3 Completeness Review.

290.3.1 Completeness Review shall assess whether a dataset, data record, evidence record, log, telemetry stream, dashboard, map, model output, baseline input, or public-safe report contains the fields, source context, metadata, permissions, restrictions, timestamps, versions, quality indicators, limitations, and correction path necessary for the intended use.

290.3.2 Incomplete data shall be marked and shall not be silently completed through assumption, interpolation, AI generation, model inference, synthetic substitution, sponsor-provided supplementation, provider-provided supplementation, or public authority implication unless the method is documented, lawful, limitation-aware, and reviewed.

290.3.3 Completeness deficiencies shall be disclosed where they materially affect interpretation, public-safe publication, public authority learning, technical baselines, dashboards, maps, AI outputs, Docket inputs, Grid inputs, GRF-facing inputs, or GRA-facing inputs.

290.4 Timeliness Review.

290.4.1 Timeliness Review shall determine whether data is current enough for the intended use and whether changed facts, changed law, changed public authority status, changed technology, changed infrastructure condition, changed provider status, changed community condition, changed risk context, or superseding records affect reliability.

290.4.2 Data may be classified as current, time-bounded, stale, superseded, archived, historical, corrected, withdrawn, or restricted based on timeliness.

290.4.3 Stale data shall not be used for high-reliance public-safe publication, public authority learning, dashboards, maps, technical baselines, AI evaluation, verifiable compute outputs, or Nexus-facing materials unless age and limitations are disclosed and competent authority approves such use.

290.5 Relevance Review.

290.5.1 Relevance Review shall determine whether data is materially connected to the purpose, research question, evidence claim, method output, observability output, AI use case, technical baseline, public authority learning material, safeguards review, publication, dashboard, map, or correction matter for which it is proposed.

290.5.2 Data shall not be treated as relevant merely because it is available, voluminous, machine-readable, recently generated, visually persuasive, sponsor-submitted, provider-submitted, public authority-adjacent, ledger-recorded, AI-generated, or technically convenient.

290.5.3 Data of limited relevance may be excluded, separately labeled, used only for background context, used only for exploratory analysis, or restricted from public-safe outputs.

290.6 Integrity Review.

290.6.1 Integrity Review shall determine whether data has been preserved from unauthorized alteration, deletion, substitution, corruption, tampering, spoofing, poisoning, unauthorized enrichment, unauthorized de-identification, unauthorized re-identification, unauthorized linkage, or unauthorized AI transformation.

290.6.2 Integrity Review shall consider checksums, hashes, signatures, audit logs, repository history, access logs, custody records, system logs, change logs, validation records, source confirmations, cryptographic receipts where appropriate, and other integrity controls.

290.6.3 Data with unresolved integrity concerns shall be quarantined, access-restricted, revalidated, corrected, superseded, withdrawn, or excluded from reliance.

290.7 Provenance Review.

290.7.1 Provenance Review shall identify how data was created, collected, generated, received, extracted, transformed, processed, enriched, combined, analyzed, stored, transferred, published, corrected, superseded, withdrawn, archived, or deleted.

290.7.2 Provenance records shall identify source systems, contributors, tools, methods, AI systems where applicable, processing steps, transformation logic, versions, timestamps, custody changes, permission changes, and known limitations.

290.7.3 Data lacking sufficient provenance shall not be used for high-reliance evidence, public authority learning, public-safe publication, technical baselines, AI evaluation, or Nexus-facing inputs unless competent authority records why such use is necessary and how limitations will prevent misleading reliance.

290.8 Source Lineage Review.

290.8.1 Source Lineage Review shall preserve traceability from data to its originating source, source chain, contributor, system, sensor, public authority, community, protected knowledge source, dataset, document, model, ledger, repository, or compute environment.

290.8.2 Source lineage shall be documented sufficiently to support review, challenge, correction, supersession, withdrawal, archive, rights management, public-safe publication, and downstream dependency notification.

290.8.3 Source lineage uncertainty shall require restriction, labeling, further review, or exclusion where reliance would otherwise be misleading.

290.9 Version Control.

290.9.1 Material datasets, data records, evidence packs, schemas, data dictionaries, dashboards, maps, models, embeddings, prompts where safe, inference records, technical baselines, software, publications, and public-safe outputs shall be version-controlled where necessary to preserve integrity, reproducibility, auditability, correctionability, and technical memory.

290.9.2 Version records shall identify version number or identifier, date, custodian, author or system source where appropriate, change reason, affected fields, affected outputs, compatibility with prior versions, review status, public-safe status, and supersession relationship.

290.9.3 No person shall replace, overwrite, silently alter, relabel, or republish material data in a way that obscures prior versions, correction history, or reliance status.

290.10 Change Log.

290.10.1 The Corporation shall maintain change logs for material datasets, repositories, dashboards, maps, technical baselines, models, inference workflows, evidence records, and public-safe publications.

290.10.2 Change logs shall identify material additions, deletions, transformations, corrections, reclassifications, access changes, publication changes, AI-use changes, permission changes, restriction changes, and archive changes.

290.10.3 Change logs shall be protected from unauthorized alteration and retained according to the applicable retention schedule.

290.11 Derived Data Controls.

290.11.1 Derived Data means data created through transformation, aggregation, calculation, inference, summarization, enrichment, linkage, normalization, de-identification, pseudonymization, geospatial processing, AI processing, model output, or other processing of source data.

290.11.2 Derived Data shall retain traceability to source data, methods, transformations, permissions, restrictions, data rights, public authority conditions, protected knowledge conditions, AI-use restrictions, and correction obligations.

290.11.3 Derived Data shall not be treated as free from source restrictions merely because it has been transformed, summarized, aggregated, embedded, de-identified, or converted into model output.

290.12 Synthetic Data Controls.

290.12.1 Synthetic Data means artificially generated data intended to resemble, simulate, augment, substitute for, or test against real data.

290.12.2 Synthetic Data shall be labeled as synthetic and shall include records identifying generation method, source dependencies, model used where applicable, purpose, limitations, re-identification risk, bias risk, representativeness limits, prohibited uses, and public-safe status.

290.12.3 Synthetic Data shall not be presented as observed data, public authority data, community input, protected knowledge, verified evidence, real-world measurement, benchmark truth, public-safe fact, or operational condition.

290.12.4 Synthetic Data derived from sensitive source data shall remain subject to review for privacy leakage, protected knowledge leakage, public authority restrictions, cyber sensitivity, infrastructure sensitivity, and AI model inversion risk.

290.13 Data Labeling Controls.

290.13.1 Data Labeling Controls shall govern labels applied to data for classification, annotation, AI training, AI evaluation, evidence classification, public-safe status, confidence, uncertainty, technology category, public authority capacity, rights-bearing status, protected knowledge status, or publication use.

290.13.2 Labels shall be accurate, consistent with controlled vocabulary, versioned where material, reviewer-supported where required, and correctionable.

290.13.3 AI-assisted labeling shall be reviewed for hallucination, bias, semantic drift, protected knowledge misclassification, public authority capacity error, finance-boundary error, certification-boundary error, recognition-boundary error, procurement-boundary error, and public-safe status error.

290.14 Error Correction.

290.14.1 Data errors shall be corrected promptly when identified, including errors in values, labels, classifications, timestamps, sources, permissions, restrictions, access classes, public-safe status, AI-use status, derived data, synthetic data, maps, dashboards, model outputs, or publications.

290.14.2 Correction shall preserve prior state where necessary for historical traceability, auditability, legal compliance, public authority trust, protected knowledge duties, and technical memory.

290.14.3 Corrected data shall trigger downstream review where affected outputs, publications, dashboards, maps, repositories, AI outputs, technical baselines, public authority learning materials, Docket inputs, Grid inputs, GRF-facing inputs, or GRA-facing inputs may have relied on the erroneous data.

290.15 Supersession.

290.15.1 Data shall be superseded when a newer, corrected, more authoritative, more complete, more accurate, more current, or more appropriate record replaces or materially qualifies an earlier record.

290.15.2 Superseded data shall be marked as superseded and shall not be used as current data unless its historical status is clearly identified and such use is appropriate.

290.15.3 Supersession shall identify affected records, affected outputs, replacement records, effective date, reason, and downstream notification requirements.

290.16 Withdrawal.

290.16.1 Data shall be withdrawn from use where it is unlawful to use, improperly permissioned, materially inaccurate, unsafe, misclassified, compromised, rights-restricted, protected-knowledge-exposing, public authority-restricted, cyber-unsafe, infrastructure-unsafe, or inconsistent with this Bylaw.

290.16.2 Withdrawal may apply to source data, derived data, synthetic data, datasets, dashboards, maps, embeddings, model outputs, repositories, public-safe reports, technical baselines, or publications.

290.16.3 Withdrawal shall include access restriction, reliance status marking, downstream dependency review, public or controlled notice where required, and archive treatment.

290.17 Retraction Where Public Outputs Are Affected.

290.17.1 Where public outputs materially relied on erroneous, fabricated, falsified, improperly permissioned, restricted, unsafe, or withdrawn data, the Corporation shall consider retraction where correction or supersession is insufficient to prevent misleading reliance.

290.17.2 Retraction may apply to public-safe reports, dashboards, maps, datasets, technical baselines, software releases, model cards, system cards, benchmark cards, Academy materials, public authority learning materials, donor reports, sponsor reports, or Nexus-facing materials.

290.17.3 Retraction records shall identify reason, scope, affected outputs, replacement status, archive status, and notice requirements while protecting restricted information.

290.18 Data Quality, Integrity, and Correction Records.

290.18.1 The Corporation shall maintain Data Quality, Integrity, and Correction Records, including data quality reviews, accuracy reviews, completeness reviews, timeliness reviews, relevance reviews, integrity reviews, provenance reviews, source lineage reviews, version records, change logs, derived data records, synthetic data records, data labeling records, error correction records, supersession records, withdrawal records, retraction records, downstream dependency records, notices, restrictions, deletions, and archive records.

---

Section 291. Retention, Deletion, Sealing, Archival, Legal Holds, and Secure Disposal

291.1 Retention Purpose.

291.1.1 Retention governance shall ensure that Corporation data, records, research materials, evidence materials, methods, public authority data, privacy records, cyber records, AI records, model records, inference records, compute records, controlled-room records, fiscal records, governance records, technical baselines, software records, publication records, correction records, and protected knowledge records are retained for lawful, public-benefit, institutional, evidentiary, audit, tax, research, public authority, safeguards, and technical memory purposes.

291.1.2 Retention shall be balanced with data minimization, privacy, protected knowledge duties, public authority restrictions, cybersecurity, confidentiality, legal holds, correctionability, and secure disposal.

291.2 Retention Schedule.

291.2.1 The Corporation shall maintain one or more retention schedules identifying categories of records, minimum retention periods, legal requirements, contractual requirements, grant requirements, public authority requirements, research requirements, privacy requirements, deletion rules, archive rules, and secure disposal rules.

291.2.2 Retention schedules shall be reviewed periodically and updated when law, grants, contracts, public authority requirements, research obligations, privacy obligations, technology, institutional operations, or risk context changes.

291.3 Minimum Retention Requirements.

291.3.1 Minimum retention requirements shall be set for governance records, fiscal records, tax records, grant records, contract records, employment records, research records, evidence records, data records, privacy records, cybersecurity records, AI governance records, model records, controlled-room records, publication records, correction records, and incident records.

291.3.2 No record shall be deleted before the applicable minimum retention period expires unless lawful authority, applicable restriction, privacy obligation, protected knowledge duty, or competent legal review permits or requires deletion.

291.4 Contractual Retention Requirements.

291.4.1 Contractual retention requirements shall be identified and followed for contracts, data-sharing agreements, licenses, vendor agreements, AI-provider agreements, cloud agreements, public authority agreements, grants, sponsorships, research agreements, collaboration agreements, and controlled-room agreements.

291.4.2 Contractual deletion, return, audit, access, archive, confidentiality, and evidence-preservation obligations shall be reflected in retention controls.

291.5 Public Authority Retention Requirements.

291.5.1 Public Authority Data and related records shall be retained, sealed, deleted, returned, transferred, archived, or disposed of according to applicable law, public authority instructions, public records considerations, contracts, grants, confidentiality obligations, security requirements, and capacity records.

291.5.2 The Corporation shall not destroy, alter, conceal, or withhold Public Authority Data to evade lawful public records, audit, grant, procurement, investigation, or public authority requirements.

291.6 Research Retention Requirements.

291.6.1 Research records shall be retained as required by law, research ethics, IRB or equivalent determinations, grant terms, publication standards, data rights, institutional policy, correction obligations, and technical memory needs.

291.6.2 Research retention shall include protocols, ethics records, consent records, source records, datasets, evidence records, method records, reviewer notes where retained, analysis records, AI-use records, compute records, publication records, challenge records, and correction records.

291.7 Grant and Funder Retention Requirements.

291.7.1 Grant and funder records shall be retained according to grant terms, funder requirements, accounting rules, tax obligations, audit requirements, donor restrictions, public authority requirements, and Board policy.

291.7.2 Grant and funder retention shall include applications, agreements, budgets, restrictions, expenditure records, deliverables, reports, correspondence, modifications, public-safe outputs, corrections, and closeout records.

291.8 Legal Holds.

291.8.1 A Legal Hold shall suspend ordinary deletion, alteration, disposal, or destruction of records where litigation, investigation, claim, audit, regulatory inquiry, public authority inquiry, subpoena, preservation request, incident, misconduct review, or legal risk requires preservation.

291.8.2 Legal Holds shall identify scope, affected custodians, affected systems, affected data, start date, responsible authority, notice requirements, preservation steps, and release conditions.

291.8.3 No person shall delete, alter, conceal, overwrite, or destroy records subject to Legal Hold.

291.9 Regulatory Holds.

291.9.1 A Regulatory Hold shall preserve records where a regulator, public authority, grantor, auditor, tax authority, charitable solicitation authority, privacy authority, cybersecurity authority, or other competent body requires or may require preservation.

291.9.2 Regulatory Holds shall be coordinated with counsel or competent authority where appropriate and shall be recorded.

291.10 Litigation Holds.

291.10.1 A Litigation Hold shall be imposed when litigation is pending, threatened, reasonably anticipated, or otherwise requires preservation of records.

291.10.2 Litigation Holds shall preserve relevant records, including email, messaging, documents, datasets, logs, repositories, model records, inference records, dashboards, maps, financial records, governance records, contracts, and communications.

291.11 Sealing.

291.11.1 Sealing shall restrict access to records that require heightened protection due to privilege, confidentiality, investigation sensitivity, public authority restriction, protected knowledge, personal information, health-sensitive data, cyber sensitivity, infrastructure sensitivity, export-control, sanctions sensitivity, misconduct review, or legal hold.

291.11.2 Sealed records shall identify sealing authority, reason, scope, access conditions, review date, and unsealing conditions where appropriate.

291.12 Archival.

291.12.1 Archival shall preserve records for institutional memory, legal compliance, research integrity, evidence integrity, technical memory, correctionability, auditability, historical traceability, and public-benefit continuity.

291.12.2 Archived records shall be classified by current, historical, superseded, corrected, withdrawn, retracted, restricted, sealed, privileged, public-safe, internal, controlled, deleted-under-policy, or preserved-under-hold status where applicable.

291.12.3 Archived records shall not be treated as current authority unless their current status is confirmed by competent record.

291.13 Deletion.

291.13.1 Deletion shall occur where required by law, contract, consent withdrawal, rights request, public authority instruction, data minimization, retention schedule, protected knowledge restriction, ethics requirement, or competent institutional decision, provided that no Legal Hold, Regulatory Hold, Litigation Hold, audit requirement, correction obligation, or preservation duty prevents deletion.

291.13.2 Deletion shall be recorded and shall identify data deleted, authority, reason, date, method, affected systems, backups where applicable, and residual limitations.

291.13.3 Deletion shall not be used to conceal misconduct, evade review, destroy evidence, suppress unfavorable findings, avoid correction, or frustrate public authority, legal, audit, research, or safeguards obligations.

291.14 Secure Disposal.

291.14.1 Secure disposal shall ensure that records and data approved for disposal are destroyed or rendered inaccessible using methods appropriate to classification, medium, system, and sensitivity.

291.14.2 Secure disposal may include secure deletion, cryptographic erasure, physical destruction, certified disposal, repository removal, access revocation, key destruction, storage wipe, or vendor-certified destruction.

291.14.3 Secure disposal of sensitive records shall be documented and may require witness, certificate, audit trail, or vendor confirmation.

291.15 Backup and Archive Alignment.

291.15.1 Retention, deletion, sealing, archival, and disposal controls shall address backups, replicas, logs, snapshots, caches, AI embeddings, derived datasets, external processors, vendor systems, cloud storage, controlled rooms, collaboration platforms, and repository mirrors where feasible and appropriate.

291.15.2 Where immediate deletion from backup systems is not technically feasible, the Corporation shall prevent restoration into active use except in accordance with deletion restrictions, legal holds, and correction obligations.

291.16 Data Subject or Participant Deletion Requests Where Applicable.

291.16.1 Data subject, participant, contributor, community, or rights-bearing person deletion requests shall be reviewed under applicable law, consent terms, research ethics, public authority restrictions, protected knowledge protocols, retention obligations, legal holds, and institutional policy.

291.16.2 The Corporation may deny, defer, limit, or partially fulfill deletion where lawful grounds require retention, but shall record the basis and provide explanation or appeal pathway where applicable.

291.17 Protected Knowledge Disposition.

291.17.1 Protected Knowledge disposition shall be governed by consent, non-consent, withdrawal, restriction, community protocols, Tribal / Indigenous protocols, safeguards review, public-safe status, legal obligations, and institutional correction duties.

291.17.2 Protected Knowledge may require return, deletion, sealing, controlled archive, non-public retention, restricted access, non-attribution, public-safe correction, or withdrawal of downstream outputs.

291.18 Retention, Deletion, Sealing, Archival, Legal Hold, and Secure Disposal Records.

291.18.1 The Corporation shall maintain Retention, Deletion, Sealing, Archival, Legal Hold, and Secure Disposal Records, including retention schedules, minimum retention records, contractual retention records, public authority retention records, research retention records, grant and funder retention records, Legal Holds, Regulatory Holds, Litigation Holds, sealing records, archival records, deletion records, secure disposal records, backup and archive alignment records, data subject or participant deletion request records, protected knowledge disposition records, corrections, restrictions, and closeout records.

---

Section 292. AI Governance Program

292.1 AI Governance Program Purpose.

292.1.1 The Corporation shall maintain an AI Governance Program to ensure that AI systems are acquired, developed, configured, approved, used, monitored, evaluated, restricted, retired, and corrected in a lawful, secure, human-governed, evidence-grounded, public-benefit, bias-aware, privacy-protective, public-safe, and correctionable manner.

292.1.2 The AI Governance Program shall apply to generative AI, agentic AI, retrieval systems, embedding systems, classifiers, summarizers, transcription tools, translation tools, anomaly detectors, routing tools, model evaluation tools, autonomous or semi-autonomous workflows, AI-enabled software, AI-assisted research tools, AI-enabled cybersecurity tools, and any system that materially produces, ranks, classifies, infers, recommends, drafts, summarizes, translates, routes, detects, or evaluates information for Corporation purposes.

292.1.3 The AI Governance Program shall preserve non-execution, public authority boundary discipline, finance-boundary discipline, certification-boundary discipline, recognition-boundary discipline, procurement neutrality, provider neutrality, sponsor non-control, public-safe publication, data / AI / cyber security, protected knowledge safeguards, research integrity, evidence integrity, verifiable intelligence, and correctionability.

292.2 AI-Use Policy.

292.2.1 The Corporation shall adopt and maintain an AI-Use Policy or equivalent recorded controls governing permitted AI uses, prohibited AI uses, approved systems, approved data classes, prohibited data classes, human review, model records, inference records, data retention, no-training requirements, prompt restrictions, output review, publication controls, incident response, and enforcement.

292.2.2 The AI-Use Policy shall prohibit AI uses that bypass human judgment, fabricate evidence, generate unsupported public claims, process restricted data in unapproved systems, disclose protected knowledge, issue public authority statements, create finance-readiness claims, create certification claims, create recognition claims, create procurement recommendations, or provide regulated advice.

292.2.3 The AI-Use Policy shall be reviewed and updated when AI systems, law, data classes, research practices, public authority interfaces, cybersecurity risk, protected knowledge duties, publication risks, or institutional operations materially change.

292.3 AI System Inventory.

292.3.1 The Corporation shall maintain an AI System Inventory identifying material AI systems used, tested, procured, deployed, or approved by the Corporation.

292.3.2 The AI System Inventory shall identify system name, provider, deployment environment, owner, custodian, model family where known, approved use cases, prohibited use cases, data classes permitted, data classes prohibited, access controls, integration points, risk class, review cycle, incident history, and retirement status.

292.4 Model Register.

292.4.1 The Corporation shall maintain a Model Register for material models used in Corporation activities, including externally hosted models, internally hosted models, open-source models, proprietary models, embedding models, retrieval models, classification models, summarization models, anomaly detection models, transcription models, translation models, evaluation models, and agentic systems.

292.4.2 The Model Register shall support model governance, approved use, risk review, AI vendor review, data protection, human review, incident response, evaluation, retirement, and correction.

292.5 AI Risk Classification.

292.5.1 Each material AI use case or AI system shall be assigned an AI Risk Classification proportionate to its data class, use case, autonomy, human review, public authority relevance, rights impact, public-safe publication risk, cybersecurity risk, protected knowledge risk, research integrity impact, evidence integrity impact, and boundary risk.

292.5.2 AI Risk Classification may include low, moderate, high, restricted, prohibited, experimental, controlled-room-only, public-authority-restricted, protected-knowledge-restricted, cyber-restricted, or other approved categories.

292.5.3 Higher-risk AI uses shall require heightened approval, documentation, human review, monitoring, testing, access controls, incident escalation, and public-safe review.

292.6 AI Use Case Approval.

292.6.1 Material AI use cases shall be approved before use by competent authority under the AI Governance Program.

292.6.2 AI use case approval shall identify purpose, system, model, users, data classes, permitted inputs, prohibited inputs, permitted outputs, prohibited outputs, human review, access controls, records, retention, publication limits, privacy review, security review, protected knowledge review, public authority boundary review, finance-boundary review, certification-boundary review, recognition-boundary review, procurement-boundary review, and incident response.

292.6.3 Experimental AI uses shall be clearly labeled, bounded, access-controlled, and barred from public reliance unless separately approved.

292.7 AI System Owner.

292.7.1 Each material AI system shall have an AI System Owner responsible for institutional purpose, approved use, risk classification, compliance, monitoring, review, and retirement.

292.7.2 AI System Owner status shall not create personal ownership, unrestricted authority, publication authority, certification authority, public authority authority, finance-readiness authority, procurement authority, or recognition authority.

292.8 Model Owner.

292.8.1 Each material model in the Model Register shall have a Model Owner responsible for the model’s permitted use, risk classification, evaluation status, known limitations, incident history, and continued suitability for approved purposes.

292.8.2 The Model Owner shall ensure that model use remains aligned with data rights, AI-use restrictions, no-training terms, public-safe requirements, and human review requirements.

292.9 Model Custodian.

292.9.1 Each material model shall have a Model Custodian responsible for technical custody, access controls, configuration, deployment records, version records, inference records, monitoring records, incident records, retirement records, and archive status.

292.9.2 Model Custodians shall preserve logs and records sufficient for correction, auditability where appropriate, and incident response.

292.10 Data Steward Interface.

292.10.1 AI governance shall interface with data stewardship to ensure that data used in AI systems is lawfully sourced, properly classified, permissioned, purpose-limited, access-controlled, retention-managed, protected from unauthorized training or embedding, and public-safe where used externally.

292.10.2 Data Stewards shall be consulted where AI use involves public authority data, personal information, health-sensitive data, rights-bearing data, cyber-sensitive data, infrastructure-sensitive data, controlled technology data, community-protected data, protected knowledge, confidential data, finance-sensitive data, or research-sensitive data.

292.11 AI Vendor Review.

292.11.1 AI vendors and AI service providers shall be reviewed before material use for security, privacy, data retention, training terms, model improvement terms, inference logging, sub-processors, cross-border processing, access controls, confidentiality, auditability, incident response, deletion, export-control, sanctions, IP, public authority restrictions, and contractual controls.

292.11.2 AI vendor review shall consider whether vendor terms permit provider training, model improvement, human review by vendor personnel, retention of prompts, retention of outputs, embedding reuse, or onward transfer.

292.11.3 AI vendors shall not be approved for sensitive data classes unless contractual, technical, and governance controls are sufficient for those classes.

292.12 AI Procurement Review.

292.12.1 Procurement of AI systems shall be subject to procurement controls, contract review, data protection review, cybersecurity review, AI governance review, public authority boundary review, IP review, accessibility review, bias review, sanctions and export-control review where applicable, and provider-neutrality review.

292.12.2 AI procurement shall not create provider lock-in, unsupported dependency, hidden data transfer, unauthorized training, proprietary enclosure of public-good assets, certification implication, public authority overclaim, or procurement neutrality violation.

292.13 AI Public-Safe Review.

292.13.1 AI-assisted outputs intended for external circulation, public-safe publication, public authority learning, dashboards, maps, technical baselines, donor reports, sponsor reports, GRF-facing inputs, GRA-facing inputs, Docket inputs, Grid inputs, or Nexus-facing materials shall receive public-safe review.

292.13.2 Public-safe review shall assess accuracy, source support, hallucination risk, bias, public authority boundary, finance boundary, certification boundary, recognition boundary, procurement neutrality, provider neutrality, protected knowledge, privacy, cyber sensitivity, infrastructure sensitivity, and correction path.

292.14 AI Human Review Requirement.

292.14.1 Human review shall be required before material AI outputs are used for evidence classification, public-safe publication, public authority learning materials, technical baselines, controlled-room outputs, dashboards, maps, external reports, donor reports, sponsor reports, Docket inputs, Grid inputs, GRF-facing inputs, GRA-facing inputs, or other high-reliance purposes.

292.14.2 Human review shall verify that AI outputs are source-supported, limitation-aware, uncertainty-aware, non-hallucinated, non-fabricated, properly classified, safe for intended use, and within the Corporation’s non-executing role.

292.14.3 Human review shall not be nominal. Reviewers shall have sufficient competence, access, time, and authority to reject, revise, restrict, or escalate AI outputs.

292.15 AI Monitoring.

292.15.1 The Corporation shall monitor material AI systems and use cases proportionate to risk.

292.15.2 AI monitoring may include review of usage logs, output quality, hallucination rates, bias concerns, data leakage indicators, prompt injection risk, retrieval quality, access patterns, vendor changes, model version changes, drift, incidents, and user compliance.

292.15.3 Monitoring shall be used to protect integrity, privacy, security, safeguards, and public-safe publication, and shall not be used for improper surveillance, retaliation, discrimination, or unauthorized personnel action.

292.16 AI Incident Response.

292.16.1 AI Incident Response shall address suspected or confirmed AI-related incidents, including unauthorized data upload, unauthorized training, prompt leakage, embedding leakage, model hallucination in material output, fabricated citations, biased output, protected knowledge exposure, public authority overclaim, finance overclaim, certification overclaim, procurement overclaim, recognition overclaim, cyber compromise, model misuse, agentic misaction, or vendor incident.

292.16.2 AI Incident Response may require model-use hold, prompt review where safe, inference record review, access revocation, vendor notice, data deletion, unlearning request where available, output correction, publication withdrawal, public-safe notice, legal review, public authority notice where required, and recurrence prevention.

292.17 AI Governance Training.

292.17.1 Persons using AI systems for Corporation work shall receive training appropriate to role and risk.

292.17.2 AI governance training may address approved tools, prohibited tools, prohibited data inputs, no-training rules, hallucination risk, bias risk, confidentiality, protected knowledge, public authority data, cyber-sensitive data, human review, publication limits, prompt hygiene, incident reporting, and non-execution boundaries.

292.18 AI Governance Records.

292.18.1 The Corporation shall maintain AI Governance Records, including AI-Use Policy records, AI System Inventory entries, Model Register entries, AI Risk Classifications, AI use case approvals, AI System Owner records, Model Owner records, Model Custodian records, Data Steward interface records, AI vendor reviews, AI procurement reviews, AI public-safe reviews, human review records, monitoring records, AI incident response records, AI governance training records, corrections, restrictions, retirements, and archive records.

---

Section 293. Model Register, Model Records, Dataset Cards, Model Cards, System Cards, Benchmark Cards, and Evaluation Harnesses

293.1 Model Register Requirement.

293.1.1 The Corporation shall maintain a Model Register for material AI models, machine learning models, embedding models, retrieval models, classification models, summarization models, transcription models, translation models, anomaly detection models, evaluation models, simulation models, digital twin models, risk models, and other models used in Corporation activities.

293.1.2 The Model Register shall support governance, traceability, permitted use, prohibited use, evaluation, public-safe review, data rights, incident response, human review, retirement, correction, and technical memory.

293.2 Model Identity.

293.2.1 Model records shall identify the model’s name, model family, system name, deployment name, provider identifier, internal identifier, and any aliases or versions used in Corporation systems or documentation.

293.2.2 Model identity shall be precise enough to distinguish among materially different models, hosted services, fine-tuned variants, internally configured variants, embedding models, retrieval systems, and agentic systems.

293.3 Model Version.

293.3.1 Model records shall identify version, release date where known, deployment date, update date, configuration state, fine-tuning status, prompt or system instruction version where material and safe, retrieval index version, embedding version, and retirement or supersession status.

293.3.2 Version changes shall be reviewed where they may affect output quality, data handling, public-safe publication, bias, hallucination risk, evaluation results, security, confidentiality, or permitted use.

293.4 Model Provider.

293.4.1 Model records shall identify the provider, vendor, host, open-source project, internal team, cloud provider, deployment environment, sub-processors where material, support model, and applicable contractual terms.

293.4.2 Provider records shall include data retention, training, model improvement, inference logging, confidentiality, security, deletion, sub-processing, cross-border transfer, audit, IP, and incident terms where applicable.

293.5 Model Purpose.

293.5.1 Model records shall identify the approved purpose of the model, including research support, summarization, classification, translation, transcription, retrieval, anomaly detection, evidence comparison, public-safe drafting, software support, cybersecurity support, accessibility support, or other approved use.

293.5.2 A model approved for one purpose shall not be used for another purpose without reuse review and approval where material.

293.6 Permitted Uses.

293.6.1 Model records shall identify permitted uses and permitted data classes.

293.6.2 Permitted uses may be conditioned on human review, controlled-room use, no-training terms, non-public use, public-safe review, limited data classes, access restrictions, output restrictions, or defined retention periods.

293.7 Prohibited Uses.

293.7.1 Model records shall identify prohibited uses, including uses involving prohibited data classes, automated final decisions, public authority decisions, public warnings, emergency commands, finance-readiness determinations, certification determinations, recognition determinations, procurement recommendations, provider rankings, legal advice, clinical advice, or unreviewed public publication.

293.7.2 Prohibited uses shall be enforced through policy, access controls, training, technical controls where feasible, monitoring, incident response, and enforcement.

293.8 Risk Class.

293.8.1 Each material model shall have a risk class reflecting data sensitivity, use case, autonomy, rights impact, public authority relevance, publication risk, cybersecurity risk, protected knowledge risk, model opacity, provider risk, integration risk, and boundary risk.

293.8.2 Risk class shall determine evaluation depth, approval authority, human review, monitoring, logging, access controls, incident thresholds, and retirement review.

293.9 Data Access.

293.9.1 Model records shall identify what data classes the model may access, ingest, retrieve, embed, store, process, generate, or output.

293.9.2 Data access records shall identify whether prompts, outputs, embeddings, retrieval results, logs, or fine-tuning data are retained, transmitted, used for model improvement, available to vendor personnel, or subject to deletion.

293.10 Evaluation Record.

293.10.1 Material models shall have evaluation records proportionate to risk and use.

293.10.2 Evaluation records may include benchmark results, task-specific evaluations, hallucination reviews, bias reviews, security reviews, privacy reviews, protected knowledge reviews, public authority boundary reviews, finance-boundary reviews, certification-boundary reviews, recognition-boundary reviews, procurement-boundary reviews, robustness tests, red-team findings, user testing, and limitations.

293.10.3 Evaluation records shall not be represented as certification, public authority approval, legal compliance approval, procurement approval, finance-readiness, recognition, rating, or safety guarantee.

293.11 Known Limitations.

293.11.1 Model records shall identify known limitations, including hallucination risk, bias, domain limits, language limits, jurisdiction limits, stale knowledge, reasoning limits, retrieval limits, context window limits, privacy limits, security limits, protected knowledge risks, public authority risks, and unsuitable uses.

293.11.2 Known limitations shall be disclosed to users and reviewers where necessary to prevent misuse or misleading reliance.

293.12 Incident History.

293.12.1 Model records shall include material incident history, including hallucination incidents, data leakage, unauthorized data use, prompt injection, biased outputs, protected knowledge exposure, public authority overclaim, finance overclaim, certification overclaim, procurement overclaim, model outage, vendor incident, or security event.

293.12.2 Incident history shall inform continued approval, restrictions, monitoring, retraining, vendor action, or retirement.

293.13 Retirement Status.

293.13.1 Model records shall identify whether a model is active, experimental, restricted, suspended, deprecated, retired, prohibited, superseded, or archived.

293.13.2 Retired or prohibited models shall not be used for Corporation work except under controlled archive or forensic review authorized by competent authority.

293.14 Dataset Cards.

293.14.1 Dataset Cards shall describe material datasets used for AI training where authorized, evaluation, benchmarking, retrieval, analysis, or evidence work.

293.14.2 Dataset Cards may include source, purpose, collection method, data classes, permissions, license, representativeness, limitations, bias, missingness, privacy, protected knowledge, public authority restrictions, cyber sensitivity, infrastructure sensitivity, retention, deletion, and correction path.

293.15 Model Cards.

293.15.1 Model Cards shall describe material models used by the Corporation, including purpose, intended use, prohibited use, training context where known and permissible to describe, evaluation, limitations, risk class, data access, human review requirements, and incident history.

293.15.2 Model Cards shall be written to prevent overclaim and shall not imply certification, recognition, finance-readiness, procurement approval, public authority approval, legal compliance, or safety guarantee.

293.16 System Cards.

293.16.1 System Cards shall describe AI systems composed of models, retrieval layers, tools, workflows, interfaces, human review, access controls, datasets, prompts, integrations, and monitoring.

293.16.2 System Cards shall identify system purpose, users, data classes, model components, workflow, human review, security controls, public-safe controls, limitations, incident response, and retirement path.

293.17 Benchmark Cards.

293.17.1 Benchmark Cards shall describe benchmarks used to evaluate models, systems, methods, datasets, public-good software, technical baselines, dashboards, maps, or evidence workflows.

293.17.2 Benchmark Cards shall identify benchmark purpose, test set, scope, limitations, version, leakage controls, gaming risk, scoring method, evaluation conditions, reviewer, and public-safe status.

293.17.3 Benchmark Cards shall not be used as rankings, certifications, procurement recommendations, finance-readiness determinations, public authority approvals, provider endorsements, or ratings.

293.18 Evaluation Harnesses.

293.18.1 Evaluation Harnesses shall be controlled tools, scripts, workflows, prompts, test environments, datasets, or procedures used to evaluate models, systems, methods, benchmarks, software, dashboards, maps, evidence workflows, or technical baselines.

293.18.2 Evaluation Harnesses shall be versioned, access-controlled, documented, protected against leakage or gaming, reviewed for bias and representativeness where appropriate, and linked to evaluation records.

293.19 Model Register and Evaluation Records.

293.19.1 The Corporation shall maintain Model Register and Evaluation Records, including model identity records, model version records, provider records, purpose records, permitted use records, prohibited use records, risk class records, data access records, evaluation records, known limitation records, incident history records, retirement status records, Dataset Cards, Model Cards, System Cards, Benchmark Cards, Evaluation Harness records, corrections, restrictions, supersessions, retirements, and archive records.

---

Section 294. AI Training, Fine-Tuning, Embeddings, Retrieval, Model Improvement, and Sensitive Data Restrictions

294.1 Training Restriction Purpose.

294.1.1 AI training restrictions shall prevent unauthorized, unsafe, unlawful, non-consensual, non-permissioned, non-public-safe, or boundary-defective use of Corporation data for AI training, model development, model improvement, fine-tuning, embedding, retrieval, inference, benchmarking, or external model enhancement.

294.1.2 AI training restrictions shall protect public authority data, rights-bearing data, personal information, health-sensitive data, cyber-sensitive data, infrastructure-sensitive data, controlled technology data, community-protected data, Tribal / Indigenous data, local and territorial knowledge, cultural and environmental knowledge, protected knowledge, confidential data, privileged data, commercially sensitive data, finance-sensitive data, research-sensitive data, and Nexus-sensitive data.

294.2 Fine-Tuning Restriction.

294.2.1 No Corporation data shall be used to fine-tune, adapt, specialize, reinforce, instruct, align, or otherwise modify a model unless express authorization, lawful basis, data rights, permission, risk review, data classification, public-safe status, and competent approval support the use.

294.2.2 Fine-tuning with sensitive data shall be prohibited unless a heightened review determines that the use is lawful, necessary, secure, permissioned, no less protective alternative exists, and residual risk is acceptable.

294.3 Embedding Restriction.

294.3.1 No Corporation data shall be converted into embeddings, vector representations, semantic indexes, retrieval indexes, model memory, or similarity-search stores unless authorized, purpose-limited, access-controlled, retention-managed, deletion-capable where feasible, and reviewed for leakage risk.

294.3.2 Embeddings derived from sensitive data shall be treated as sensitive and shall remain subject to source restrictions, data rights, privacy obligations, public authority restrictions, protected knowledge safeguards, and deletion obligations where feasible.

294.4 Retrieval-Augmented Generation Restriction.

294.4.1 Retrieval-Augmented Generation systems using Corporation data shall be approved before material use and shall be governed by data classification, access control, source filtering, retrieval logging where appropriate, prompt-injection controls, output review, hallucination review, public-safe review, and correction paths.

294.4.2 Retrieval systems shall not retrieve restricted, confidential, privileged, public authority, health-sensitive, cyber-sensitive, infrastructure-sensitive, community-protected, or protected knowledge data for unauthorized users or unapproved purposes.

294.5 Model Improvement Restriction.

294.5.1 Corporation data shall not be used by any AI vendor, provider, processor, collaborator, sponsor, donor, funder, public authority, national company, Project SPV, enterprise actor, or internal team for model improvement, training, reinforcement, evaluation, product development, or benchmarking unless expressly authorized by competent record.

294.5.2 Default model improvement by vendors or platforms shall be disabled or contractually prohibited where sensitive or confidential data is involved, unless competent authority approves otherwise after review.

294.6 Sensitive Nexus Data Restriction.

294.6.1 Sensitive Nexus Data includes non-public Nexus architecture, controlled vocabulary, technical baselines, Docket materials, Grid materials, GRF-facing inputs, GRA-facing inputs, public authority learning materials, controlled-room materials, evidence packs, proof receipts, strategic plans, implementation materials, consortium records, national company interface records, Project SPV interface records, and non-public public-good stack materials.

294.6.2 Sensitive Nexus Data shall not be used for AI training, fine-tuning, embeddings, retrieval, model improvement, public release, sponsor benefit, provider development, capital-reader materials, or external sharing unless authorized, classified, public-safe reviewed, and boundary-reviewed.

294.7 Public Authority Data Restriction.

294.7.1 Public Authority Data shall not be used for AI training, fine-tuning, embeddings, model improvement, external retrieval, or AI-assisted publication unless public authority permission, lawful basis, contract terms, data classification, AI-use approval, no-training controls, and public-safe review support the use.

294.7.2 Public Authority Data shall not be used to generate public authority statements, official interpretations, public warnings, emergency commands, procurement advice, regulatory advice, grant determinations, public finance determinations, or sovereign decisions.

294.8 Rights-Bearing Data Restriction.

294.8.1 Rights-Bearing Data and personal information shall not be used for AI training, fine-tuning, embeddings, model improvement, automated classification, profiling, public-safe mapping, or external sharing unless lawful basis, notice, consent where required, rights review, bias review, privacy review, civil rights review, and public-safe review support the use.

294.8.2 Rights-Bearing Data shall not be processed through AI in a manner that creates discriminatory impact, inaccessible participation, retaliation risk, unjustified profiling, opaque classification, unsupported inference, or misleading public interpretation.

294.9 Health-Sensitive Data Restriction.

294.9.1 Health-Sensitive Data, Public Health Data, Clinical or Medical Data, youth health data, vulnerable-population health data, and health-system data shall not be used for AI training, fine-tuning, embeddings, retrieval, model improvement, or public AI outputs unless heightened privacy, ethics, public authority, security, bias, and public-safe review authorize such use.

294.9.2 AI outputs derived from Health-Sensitive Data shall not be used as clinical advice, diagnosis, treatment guidance, public health order, emergency command, public warning, or substitute for licensed professional judgment or competent public authority action.

294.10 Cyber-Sensitive and Infrastructure-Sensitive Data Restriction.

294.10.1 Cyber-Sensitive Data and Infrastructure-Sensitive Data shall not be used for AI training, fine-tuning, embeddings, retrieval, model improvement, external AI analysis, or public generation where such use may expose vulnerabilities, credentials, system architecture, operational weaknesses, critical locations, infrastructure dependencies, incident details, or security-sensitive information.

294.10.2 AI processing of cyber-sensitive or infrastructure-sensitive data shall require cyber review, access restriction, public-safe review, prompt-injection controls, output filtering, logging, and incident escalation.

294.11 Community-Protected and Tribal / Indigenous Knowledge Data Restriction.

294.11.1 Community-Protected Data, Tribal / Indigenous Data, Indigenous Knowledge, local knowledge, territorial knowledge, cultural knowledge, environmental knowledge, sacred knowledge, and Protected Knowledge Data shall not be used for AI training, fine-tuning, embeddings, retrieval, model improvement, summarization, translation, mapping, or public generation unless lawful authority, protocol permission where applicable, safeguards review, no-training controls, attribution or non-attribution terms, withdrawal pathways, and public-safe approval support the use.

294.11.2 AI processing shall be prohibited where it risks extraction, decontextualization, semantic drift, cultural harm, ecological harm, community exposure, public authority misuse, sponsor misuse, provider development, or loss of control over protected knowledge.

294.12 Confidential, Privileged, Commercially Sensitive, and Finance-Sensitive Data Restriction.

294.12.1 Confidential Data, privileged materials, attorney-client materials, work product, commercially sensitive data, procurement-sensitive data, sponsor-sensitive data, donor-sensitive data, funder-sensitive data, finance-sensitive data, non-public compensation data, bank data, and capital-reader materials shall not be used for AI training, fine-tuning, embeddings, retrieval, model improvement, or external AI processing unless authorized by competent record and protected by appropriate contractual, technical, and access controls.

294.12.2 Privileged materials shall not be processed through AI systems in a manner that risks privilege waiver, vendor access, uncontrolled retention, training, disclosure, or unauthorized sharing.

294.13 Express Authorization Requirement.

294.13.1 AI training, fine-tuning, embedding, retrieval indexing, model improvement, or external AI processing involving sensitive data shall require express authorization by competent authority.

294.13.2 Express authorization shall identify data, purpose, model or system, provider, users, data class, permitted use, prohibited use, retention, deletion, no-training terms, human review, public-safe limits, incident response, and review date.

294.13.3 General access to data shall not constitute authorization for AI training, fine-tuning, embeddings, retrieval indexing, model improvement, or external AI processing.

294.14 Recorded Purpose Requirement.

294.14.1 Each authorized AI training, fine-tuning, embedding, retrieval, or model improvement activity shall have a recorded purpose.

294.14.2 The recorded purpose shall be specific, public-benefit aligned, lawful, necessary, proportionate, data-minimized, time-limited where appropriate, and consistent with source permissions and restrictions.

294.14.3 Open-ended, speculative, future-use, provider-development, sponsor-benefit, capital-reader, or convenience-based purposes shall not satisfy this requirement.

294.15 Deletion and Unlearning Path Where Available.

294.15.1 Where data is used in AI systems, the Corporation shall identify deletion, removal, suppression, index deletion, embedding deletion, cache deletion, fine-tune deletion, unlearning, model retirement, or vendor deletion pathways where available.

294.15.2 Where full deletion or unlearning is not technically available, the Corporation shall record the limitation, restrict further use, suppress retrieval where feasible, replace affected indexes or models where appropriate, notify affected parties where required, and prevent misleading claims about deletion.

294.16 Embedding Store Access and Deletion Controls.

294.16.1 Embedding stores, vector databases, semantic indexes, retrieval indexes, and model memory stores containing or derived from Corporation data shall be access-controlled, classified, logged where appropriate, retention-managed, deletion-capable where feasible, and protected from unauthorized export, query, scraping, joining, or inference.

294.16.2 Deletion of source data shall trigger review of derived embeddings, indexes, caches, retrieval records, and model memory where applicable.

294.17 Leakage Risk Review.

294.17.1 Leakage Risk Review shall assess whether AI training, fine-tuning, embeddings, retrieval systems, model improvement, prompts, inference logs, outputs, evaluation sets, or vendor systems may expose restricted data, reveal protected knowledge, disclose personal information, enable re-identification, leak cyber-sensitive data, reveal infrastructure-sensitive data, or disclose confidential or privileged material.

294.17.2 Where leakage risk is unacceptable, the activity shall be denied, redesigned, restricted, localized, aggregated, redacted, routed through a controlled room, or terminated.

294.18 AI Training, Fine-Tuning, Embedding, Retrieval, and Model Improvement Records.

294.18.1 The Corporation shall maintain AI Training, Fine-Tuning, Embedding, Retrieval, and Model Improvement Records, including training restriction records, fine-tuning reviews, embedding reviews, retrieval-augmented generation reviews, model improvement reviews, Sensitive Nexus Data restriction records, Public Authority Data restriction records, Rights-Bearing Data restriction records, Health-Sensitive Data restriction records, Cyber-Sensitive and Infrastructure-Sensitive Data restriction records, Community-Protected and Tribal / Indigenous Knowledge Data restriction records, confidential / privileged / commercially sensitive / finance-sensitive data restriction records, express authorization records, recorded purpose records, deletion and unlearning path records, embedding store access and deletion controls, leakage risk reviews, incidents, corrections, restrictions, denials, and archive records.

Section 295. Inference Records, Compute Workload Records, Verifiable Compute, Proof Receipts, and Verifiable Intelligence Outputs

295.1 Inference Record Purpose.

295.1.1 Inference Records shall preserve the traceability, reviewability, auditability where appropriate, limitation discipline, and correctionability of material AI-assisted outputs generated, received, relied upon, routed, published, archived, or used by the Corporation.

295.1.2 Inference Records shall ensure that AI-assisted outputs remain linked to source records, model identity, version, prompt or input context where safe, retrieval sources, data permissions, human review, confidence classification, limitations, public-safe status, and correction path.

295.1.3 Inference Records shall prevent AI outputs from being treated as ungrounded institutional truth, hidden evidence, unreviewed public authority material, finance-facing determination, certification-like output, recognition-like output, procurement recommendation, public warning, emergency command, or enterprise execution instruction.

295.2 Compute Workload Record Purpose.

295.2.1 Compute Workload Records shall document material computational work performed by or for the Corporation, including AI inference, data processing, evidence analysis, model evaluation, simulation, digital twin execution, benchmark execution, observability processing, dashboard generation, geospatial processing, cryptographic proof generation, public-good software testing, and technical baseline generation.

295.2.2 Compute Workload Records shall support reproducibility where appropriate, verifiability where feasible, source-lineage discipline, cybersecurity, access control, privacy, public authority restrictions, protected knowledge safeguards, correctionability, and technical memory.

295.3 Verifiable Compute Purpose.

295.3.1 Verifiable Compute shall support trusted execution records for computational workflows by documenting inputs, permissions, environment, code, tools, models, workflows, execution parameters, outputs, reviewers, limitations, and correction path.

295.3.2 Verifiable Compute may include hashes, timestamps, execution logs, signed artifacts, dependency records, environment records, container records, repository commit references, workflow receipts, cryptographic receipts where appropriate, and reproducibility packages where lawful and safe.

295.3.3 Verifiable Compute shall support evidence integrity and technical memory and shall not constitute certification, security guarantee, legal compliance approval, public authority approval, finance-readiness, procurement approval, recognition, rating, or operational authorization.

295.4 Proof Receipt Purpose.

295.4.1 A Proof Receipt shall be a record artifact evidencing that a defined input, process, review, computation, submission, validation step, evidence-pack step, controlled-room step, or publication step occurred according to recorded conditions.

295.4.2 A Proof Receipt may document presence, timestamp, source reference, workflow state, reviewer state, version state, custody state, access state, or computation state, but shall not by itself prove substantive truth, safety, legal compliance, public authority adoption, recognition, finance-readiness, procurement approval, certification, rating, public warning, emergency command, or execution authority.

295.4.3 Proof Receipts shall be bounded by their recorded scope, source records, workflow conditions, limitations, confidence classification, and correction path.

295.5 Verifiable Intelligence Output Purpose.

295.5.1 Verifiable Intelligence Outputs shall be AI-assisted or analytically derived outputs that are traceable to source records, model or method records, inference records, compute records, human review records, limitations, confidence classification, public-safe status, and correction path.

295.5.2 Verifiable Intelligence Outputs may support research, evidence classification, summarization, observability, anomaly detection, public authority learning, public-good technical assets, public-safe drafting, and Nexus-facing evidence infrastructure.

295.5.3 Verifiable Intelligence Outputs shall remain subordinate to evidence records, methods records, human review, lawful authority, public-safe review, and correctionability.

295.6 Material AI Output Record Requirement.

295.6.1 Each material AI output used for evidence classification, publication drafting, public authority learning, technical baselines, dashboards, maps, Docket inputs, Grid inputs, GRF-facing inputs, GRA-facing inputs, controlled-room outputs, donor or sponsor reports, or external communications shall have an Inference Record or equivalent record.

295.6.2 Material AI outputs shall not be used in high-reliance contexts unless reviewed, classified, limitation-aware, and traceable to source records.

295.7 Material Compute Workload Record Requirement.

295.7.1 Each material compute workload affecting evidence, research, public-safe publication, AI evaluation, public-good software, technical baselines, dashboards, maps, controlled-room outputs, or Nexus-facing materials shall have a Compute Workload Record.

295.7.2 Compute Workload Records shall be proportionate to risk, reliance, reproducibility need, public authority relevance, privacy sensitivity, cyber sensitivity, protected knowledge implications, and publication significance.

295.8 Input Data Record.

295.8.1 Input Data Records shall identify datasets, documents, records, telemetry, logs, prompts, retrieval sources, model inputs, sensor inputs, public authority data, protected knowledge, or other inputs used in inference or compute workloads.

295.8.2 Input Data Records shall include data class, source lineage, provenance, permissions, restrictions, public-safe status, AI-use status, access class, and correction status where applicable.

295.9 Authority Record.

295.9.1 Authority Records shall identify the lawful, institutional, contractual, research, public authority, contributor, community, Tribal / Indigenous, license, consent, or policy authority supporting inference or compute use.

295.9.2 Absence of sufficient authority shall require hold, restriction, clarification, deletion where appropriate, or refusal of the inference or compute activity.

295.10 Environment Record.

295.10.1 Environment Records shall identify the compute environment, cloud environment, local environment, container, runtime, hardware where material, region where material, security configuration, access restrictions, and relevant dependencies.

295.10.2 Environment Records shall be sufficient to support review, reproducibility where appropriate, security review, cross-border review, and correction.

295.11 Model, Code, Tool, or Workflow Record.

295.11.1 Material inference and compute records shall identify the model, code, tool, script, workflow, prompt framework, retrieval system, benchmark harness, digital twin, dashboard generator, geospatial tool, or other system used.

295.11.2 Such records shall include version, custodian, permitted use, known limitations, configuration where safe, repository reference where applicable, and approval status.

295.12 Execution Record.

295.12.1 Execution Records shall identify when, by whom or by what authorized system, under what workflow, with what parameters, and for what approved purpose the inference or compute activity occurred.

295.12.2 Execution Records may include timestamps, job identifiers, logs, hashes, runtime parameters, workflow receipts, access records, and error records.

295.13 Output Record.

295.13.1 Output Records shall preserve the generated output, output identifier, version, date, classification, public-safe status, permitted uses, prohibited uses, downstream dependencies, and whether the output is draft, reviewed, accepted, rejected, corrected, superseded, withdrawn, or archived.

295.13.2 Outputs shall not be used beyond their recorded status and permitted use.

295.14 Reviewer Record.

295.14.1 Reviewer Records shall identify the person, role, committee, workflow, or competent authority responsible for reviewing material inference or compute outputs.

295.14.2 Reviewer Records shall identify review date, scope, conflicts, acceptance, rejection, conditions, limitations, corrections, public-safe determinations, and escalation where applicable.

295.15 Confidence, Limitation, and Classification Record.

295.15.1 Material inference, compute, proof receipt, and verifiable intelligence outputs shall identify confidence, uncertainty, limitations, evidence class, access class, data class, public-safe class, AI-use class, boundary limitations, and correction status where applicable.

295.15.2 Confidence and limitation records shall be clear enough to prevent misuse as certification, recognition, finance-readiness, procurement approval, public authority decision, rating, public warning, emergency command, or guarantee.

295.16 Correction Path.

295.16.1 Each material inference, compute workload, proof receipt, and verifiable intelligence output shall include a correction path identifying custodian, challenge mechanism, review route, affected-output review, supersession rule, withdrawal rule, notice rule, and archive treatment.

295.16.2 Correction shall be triggered by source correction, model correction, prompt defect, data-rights change, public authority clarification, protected knowledge restriction, hallucination, compute error, workflow defect, security incident, or public-safe defect.

295.17 No Proof Receipt as Public Authority Decision, Certification, Recognition, Finance-Readiness, Procurement Approval, Rating, Public Warning, or Emergency Command.

295.17.1 No Proof Receipt, Inference Record, Compute Workload Record, Verifiable Compute artifact, Verifiable Intelligence Output, execution log, timestamp, hash, signed artifact, dashboard output, model output, benchmark output, or technical receipt shall constitute or be represented as public authority decision, certification, recognition, finance-readiness, insurance-readiness, procurement approval, rating, investment suitability, public finance approval, legal compliance approval, public warning, emergency command, safety approval, provider preference, guarantee, or enterprise execution.

295.17.2 Where such artifacts may be misread, the Corporation shall apply limitation language, restrict access, redesign the artifact, correct the claim, or withdraw the artifact.

295.18 Inference, Compute, Proof Receipt, and Verifiable Intelligence Records.

295.18.1 The Corporation shall maintain Inference, Compute, Proof Receipt, and Verifiable Intelligence Records, including inference records, compute workload records, verifiable compute records, proof receipt records, verifiable intelligence output records, material AI output records, input data records, authority records, environment records, model / code / tool / workflow records, execution records, output records, reviewer records, confidence / limitation / classification records, correction paths, boundary limitation records, corrections, supersessions, withdrawals, and archive records.

---

Section 296. Human Review, AI Output Limits, Agentic AI Controls, and Automation Boundaries

296.1 Human Review Purpose.

296.1.1 Human review shall ensure that material AI outputs, automated outputs, agentic actions, routing decisions, public-safe drafts, evidence classifications, compute outputs, and technical artifacts remain subject to competent human responsibility, institutional judgment, lawful authority, public-safe review, and correctionability.

296.1.2 Human review shall preserve research integrity, evidence integrity, public authority boundary discipline, finance-boundary discipline, certification-boundary discipline, recognition-boundary discipline, procurement neutrality, provider neutrality, privacy, cybersecurity, protected knowledge safeguards, civil rights, accessibility, and non-execution.

296.2 Qualified Human Review for Material Outputs.

296.2.1 Material AI outputs shall be reviewed by qualified human reviewers with sufficient competence, authority, access, time, and independence to assess factual support, source grounding, classification, limitations, uncertainty, public-safe status, data rights, protected knowledge, public authority meaning, and boundary risks.

296.2.2 Qualified review shall not be satisfied by nominal approval, automated acceptance, superficial reading, unrecorded chat review, sponsor review, provider review, or unsupported reliance on model confidence.

296.3 Evidence Output Review.

296.3.1 AI-assisted evidence outputs, evidence classifications, summaries, extractions, anomaly flags, confidence suggestions, and corroboration suggestions shall be reviewed before use in material research, publications, public authority learning, dashboards, maps, technical baselines, or Nexus-facing inputs.

296.3.2 Evidence Output Review shall verify source support, provenance, permission, classification, hallucination risk, uncertainty, limitations, public-safe status, and correction path.

296.4 Public Claim Review.

296.4.1 Public claims generated, assisted, summarized, translated, or routed by AI shall be reviewed for accuracy, evidence support, limitation language, legal boundary discipline, public authority boundary discipline, finance boundary discipline, certification boundary discipline, recognition boundary discipline, procurement neutrality, provider neutrality, and public-safe framing.

296.4.2 Public claims shall not be released merely because AI produced confident language or because prior similar language was approved.

296.5 Public Authority Material Review.

296.5.1 AI-assisted materials involving public authorities shall receive human review for capacity classification, official-status ambiguity, public records implications, procurement implications, grant implications, regulatory implications, public finance implications, emergency-management implications, public warning implications, and sovereign-obligation implications.

296.5.2 Public authority materials shall not be sent, published, summarized, or relied upon through autonomous AI action unless specifically authorized and reviewed.

296.6 Docket and Grid Input Review.

296.6.1 AI-assisted Docket or Grid inputs shall be reviewed to ensure that they are evidence-supported, properly classified, limitation-aware, non-approving, non-guaranteeing, non-certifying, non-recognizing, non-finance-readiness, and correctionable.

296.6.2 Docket or Grid input review shall confirm that AI outputs do not create routeability approval, Grid guarantee, Docket approval, public authority approval, procurement approval, provider preference, or execution authority.

296.7 GRF and GRA Interface Output Review.

296.7.1 AI-assisted outputs routed to or referencing The Global Risks Forum (GRF) or The Global Risks Alliance (GRA) shall be reviewed for role separation, legal separateness, recognition-boundary discipline, finance-boundary discipline, public-safe claims, and non-execution.

296.7.2 AI outputs shall not represent GCRI US evidence as GRF recognition or GRA finance-readiness unless a competent separate GRF or GRA record supports that status.

296.8 Publication Review.

296.8.1 AI-assisted publications, public-safe summaries, dashboards, maps, model cards, system cards, benchmark cards, technical baselines, Academy materials, donor reports, sponsor reports, and public authority learning materials shall receive publication review before external release.

296.8.2 Publication Review shall include accuracy, source support, hallucination, bias, limitations, uncertainty, public-safe status, data / AI / cyber review, protected knowledge review, public authority boundary review, and correction path.

296.9 Agentic AI Tool Permission Controls.

296.9.1 Agentic AI systems capable of taking actions, using tools, querying systems, sending communications, modifying records, accessing repositories, creating files, opening tickets, triggering workflows, publishing content, or executing code shall be permissioned through explicit tool controls.

296.9.2 Agentic AI permissions shall be limited to approved tools, approved data classes, approved systems, approved actions, approved time periods, and approved users.

296.9.3 Agentic systems shall not receive broad or standing authority where narrower permissions, sandboxing, read-only access, approval gates, or human-in-the-loop controls can satisfy the purpose.

296.10 Agentic AI Logs.

296.10.1 Agentic AI actions shall be logged proportionate to risk, including prompts or instruction descriptions where safe, tools invoked, systems accessed, records accessed, actions attempted, actions completed, outputs generated, approvals obtained, errors, and human overrides.

296.10.2 Agentic logs shall be protected, retained, reviewed, and used for incident response, correction, monitoring, and accountability.

296.11 Approval Gates.

296.11.1 Approval Gates shall be required before agentic AI performs material actions affecting publication, external communication, public authority materials, data access changes, repository release, software release, technical baseline release, dataset release, dashboard release, map release, controlled-room access, financial systems, contract routing, or sensitive records.

296.11.2 Approval Gates shall require competent human approval, recorded authority, and conditions appropriate to risk.

296.12 Prohibited Agent Actions.

296.12.1 Agentic AI shall be prohibited from autonomously:

296.12.1(a) publishing, releasing, or externally circulating material outputs; 296.12.1(b) sending public authority communications that may imply institutional position or public authority meaning; 296.12.1(c) making finance, procurement, certification, recognition, rating, public warning, or emergency-command statements; 296.12.1(d) granting, modifying, or revoking sensitive access without approval; 296.12.1(e) deleting, altering, sealing, or withdrawing records without approval; 296.12.1(f) committing production code, releasing software, rotating secrets, or changing security settings without approval except under approved emergency automation; 296.12.1(g) processing prohibited data classes through unapproved tools; and 296.12.1(h) taking any action reserved to the Board, officers, public authorities, counsel, auditors, or other competent human actors.

296.13 Kill Switches.

296.13.1 Material agentic AI systems shall include practical suspension, disablement, revocation, or kill-switch mechanisms proportionate to risk.

296.13.2 Kill switches may include access revocation, tool disablement, API key revocation, workflow pause, model-use hold, network restriction, repository lock, publication hold, and administrator override.

296.13.3 Kill-switch use shall be logged and reviewed after activation.

296.14 No Autonomous Public Authority Communication Without Authorization.

296.14.1 AI systems shall not autonomously communicate with public authorities, public officials, regulators, procurement bodies, grantors, emergency-management bodies, public finance actors, or public authority participants in a manner that states, implies, commits, or negotiates institutional position unless expressly authorized by competent human authority.

296.14.2 Automated scheduling, clerical routing, or receipt acknowledgment may be permitted only where it cannot reasonably be interpreted as substantive public authority communication or commitment.

296.15 No Autonomous Finance, Procurement, Certification, Recognition, Public Warning, or Public-Safe Publication Action.

296.15.1 AI systems shall not autonomously issue finance-readiness materials, investment-facing materials, procurement recommendations, provider rankings, certification-like statements, recognition-like statements, public warnings, emergency commands, public-safe publications, external reports, donor reports, sponsor reports, Docket inputs, Grid inputs, GRF-facing inputs, or GRA-facing inputs.

296.15.2 Any AI-assisted draft in these areas shall remain draft until reviewed and approved by competent human authority.

296.16 AI Output Limitation Language.

296.16.1 AI-assisted outputs shall include limitation language where necessary to prevent misleading reliance, including statements that outputs are AI-assisted, source-dependent, human-reviewed or not yet human-reviewed, preliminary, limited, non-certifying, non-recognizing, non-finance-readiness, non-procurement, non-public-authority, non-public-warning, and correctionable.

296.16.2 Limitation language shall not be used to excuse insufficient review, unsafe publication, privacy violation, protected knowledge exposure, or boundary overclaim.

296.17 Human Review and Automation Boundary Records.

296.17.1 The Corporation shall maintain Human Review and Automation Boundary Records, including human review policies, qualified reviewer records, evidence output reviews, public claim reviews, public authority material reviews, Docket and Grid input reviews, GRF and GRA interface output reviews, publication reviews, agentic AI permission records, agentic AI logs, approval gate records, prohibited action records, kill-switch records, autonomous communication restrictions, AI output limitation language, incidents, corrections, restrictions, suspensions, and archive records.

---

Section 297. Bias, Civil Rights, Accessibility, Safety, Hallucination, Drift, Prompt Injection, and Red-Team Review

297.1 Bias Review Purpose.

297.1.1 Bias Review shall identify, assess, mitigate, disclose where appropriate, and correct bias in data, AI systems, models, embeddings, retrieval systems, evidence classifications, observability outputs, dashboards, maps, public authority learning materials, technical baselines, and public-safe publications.

297.1.2 Bias Review shall address source bias, selection bias, measurement bias, sampling bias, labeling bias, historical bias, algorithmic bias, geographic bias, language bias, disability-related bias, community underrepresentation, public authority data bias, and sponsor or provider influence.

297.2 Civil Rights Review.

297.2.1 Civil Rights Review shall assess whether data use, AI-assisted processing, research outputs, dashboards, maps, classification systems, public-safe publications, or public authority learning materials may create discriminatory impact, unequal access, stigmatization, exclusion, retaliation risk, disparate error, or misuse against protected classes or rights-bearing persons.

297.2.2 Where civil rights risk is material, the Corporation shall redesign, restrict, aggregate, correct, withdraw, or refuse the activity unless sufficient mitigation exists.

297.3 Accessibility Review.

297.3.1 Accessibility Review shall ensure that digital tools, AI outputs, dashboards, maps, reports, learning materials, controlled-room procedures, grievance pathways, publications, and technical assets are accessible where required or appropriate.

297.3.2 Accessibility Review shall address disability access, plain language where appropriate, language access, assistive technology compatibility, captioning, alternative text, readable formats, and inclusive participation.

297.4 Safety Review.

297.4.1 Safety Review shall assess whether data, AI systems, outputs, technical baselines, dashboards, maps, publications, cyber materials, geospatial materials, biosecurity materials, infrastructure materials, or public authority learning materials could increase harm, misuse, panic, unsafe reliance, operational exposure, protected knowledge exposure, or public authority confusion.

297.4.2 Safety Review may require restriction, redaction, controlled release, public-safe framing, additional human review, public authority boundary language, or non-public handling.

297.5 Hallucination Review.

297.5.1 Hallucination Review shall assess whether AI outputs contain fabricated facts, sources, citations, legal claims, public authority actions, technical standards, data values, research findings, method descriptions, confidence claims, or quotations.

297.5.2 AI outputs that cannot be grounded in source records shall be corrected, rejected, restricted, or labeled as unsupported and shall not be used in material outputs.

297.6 Model Drift Review.

297.6.1 Model Drift Review shall assess whether model performance, output quality, bias, hallucination risk, retrieval behavior, classification behavior, or safety characteristics have materially changed over time.

297.6.2 Drift may result from model updates, provider changes, retrieval index changes, data distribution changes, prompt changes, tool changes, system integration changes, or user behavior changes.

297.6.3 Material drift shall require re-evaluation, restriction, monitoring, limitation update, or model retirement.

297.7 Prompt Injection Review.

297.7.1 Prompt Injection Review shall assess whether AI systems are vulnerable to malicious, manipulative, hidden, indirect, embedded, retrieved, document-based, webpage-based, dataset-based, user-supplied, or tool-supplied instructions that override approved system rules, expose data, alter outputs, misuse tools, or bypass access controls.

297.7.2 Prompt injection controls may include input sanitization, source isolation, instruction hierarchy, tool permission limits, retrieval filtering, output inspection, sandboxing, human review, and monitoring.

297.8 Data Poisoning Review.

297.8.1 Data Poisoning Review shall assess whether training data, fine-tuning data, embeddings, retrieval indexes, datasets, logs, telemetry, public submissions, provider data, sponsor data, or public authority inputs have been manipulated to distort AI outputs, classifications, evidence conclusions, or observability results.

297.8.2 Suspected poisoned data shall be held, quarantined, investigated, corrected, restricted, deleted where appropriate, or excluded from reliance.

297.9 Adversarial Input Review.

297.9.1 Adversarial Input Review shall assess attempts to manipulate AI systems, evidence systems, dashboards, maps, benchmarks, public-safe reports, or technical baselines through crafted inputs, spoofed signals, evasion, jailbreaks, prompt attacks, benchmark gaming, false metadata, manipulated documents, or malicious files.

297.9.2 Adversarial inputs shall be logged, investigated, and used to strengthen safeguards where appropriate.

297.10 Red-Team Review.

297.10.1 Red-Team Review may be used to evaluate data systems, AI systems, models, controlled rooms, public-good software, dashboards, maps, public-safe publications, evidence workflows, cyber controls, and technical baselines against misuse, bias, hallucination, leakage, prompt injection, data poisoning, spoofing, boundary overclaim, and public-safe failure.

297.10.2 Red-team activities shall be authorized, scoped, logged, safe, legally compliant, and controlled to avoid unauthorized access, harm, or public confusion.

297.11 Public Authority Impact Review.

297.11.1 Public Authority Impact Review shall assess whether data or AI outputs may be misunderstood as official adoption, public authority decision, regulatory finding, procurement action, grant action, public finance approval, emergency command, public warning, or sovereign obligation.

297.11.2 Outputs with public authority impact risk shall include capacity classification, limitation language, human review, and public-safe handling.

297.12 Community Safeguards Review.

297.12.1 Community Safeguards Review shall assess whether data or AI outputs may expose, stigmatize, misrepresent, exclude, retaliate against, or otherwise harm communities, vulnerable populations, protected participants, local groups, or rights-bearing persons.

297.12.2 Safeguards measures may include aggregation, non-attribution, redaction, delayed release, controlled access, community review, grievance pathways, correction, or withdrawal.

297.13 Protected Knowledge Review.

297.13.1 Protected Knowledge Review shall assess whether data or AI systems may expose, extract, translate, generalize, embed, infer, map, train on, publish, or transfer Tribal / Indigenous, local, territorial, cultural, environmental, sacred, ecological, community-protected, or otherwise protected knowledge.

297.13.2 Where risk is material and unresolved, the activity shall be held, restricted, redesigned, or refused.

297.14 Error Logging.

297.14.1 Material errors, hallucinations, bias incidents, drift findings, prompt injection attempts, data poisoning concerns, adversarial inputs, red-team findings, civil rights concerns, accessibility defects, public authority overclaims, protected knowledge exposures, and public-safe failures shall be logged.

297.14.2 Error logs shall identify affected systems, affected outputs, severity, cause where known, mitigation, correction, recurrence-prevention action, and closeout status.

297.15 Mitigation.

297.15.1 Mitigation may include data correction, model restriction, prompt revision, retrieval filtering, access restriction, additional human review, bias correction, accessibility redesign, output limitation language, public-safe redaction, controlled release, monitoring, training, vendor action, or withdrawal.

297.15.2 Mitigation shall be tested where feasible and monitored where residual risk remains.

297.16 Restriction, Suspension, Retirement, or Replacement.

297.16.1 AI systems, datasets, models, embeddings, retrieval indexes, dashboards, maps, technical baselines, public-good software, workflows, or outputs may be restricted, suspended, retired, or replaced where bias, civil rights risk, accessibility defect, safety risk, hallucination, drift, prompt injection, data poisoning, adversarial vulnerability, red-team finding, protected knowledge risk, or public-safe risk cannot be adequately mitigated.

297.16.2 Retirement or replacement shall include downstream dependency review, user notice where appropriate, archive treatment, and correction of affected outputs.

297.17 Bias, Safety, Drift, and Red-Team Records.

297.17.1 The Corporation shall maintain Bias, Safety, Drift, and Red-Team Records, including bias reviews, civil rights reviews, accessibility reviews, safety reviews, hallucination reviews, model drift reviews, prompt injection reviews, data poisoning reviews, adversarial input reviews, red-team reviews, public authority impact reviews, community safeguards reviews, protected knowledge reviews, error logs, mitigation records, restriction records, suspension records, retirement records, replacement records, corrections, withdrawals, and archive records.

---

Section 298. Cybersecurity Program

298.1 Cybersecurity Program Purpose.

298.1.1 The Corporation shall maintain a Cybersecurity Program proportionate to its mission, size, systems, data classes, public authority interfaces, public-good software, AI use, controlled rooms, repositories, research activities, financial systems, vendors, and risk profile.

298.1.2 The Cybersecurity Program shall protect confidentiality, integrity, availability, resilience, evidence integrity, research integrity, public authority trust, privacy, protected knowledge, public-good technical assets, financial controls, and public-safe publication.

298.1.3 The Cybersecurity Program shall not convert the Corporation into a managed security services provider, cyber regulator, incident commander, public warning authority, cyber-certification body, cyber insurer, or cyber-compliance approver.

298.2 Cybersecurity Policy.

298.2.1 The Corporation shall adopt and maintain cybersecurity policies or equivalent controls addressing access control, acceptable use, secure configuration, endpoint security, network security, cloud security, application security, repository security, data security, logging, monitoring, vulnerability management, incident response, backup, disaster recovery, vendor security, training, and enforcement.

298.2.2 Cybersecurity policies shall be reviewed periodically and updated when systems, threats, law, data classes, AI tools, public authority interfaces, vendors, or operations materially change.

298.3 Security Baseline.

298.3.1 The Corporation shall maintain a Security Baseline defining minimum controls for systems, devices, repositories, cloud environments, AI tools, controlled rooms, collaboration platforms, financial systems, public-good software, and sensitive data environments.

298.3.2 Security Baselines may include authentication, encryption, logging, backup, patching, access review, secure configuration, vulnerability review, incident escalation, and data classification requirements.

298.4 Asset Inventory.

298.4.1 The Corporation shall maintain an Asset Inventory for material systems, devices, repositories, cloud environments, applications, AI systems, datasets, controlled rooms, credentials, domains, services, and public-good technical assets.

298.4.2 Asset Inventory records shall identify owner, custodian, classification, purpose, location, access class, vendor, security baseline, review cycle, and retirement status.

298.5 Identity and Access Management.

298.5.1 Identity and Access Management shall ensure verified identity, least privilege, role-based access, attribute-based access where appropriate, multi-factor authentication, privileged access control, access logging, timely revocation, and periodic access review.

298.5.2 Identity controls shall apply to personnel, contractors, fellows, volunteers, collaborators, vendors, providers, public authority participants, service accounts, AI agents, and automated workflows.

298.6 Secure Configuration.

298.6.1 Systems shall be configured securely according to approved baselines, including disabling unnecessary services, enforcing least privilege, protecting administrative interfaces, logging security events, applying encryption where appropriate, and hardening cloud, repository, endpoint, and application settings.

298.6.2 Configuration changes shall be controlled, logged, reviewed where material, and reversible where appropriate.

298.7 Endpoint Security.

298.7.1 Endpoint Security shall protect laptops, desktops, mobile devices, servers, developer workstations, controlled-room devices, and other endpoints used for Corporation work.

298.7.2 Endpoint controls may include device management, encryption, screen lock, patching, malware protection, local storage restrictions, remote wipe, restricted administrator rights, and prohibition of unapproved personal-device use for sensitive data.

298.8 Network Security.

298.8.1 Network Security shall protect Corporation networks, remote access, VPNs where used, wireless networks, controlled-room networks, cloud networks, APIs, and inter-system connections.

298.8.2 Network controls may include segmentation, secure remote access, firewall rules, intrusion detection where appropriate, secure DNS, monitoring, encrypted transmission, and restriction of public exposure.

298.9 Cloud Security.

298.9.1 Cloud Security shall govern cloud platforms, storage, compute, AI services, logs, identities, keys, backups, data regions, and integrations.

298.9.2 Cloud environments shall be configured for least privilege, encryption where appropriate, logging, backup, secure networking, region controls, deletion controls, vendor review, and incident response.

298.10 Application Security.

298.10.1 Application Security shall govern software developed, acquired, configured, hosted, or used by the Corporation.

298.10.2 Application controls may include secure development lifecycle, authentication, authorization, input validation, dependency review, vulnerability testing, logging, secrets management, secure APIs, privacy review, and public-safe release review.

298.11 Repository Security.

298.11.1 Repository Security shall protect code, documentation, datasets, models, technical baselines, public-good software, methods, schemas, and configuration records stored in repositories.

298.11.2 Repository controls shall include access review, branch protections where appropriate, code review, dependency review, secret scanning, signing where appropriate, issue controls, release controls, and archive controls.

298.12 Data Security.

298.12.1 Data Security shall protect data according to classification, including encryption where appropriate, access control, data loss prevention where feasible, backup, secure transfer, public-safe review, retention, deletion, and incident response.

298.12.2 Data Security shall apply to public authority data, personal information, health-sensitive data, cyber-sensitive data, infrastructure-sensitive data, controlled technology, finance-sensitive data, commercially sensitive data, research-sensitive data, and protected knowledge.

298.13 Logging and Monitoring.

298.13.1 Material systems shall generate logs proportionate to risk, including access logs, administrative logs, repository logs, cloud logs, AI-use logs, controlled-room logs, financial system logs, security logs, and incident logs.

298.13.2 Monitoring shall detect unusual access, suspicious downloads, failed logins, privilege misuse, repository secrets, anomalous AI use, data exfiltration indicators, and system compromise.

298.14 Vulnerability Management.

298.14.1 The Corporation shall maintain vulnerability management practices for material systems, software, repositories, dependencies, cloud services, AI tools, and public-good technical assets.

298.14.2 Vulnerability management may include scanning, dependency review, patching, severity classification, remediation tracking, coordinated disclosure where applicable, compensating controls, and public-safe advisories where appropriate.

298.15 Incident Response.

298.15.1 Cybersecurity incidents shall be identified, contained, investigated, remediated, reported where required, corrected, and reviewed for recurrence prevention.

298.15.2 Incident response shall protect evidence, logs, privilege, confidentiality, public authority trust, privacy, protected knowledge, operational continuity, and public-safe communication.

298.16 Backup, Disaster Recovery, and Business Continuity.

298.16.1 The Corporation shall maintain backup, disaster recovery, and business continuity controls proportionate to its systems, records, public authority interfaces, research activities, public-good technical assets, financial systems, controlled rooms, and institutional obligations.

298.16.2 Backups shall be protected, tested where appropriate, access-controlled, retention-aligned, and restorable where necessary.

298.17 Vendor Security.

298.17.1 Vendors, cloud providers, AI providers, processors, managed platforms, repository services, financial platforms, and other service providers shall be reviewed for security controls proportionate to data class and operational importance.

298.17.2 Vendor security review shall include confidentiality, security practices, incident notice, sub-processors, access controls, data location, deletion, audit rights where appropriate, and termination assistance.

298.18 Cybersecurity Training.

298.18.1 Persons with access to Corporation systems or data shall receive cybersecurity training appropriate to role and risk.

298.18.2 Training may address phishing, credential security, data classification, secure sharing, AI tool risks, repository security, protected knowledge, incident reporting, device security, and public authority data handling.

298.19 Cybersecurity Program Records.

298.19.1 The Corporation shall maintain Cybersecurity Program Records, including cybersecurity policies, Security Baselines, Asset Inventory records, identity and access records, secure configuration records, endpoint security records, network security records, cloud security records, application security records, repository security records, data security records, logging and monitoring records, vulnerability management records, incident response records, backup / disaster recovery / business continuity records, vendor security records, training records, corrections, restrictions, retirements, and archive records.

---

Section 299. Secure Development, Repository Security, Supply-Chain Assurance, SBOM, Signing, Provenance, Secrets, Keys, Tokens, and Release Controls

299.1 Secure Development Purpose.

299.1.1 Secure Development shall protect public-good software, technical baselines, schemas, APIs, SDKs, dashboards, maps, evidence systems, AI systems, controlled-room tools, repositories, documentation, and releases from security defects, supply-chain compromise, license defects, provenance defects, unauthorized access, secret exposure, and unsafe publication.

299.1.2 Secure Development shall support public-benefit reuse, anti-enclosure, provider neutrality, research integrity, evidence integrity, cybersecurity, public-safe publication, and correctionability.

299.2 Secure Development Lifecycle.

299.2.1 The Corporation shall maintain secure development lifecycle controls proportionate to system risk, including planning, design review, threat review where appropriate, code review, dependency review, testing, vulnerability review, release approval, provenance, documentation, monitoring, patching, and retirement.

299.2.2 Secure development controls shall apply to internal and external contributors where they contribute to Corporation-controlled repositories or public-good technical assets.

299.3 Code Review.

299.3.1 Material code changes shall be reviewed before release or deployment according to repository policy.

299.3.2 Code review shall assess correctness, security, privacy, data handling, AI-use implications, dependency changes, secrets exposure, license implications, public authority data implications, protected knowledge implications, and public-safe documentation.

299.4 Dependency Review.

299.4.1 Dependencies shall be reviewed for security, maintenance, provenance, license, known vulnerabilities, transitive risk, typosquatting, malicious package risk, abandonment, and compatibility.

299.4.2 High-risk dependencies may require replacement, pinning, sandboxing, additional review, or exclusion.

299.5 License Review.

299.5.1 Software, datasets, models, documentation, schemas, code snippets, libraries, and technical baselines shall be reviewed for license compatibility before use, release, redistribution, embedding, publication, or integration.

299.5.2 License Review shall preserve open technical baseline integrity, public-good availability, attribution, anti-enclosure intent, and avoidance of unauthorized proprietary use.

299.6 Vulnerability Review.

299.6.1 Material software, repositories, dependencies, APIs, dashboards, maps, controlled-room tools, AI tools, and release artifacts shall be reviewed for vulnerabilities proportionate to risk.

299.6.2 Vulnerabilities shall be classified, remediated, mitigated, accepted by competent authority with record, or disclosed through coordinated disclosure where appropriate.

299.7 SBOM Requirement Where Appropriate.

299.7.1 A Software Bill of Materials may be required for material public-good software, controlled-room tools, high-risk systems, public authority learning tools, cybersecurity-sensitive tools, or externally released technical assets.

299.7.2 SBOM records shall identify components, versions, suppliers, licenses, known vulnerabilities where available, and update status.

299.8 Artifact Signing Where Appropriate.

299.8.1 Release artifacts, software packages, containers, datasets, model artifacts, proof receipts, technical baselines, or documentation may be signed where appropriate to preserve integrity, provenance, and trust.

299.8.2 Signing keys shall be protected, access-controlled, rotated where appropriate, and revoked or replaced after compromise.

299.9 Provenance Records.

299.9.1 Provenance Records shall identify source repository, commit, build process, dependencies, contributors, review approvals, build environment, signing status, release date, and artifact hash where appropriate.

299.9.2 Provenance Records shall support traceability, correction, vulnerability response, supply-chain assurance, and technical memory.

299.10 Build Integrity.

299.10.1 Build processes shall be protected against unauthorized alteration, dependency substitution, secret exposure, malicious injection, environment drift, and artifact tampering.

299.10.2 Build Integrity controls may include controlled build environments, pinned dependencies, reproducible builds where appropriate, build logs, checksums, artifact signing, and release approvals.

299.11 Release Approval.

299.11.1 Material releases shall require release approval by competent authority.

299.11.2 Release Approval shall verify review status, tests, vulnerabilities, license status, documentation, public-safe language, data handling, AI-use restrictions, provenance, rollback plan, and correction path.

299.12 Rollback Plan.

299.12.1 Material releases shall include a rollback, withdrawal, supersession, patch, or correction plan proportionate to risk.

299.12.2 Rollback plans shall identify responsible persons, triggers, affected users, communication route, repository action, archive status, and dependency impacts.

299.13 Secrets Management.

299.13.1 Secrets, including passwords, API keys, private keys, tokens, credentials, certificates, connection strings, recovery codes, signing materials, and sensitive configuration values, shall be stored, accessed, rotated, revoked, and disposed of through approved secrets management controls.

299.13.2 Secrets shall not be committed to repositories, embedded in public artifacts, shared through unapproved channels, entered into unapproved AI systems, or stored in plaintext where secure alternatives are available.

299.14 Key Management.

299.14.1 Cryptographic keys shall be generated, stored, used, rotated, backed up, revoked, and destroyed according to sensitivity and purpose.

299.14.2 Key access shall be limited, logged where appropriate, and reviewed periodically.

299.15 Token Management.

299.15.1 Tokens shall be scoped, time-limited where feasible, access-controlled, monitored, rotated, and revoked when no longer needed or upon suspected compromise.

299.15.2 Tokens granting repository, cloud, AI, financial, controlled-room, or public authority data access shall receive heightened protection.

299.16 Credential Management.

299.16.1 Credentials shall be unique, protected, never shared except through approved service account controls, and revoked promptly when access is no longer required.

299.16.2 Credential compromise shall trigger incident response, rotation, access review, and affected-system review.

299.17 Recovery Code Management.

299.17.1 Recovery codes, backup credentials, emergency access credentials, and account recovery materials shall be protected as sensitive secrets.

299.17.2 Recovery materials shall be stored securely, access-restricted, periodically reviewed, and updated after personnel changes or suspected compromise.

299.18 Repository Access Controls.

299.18.1 Repository access shall be role-based, least-privilege, reviewed periodically, and protected by multi-factor authentication where appropriate.

299.18.2 Sensitive repositories may require branch protection, required review, signed commits where appropriate, secret scanning, dependency scanning, restricted external collaboration, no-public-fork rules, and release gates.

299.19 Secure Release Records.

299.19.1 The Corporation shall maintain Secure Release Records, including secure development lifecycle records, code review records, dependency review records, license review records, vulnerability review records, SBOM records where applicable, artifact signing records, provenance records, build integrity records, release approval records, rollback plans, secrets management records, key management records, token management records, credential management records, recovery code records, repository access control records, incidents, corrections, patches, withdrawals, supersessions, and archive records.

---

Section 300. Controlled Rooms, Clean Rooms, Data Rooms, Evidence Rooms, Public Authority Rooms, and No-Download Rooms

300.1 Controlled-Room Purpose.

300.1.1 Controlled Rooms shall provide secure, purpose-limited, access-controlled, logged, and governed environments for reviewing sensitive data, evidence, public authority materials, confidential records, protected knowledge, cyber-sensitive data, infrastructure-sensitive data, rights-bearing data, health-sensitive data, controlled technology, finance-sensitive materials, and other restricted materials.

300.1.2 Controlled Rooms shall support research integrity, evidence integrity, public authority trust, public-safe publication, data minimization, access control, privacy, cybersecurity, protected knowledge safeguards, and correctionability.

300.2 Clean-Room Purpose.

300.2.1 Clean Rooms shall allow controlled analysis or comparison of restricted data while preventing unauthorized disclosure, copying, extraction, re-identification, commingling, or reuse.

300.2.2 Clean Rooms may support compute-to-data, output review, privacy-preserving analysis, cross-institution comparison, public authority review, or protected knowledge review.

300.3 Data-Room Purpose.

300.3.1 Data Rooms shall provide structured access to defined datasets, records, evidence packs, technical materials, public authority materials, or diligence-supporting materials under controlled conditions.

300.3.2 Data Rooms shall not be used to create investment solicitation, finance-readiness, underwriting approval, rating, procurement approval, certification, recognition, or public authority approval.

300.4 Evidence-Room Purpose.

300.4.1 Evidence Rooms shall provide controlled review environments for evidence packs, source records, provenance records, custody records, validation materials, review notes, challenge materials, correction materials, and public-safe publication support.

300.4.2 Evidence Room access shall preserve evidence integrity, reviewer independence, source restrictions, correctionability, and public-safe handling.

300.5 Public Authority Room Purpose.

300.5.1 Public Authority Rooms shall support lawful public authority learning, review, capacity-classified participation, evidence literacy, methods literacy, observability literacy, controlled viewing, and public-safe interpretation.

300.5.2 Public Authority Rooms shall not create public authority delegation, official adoption, regulatory approval, procurement approval, grant approval, public finance approval, public warning, emergency command, sovereign obligation, or public-private partnership unless separately created by the competent public authority through lawful process outside the Corporation’s non-execution role.

300.6 No-Download Room Purpose.

300.6.1 No-Download Rooms shall permit authorized viewing, querying, analysis, or review while prohibiting downloading, copying, exporting, scraping, photographing, embedding, training, or transferring underlying materials.

300.6.2 No-Download Rooms shall include user obligations, technical controls where feasible, logging, output review, watermarking where appropriate, and access revocation rules.

300.7 Room Charter.

300.7.1 Each material Controlled Room shall have a Room Charter identifying purpose, scope, room type, data classes, permitted participants, prohibited participants where applicable, access conditions, permitted uses, prohibited uses, owner, custodian, review process, logging, output review, publication rules, closeout rules, and correction path.

300.7.2 The Room Charter shall include non-endorsement, non-certification, non-recognition, non-finance-readiness, non-procurement, non-public-authority, non-public-warning, and non-execution language where necessary.

300.8 Room Owner.

300.8.1 The Room Owner shall be responsible for the institutional purpose, scope, authorization, risk classification, participant eligibility, and continuing appropriateness of the Controlled Room.

300.8.2 Room Owner status shall not create private control, sponsor control, provider control, public authority authority, finance-readiness authority, certification authority, recognition authority, procurement authority, or publication authority beyond recorded delegation.

300.9 Room Custodian.

300.9.1 The Room Custodian shall be responsible for operational custody, access controls, logs, exhibit handling, technical configuration, user notices, output review routing, incident escalation, closeout, and records.

300.9.2 The Room Custodian shall ensure that room controls remain aligned with data classification, Room Charter, access rules, public-safe rules, and correction obligations.

300.10 Room Access Criteria.

300.10.1 Access to a Controlled Room shall be based on authorized purpose, need to know, eligibility, role, data class, confidentiality obligation, public authority capacity, conflict review, security review, safeguards review, and agreement to room rules.

300.10.2 Room access shall not be granted as a donor benefit, sponsor benefit, provider benefit, investor benefit, public authority privilege, media privilege, relationship benefit, or commercial advantage.

300.11 Room Participant Screening.

300.11.1 Participants may be screened for identity, affiliation, role, authority, conflicts, confidentiality obligations, export-control or sanctions considerations where applicable, public authority capacity, security risk, and safeguards risk.

300.11.2 Screening shall be proportionate, non-discriminatory, privacy-aware, and recorded.

300.12 Capacity Classification.

300.12.1 Participants shall be capacity-classified where their affiliation or role may affect meaning, including public authority participants, public officials, regulators, procurement personnel, grant personnel, public finance readers, sponsors, providers, investors, national company participants, Project SPV participants, community representatives, and academic collaborators.

300.12.2 Capacity classification shall be reflected in room records and publication language where material.

300.13 Conflict Review.

300.13.1 Room participation shall be reviewed for conflicts of interest, sponsor influence, provider influence, procurement conflicts, finance conflicts, public authority conflicts, research conflicts, reviewer conflicts, and related-party interests.

300.13.2 Conflict mitigation may include recusal, access restriction, separate rooms, redaction, independent review, nondisclosure, or exclusion.

300.14 Competition and Antitrust Review.

300.14.1 Controlled Rooms involving competitors, providers, vendors, industry participants, sponsors, capital actors, national companies, Project SPVs, or market participants shall be reviewed for competition and antitrust risk where applicable.

300.14.2 Room rules shall prohibit improper coordination, price discussion, market allocation, bid coordination, procurement collusion, exchange of competitively sensitive information beyond approved scope, or misuse of public-good technical processes for market advantage.

300.15 Public Authority Boundary Review.

300.15.1 Controlled Rooms involving public authorities shall be reviewed for public authority boundary risk, public records implications, open meetings, procurement rules, grant rules, government ethics, official-capacity ambiguity, public finance implications, public warning implications, and emergency-command implications.

300.15.2 Public authority participation shall not be represented as endorsement, official adoption, procurement approval, grant approval, regulatory action, public finance approval, public warning, emergency command, or sovereign decision.

300.16 Finance Boundary Review.

300.16.1 Controlled Rooms involving capital readers, funders, insurers, lenders, underwriters, banks, public finance actors, national companies, Project SPVs, or finance-facing materials shall be reviewed for finance-boundary risk.

300.16.2 Finance Boundary Review shall prevent room materials from being used as investment advice, securities offering, broker-dealer activity, finder activity, rating, underwriting approval, insurance approval, bankability, investability, public finance approval, or GRA action unless separately and lawfully authorized by the competent body outside the Corporation’s non-finance role.

300.17 Data / AI / Cyber / Privacy Review.

300.17.1 Controlled Rooms shall be reviewed for data classification, privacy, AI-use restrictions, cyber sensitivity, infrastructure sensitivity, access control, logging, output review, cross-border access, export-control, sanctions, and incident response.

300.17.2 AI tools shall not be used in a Controlled Room unless approved for the room’s data classes and expressly permitted by the Room Charter.

300.18 Protected Knowledge Review.

300.18.1 Controlled Rooms containing Community-Protected Data, Tribal / Indigenous Data, local knowledge, territorial knowledge, cultural knowledge, environmental knowledge, sacred knowledge, or Protected Knowledge shall receive Protected Knowledge Review.

300.18.2 Protected Knowledge Review may require participant restriction, non-attribution, no-download access, redaction, community or Tribal review where applicable, AI-use prohibition, mapping prohibition, and special correction pathways.

300.19 Room Logging and Monitoring.

300.19.1 Controlled Room access and material activity shall be logged proportionate to risk.

300.19.2 Logs may include participant access, time, materials viewed, searches, downloads where permitted, exports, output requests, AI use where permitted, notes submitted, exhibits added, administrative changes, and incidents.

300.19.3 Monitoring shall protect integrity, confidentiality, privacy, public authority trust, protected knowledge, and cybersecurity, and shall not be used for retaliation, improper surveillance, or discrimination.

300.20 Room Exhibit Handling.

300.20.1 Materials placed in a Controlled Room shall be registered as room exhibits or equivalent records, including title, source, data class, authority, permissions, restrictions, public-safe status, access class, version, custodian, and correction path.

300.20.2 Room exhibits shall not be copied, exported, cited, summarized, trained on, embedded, published, or transferred outside room rules.

300.21 Room Closeout.

300.21.1 Controlled Room closeout shall identify final status of participants, access revocation, exhibit disposition, output review, correction obligations, deletion or retention requirements, public-safe publication status, unresolved restrictions, incidents, and archive status.

300.21.2 Closeout shall occur when room purpose is complete, authority expires, risk changes, project ends, access is no longer needed, or competent authority orders closure.

300.22 No Room Participation as Endorsement, Certification, Recognition, Finance-Readiness, Procurement Approval, Public Authority Decision, Rating, Public Warning, or Execution Authority.

300.22.1 Participation in, access to, contribution to, hosting of, sponsorship of, review within, or receipt of materials from a Controlled Room, Clean Room, Data Room, Evidence Room, Public Authority Room, or No-Download Room shall not constitute or be represented as endorsement, certification, recognition, finance-readiness, insurance-readiness, procurement approval, public authority decision, rating, public warning, emergency command, legal compliance approval, provider preference, investment suitability, public finance approval, guarantee, partnership, agency, or execution authority.

300.22.2 Room materials and participant communications shall include limitation language where necessary to prevent misuse.

300.23 Controlled-Room Records.

300.23.1 The Corporation shall maintain Controlled-Room Records, including Controlled Room purpose records, Clean Room records, Data Room records, Evidence Room records, Public Authority Room records, No-Download Room records, Room Charters, Room Owner records, Room Custodian records, access criteria, participant screening records, capacity classifications, conflict reviews, competition and antitrust reviews, public authority boundary reviews, finance boundary reviews, data / AI / cyber / privacy reviews, protected knowledge reviews, logs, monitoring records, exhibit records, output review records, incident records, closeout records, limitation records, corrections, restrictions, deletions, and archive records.

Section 301. Third-Party Technology Providers, Data Processors, AI Providers, Cloud Providers, Cybersecurity Providers, Repository Providers, and Vendor Security

301.1 Third-Party Technology Review Purpose.

301.1.1 Third-party technology review shall ensure that any vendor, provider, processor, platform, cloud service, AI service, cybersecurity service, repository service, hosting service, software service, analytics service, collaboration service, payment service, identity service, or other external technology provider used by the Corporation is reviewed, approved, contracted, monitored, restricted, exited, and recorded in a manner consistent with the Corporation’s public-benefit purpose, legal obligations, data governance, privacy duties, AI governance, cybersecurity requirements, public authority trust, protected knowledge safeguards, research integrity, evidence integrity, public-safe publication, and correctionability.

301.1.2 Third-party technology review shall apply before material access is granted to Corporation data, public authority data, personal information, health-sensitive data, rights-bearing data, cyber-sensitive data, infrastructure-sensitive data, community-protected data, protected knowledge, confidential information, privileged information, finance-sensitive information, research-sensitive information, public-good software, technical baselines, repositories, controlled rooms, models, inference records, compute workloads, dashboards, maps, or systems.

301.1.3 No third-party provider shall be treated as approved merely because it is widely used, well-known, recommended by a sponsor, used by a public authority, used by a university, bundled with another service, low-cost, technically convenient, previously used informally, or operationally urgent.

301.2 Data Processor Review.

301.2.1 Data processor review shall determine whether a third party that processes data for or on behalf of the Corporation has adequate legal, contractual, technical, organizational, privacy, cybersecurity, incident-response, retention, deletion, confidentiality, subprocessing, and audit controls for the applicable data classes and purposes.

301.2.2 Data processor review shall assess permitted processing, prohibited processing, data location, access by processor personnel, onward transfer, subprocessor use, AI use, model training, model improvement, logging, data retention, deletion, return, breach notification, security controls, confidentiality, and termination assistance.

301.2.3 Data processors shall not use Corporation data for their own purposes, analytics, product development, AI training, model improvement, benchmarking, marketing, resale, external sharing, or sponsor or provider benefit unless expressly authorized by competent record and lawful basis.

301.3 AI Provider Review.

301.3.1 AI provider review shall assess any third-party AI model, AI platform, AI tool, agentic system, transcription system, translation system, embedding service, retrieval service, summarization system, classification system, anomaly detection system, model evaluation platform, or AI-enabled workflow before material use.

301.3.2 AI provider review shall include review of model use terms, training terms, no-training controls, fine-tuning controls, prompt retention, output retention, inference logging, human review by provider personnel, embeddings, model improvement, subprocessor access, data location, cross-border processing, security, confidentiality, deletion, IP, incident notice, bias risks, hallucination risks, prompt-injection risks, data leakage risks, and public-safe output controls.

301.3.3 AI providers shall not be used for restricted, confidential, privileged, public authority, health-sensitive, rights-bearing, cyber-sensitive, infrastructure-sensitive, controlled technology, community-protected, or protected knowledge data unless the provider is approved for that data class and all required no-training, confidentiality, access, logging, deletion, and human-review controls are in place.

301.4 Cloud Provider Review.

301.4.1 Cloud provider review shall assess cloud infrastructure, storage, compute, networking, identity, logging, backup, AI services, database services, analytics services, serverless services, key-management services, and managed platforms used by the Corporation.

301.4.2 Cloud provider review shall consider data location, region controls, encryption, key management, access control, logging, monitoring, backup, disaster recovery, isolation, privileged access, subprocessor use, compliance commitments, breach notice, audit rights where appropriate, deletion, termination assistance, and service continuity.

301.4.3 Cloud configurations shall be governed by least privilege, secure baseline, data classification, region restriction where required, public authority restrictions, cross-border controls, controlled-room rules, and incident-response readiness.

301.5 Cybersecurity Provider Review.

301.5.1 Cybersecurity provider review shall assess providers of security tooling, managed detection, vulnerability scanning, penetration testing, incident response, identity security, endpoint security, logging, monitoring, repository security, cloud security, threat intelligence, or related cyber services.

301.5.2 Cybersecurity providers may receive access to highly sensitive systems, logs, vulnerabilities, credentials, configurations, incident records, public authority data, protected knowledge, or infrastructure-sensitive data only under written authority, least-privilege controls, confidentiality, logging, public-safe handling, and incident-response conditions.

301.5.3 Use of a cybersecurity provider shall not convert the Corporation into a managed security services provider, public authority cyber command, cyber certification body, incident commander, regulator, or cyber-insurance authority.

301.6 Repository Provider Review.

301.6.1 Repository provider review shall assess providers used for code, documents, datasets, models, schemas, technical baselines, public-good software, issue tracking, release management, package publishing, version control, and collaboration.

301.6.2 Repository provider review shall address access controls, multi-factor authentication, branch protections, secret scanning, dependency alerts, audit logs, external collaborator controls, public / private repository controls, release controls, artifact signing support, backup, deletion, export, incident notice, and account recovery.

301.6.3 Repository providers shall be configured to prevent accidental public release of restricted code, secrets, credentials, public authority data, protected knowledge, controlled technology, confidential materials, or sensitive technical baselines.

301.7 Software Vendor Review.

301.7.1 Software vendor review shall assess vendors providing software used for research, evidence management, document management, records, communications, collaboration, finance, HR, CRM, learning, dashboards, maps, analytics, development, security, AI, or administrative functions.

301.7.2 Software vendor review shall consider data classes processed, privacy terms, AI features, default training or analytics, integrations, access controls, export options, deletion options, security controls, support access, subprocessor use, incident notice, IP terms, accessibility, availability, and exit risks.

301.7.3 Software vendors shall not be approved where their terms, architecture, security posture, or AI-use defaults materially conflict with Corporation obligations unless risks are remediated or use is restricted to non-sensitive contexts.

301.8 Hosting Provider Review.

301.8.1 Hosting provider review shall assess providers hosting websites, publications, portals, dashboards, maps, APIs, data rooms, controlled rooms, applications, public-good software documentation, Academy materials, repositories, or other public or controlled interfaces.

301.8.2 Hosting review shall include availability, security, access controls, logging, content takedown, backup, data location, CDN use, privacy, cookies or tracking, incident response, defacement risk, denial-of-service risk, and public-safe publication controls.

301.8.3 Public-facing hosting shall be configured to prevent unauthorized publication, misleading outdated content, uncontrolled indexing of restricted materials, and public authority, finance, certification, recognition, procurement, or public-warning overclaim.

301.9 Subprocessor Review.

301.9.1 Subprocessor review shall identify third parties used by vendors or providers to process, host, store, transmit, support, secure, analyze, or access Corporation data.

301.9.2 Subprocessor review shall assess subprocessor identity, location, role, data classes, security, privacy, AI use, onward transfer, incident notice, deletion, audit, and contractual flow-down obligations.

301.9.3 The Corporation may prohibit, restrict, require notice of, or require approval for subprocessors where data sensitivity, public authority restrictions, protected knowledge, cyber risk, cross-border transfer, sanctions, export-control, or privacy obligations require heightened control.

301.10 Contractual Security Requirements.

301.10.1 Third-party technology contracts shall include security requirements proportionate to data class, system criticality, public authority obligations, AI use, privacy risk, cyber risk, protected knowledge, controlled technology, and public-safe publication risk.

301.10.2 Contractual security requirements may include confidentiality, least privilege, encryption, access logging, vulnerability management, secure development, incident notice, breach cooperation, deletion, return, subprocessor controls, audit or assurance rights, business continuity, backup, and termination assistance.

301.10.3 Where a vendor’s standard terms are insufficient, the Corporation shall restrict use, negotiate supplemental terms, avoid sensitive data, use compensating controls, or reject the vendor.

301.11 Data Protection Addendum Where Appropriate.

301.11.1 A data protection addendum, data processing agreement, business associate agreement where applicable, public authority data addendum, research data agreement, controlled-room agreement, or equivalent instrument shall be used where required or appropriate.

301.11.2 Such instruments shall address data categories, purpose, lawful basis, processing instructions, confidentiality, security, subprocessing, cross-border transfer, data subject rights, breach notice, retention, deletion, return, audit, AI-use restrictions, and protected knowledge restrictions where applicable.

301.12 AI Use and Training Restrictions.

301.12.1 Third-party technology providers shall be prohibited from using Corporation data for AI training, fine-tuning, model improvement, embeddings, benchmarking, product development, analytics, or human review beyond support obligations unless expressly authorized by competent record.

301.12.2 Contracts and configurations shall disable default training, retention, model improvement, or telemetry use where sensitive or confidential data is involved, unless competent authority approves a bounded exception.

301.12.3 Vendor AI features shall be treated as AI systems subject to the Corporation’s AI governance program, even where embedded in ordinary office, repository, cloud, security, or collaboration tools.

301.13 Cybersecurity Requirements.

301.13.1 Third-party providers shall maintain cybersecurity controls appropriate to the services provided and the data processed.

301.13.2 Required controls may include secure identity, multi-factor authentication, access logging, encryption, vulnerability management, secure development, incident response, backup, disaster recovery, business continuity, personnel screening where appropriate, security training, and third-party assurance reports where appropriate.

301.14 Incident Notification Requirements.

301.14.1 Third-party providers shall be required, where appropriate, to notify the Corporation promptly of suspected or confirmed incidents affecting Corporation data, systems, accounts, repositories, AI outputs, cloud environments, controlled rooms, public authority data, protected knowledge, or service availability.

301.14.2 Incident notice requirements shall define timing, content, escalation pathway, cooperation obligations, evidence preservation, remediation support, notification support, and post-incident reporting.

301.15 Audit and Assurance Rights Where Appropriate.

301.15.1 Contracts with material third-party technology providers shall include audit, assurance, certification-review, security-report, penetration-test summary, compliance-attestation, or equivalent rights where appropriate to risk and negotiating practicality.

301.15.2 Audit and assurance rights shall be used to support governance and shall not be represented as certification, public authority approval, procurement approval, finance-readiness, or guarantee by the Corporation.

301.16 Exit and Portability Rights.

301.16.1 Third-party technology arrangements shall include exit and portability rights where appropriate, including data export, data deletion, return of records, migration assistance, account closure, repository transfer, key revocation, backup deletion, and post-termination confidentiality.

301.16.2 Exit planning shall address vendor lock-in, public-good technical asset continuity, repository continuity, data rights, public authority restrictions, protected knowledge, AI embeddings, model records, logs, and controlled-room closure.

301.17 Vendor Monitoring.

301.17.1 The Corporation shall monitor material third-party technology providers proportionate to risk.

301.17.2 Vendor monitoring may include periodic security review, contract review, subprocessor updates, incident monitoring, service availability review, privacy terms review, AI-terms review, access review, data deletion verification, and renewal review.

301.17.3 Material vendor changes, deteriorating risk posture, unfavorable AI-use term changes, security incidents, public authority restrictions, sanctions or export-control concerns, or unacceptable subprocessor changes may require suspension, restriction, replacement, or termination.

301.18 Third-Party Technology Provider Records.

301.18.1 The Corporation shall maintain Third-Party Technology Provider Records, including technology review records, data processor reviews, AI provider reviews, cloud provider reviews, cybersecurity provider reviews, repository provider reviews, software vendor reviews, hosting provider reviews, subprocessor reviews, contractual security requirements, data protection addenda, AI-use and training restriction records, cybersecurity requirement records, incident notification terms, audit and assurance records, exit and portability records, vendor monitoring records, suspensions, terminations, corrections, restrictions, and archive records.

---

Section 302. Data, AI, Cybersecurity, Privacy, and Technical Incident Taxonomy

302.1 Incident Taxonomy Purpose.

302.1.1 The Corporation shall maintain an incident taxonomy to classify, triage, investigate, contain, correct, notify, and learn from data, AI, cybersecurity, privacy, public authority, protected knowledge, technical, publication, repository, supply-chain, controlled-room, and public-safe incidents.

302.1.2 The taxonomy shall support consistent escalation, severity classification, legal review, privilege protection, public-safe communication, regulatory assessment, public authority trust, data subject protection, community safeguards, evidence preservation, correctionability, and recurrence prevention.

302.1.3 Incident classification shall be based on substance, risk, affected data, affected systems, affected persons, affected communities, affected public authorities, affected outputs, and potential harm, not merely on the label used by the person reporting the event.

302.2 Data Incident.

302.2.1 A Data Incident means any suspected or confirmed event involving unauthorized, unlawful, accidental, unsafe, or policy-inconsistent access, collection, use, disclosure, alteration, deletion, loss, corruption, transfer, publication, classification, retention, AI processing, or disposal of data.

302.2.2 Data Incidents may involve public data, internal data, confidential data, restricted data, rights-bearing data, public authority data, research data, evidence records, datasets, logs, model outputs, embeddings, dashboards, maps, repositories, or technical baselines.

302.3 Privacy Incident.

302.3.1 A Privacy Incident means any suspected or confirmed event involving personal information, sensitive personal information, rights-bearing data, consent, notice, data subject rights, privacy restriction, unauthorized processing, unauthorized disclosure, re-identification, linkage, profiling, or privacy-relevant publication risk.

302.3.2 Privacy Incidents shall be reviewed for applicable breach notification, data subject notice, public authority notice, contractual notice, correction, deletion, restriction, and recurrence-prevention obligations.

302.4 Personal Information Breach.

302.4.1 A Personal Information Breach means a suspected or confirmed incident involving unauthorized access to, acquisition of, disclosure of, loss of, alteration of, destruction of, or compromise of personal information or sensitive personal information, as determined under applicable law.

302.4.2 Personal Information Breaches shall be assessed under applicable federal, state, territorial, contractual, public authority, sectoral, and cross-border notification requirements.

302.5 Public Authority Data Incident.

302.5.1 A Public Authority Data Incident means any suspected or confirmed unauthorized or unsafe access, use, disclosure, alteration, deletion, publication, AI processing, transfer, or misclassification of Public Authority Data.

302.5.2 Public Authority Data Incidents shall be reviewed for public authority notice, public records implications, confidentiality, procurement implications, grant implications, public finance implications, public safety implications, emergency-management implications, public-safe communication, and public authority trust.

302.6 Health-Sensitive Data Incident.

302.6.1 A Health-Sensitive Data Incident means any suspected or confirmed incident involving health-sensitive, public health, clinical, medical, youth health, vulnerable-population, biosecurity, health-system, or health-adjacent data.

302.6.2 Such incidents shall receive heightened privacy, legal, ethics, public authority, community-safeguards, and public-safe review.

302.7 Cyber-Sensitive Data Incident.

302.7.1 A Cyber-Sensitive Data Incident means any suspected or confirmed incident involving credentials, keys, tokens, secrets, vulnerabilities, exploit details, security logs, incident records, threat intelligence, repository data, cloud data, AI-system logs, or cybersecurity-sensitive information.

302.7.2 Cyber-Sensitive Data Incidents may require immediate containment, credential rotation, key rotation, token revocation, repository hold, system isolation, vulnerability remediation, coordinated disclosure, and law enforcement or public authority interface where required or appropriate.

302.8 Infrastructure-Sensitive Data Incident.

302.8.1 An Infrastructure-Sensitive Data Incident means any suspected or confirmed incident involving critical infrastructure, utilities, telecom, energy, water, food, ports, transportation, public works, emergency systems, public health infrastructure, cyber-physical systems, geospatial layers, digital twins, AI-RAN / O-RAN systems, or operational dependency information.

302.8.2 Infrastructure-Sensitive Data Incidents shall be reviewed for public-safe mapping, cyber risk, physical security risk, public authority restrictions, community harm, national security sensitivity, and misuse risk.

302.9 Community-Protected or Protected Knowledge Incident.

302.9.1 A Community-Protected or Protected Knowledge Incident means any suspected or confirmed event involving unauthorized access, disclosure, mapping, publication, AI processing, translation, extraction, attribution, non-attribution breach, transfer, commercialization, or misuse of community-protected, Tribal / Indigenous, local, territorial, cultural, environmental, sacred, ecological, or otherwise protected knowledge.

302.9.2 Such incidents may require immediate hold, access restriction, community or Tribal review where applicable, publication withdrawal, AI index deletion where feasible, public-safe correction, controlled notice, grievance response, remedy, and non-retaliation protections.

302.10 AI Incident.

302.10.1 An AI Incident means any suspected or confirmed event involving AI system misuse, unauthorized AI use, prohibited data input, unauthorized training, model improvement without authority, embedding leakage, prompt leakage, hallucinated material output, biased output, unsafe output, agentic misaction, model drift, prompt injection, data poisoning, retrieval failure, or AI-related public-safe defect.

302.10.2 AI Incidents shall be reviewed under AI governance, data governance, privacy, cybersecurity, protected knowledge, public authority boundary, finance boundary, certification boundary, recognition boundary, procurement neutrality, and publication integrity controls.

302.11 Hallucination Incident.

302.11.1 A Hallucination Incident means an AI-generated or AI-assisted output that fabricates, invents, misstates, or materially distorts facts, sources, citations, legal claims, public authority actions, evidence records, method descriptions, technical standards, model outputs, confidence scores, or quotations in a material context.

302.11.2 Hallucination Incidents shall trigger affected-output review, correction, user notice where appropriate, reviewer training, model restriction, prompt revision, retrieval correction, or publication withdrawal where required.

302.12 Unsafe Output Incident.

302.12.1 An Unsafe Output Incident means a data, AI, cyber, dashboard, map, publication, technical baseline, public-good software, controlled-room, observability, or research output that may cause harm through exposure, overclaim, false reassurance, false alarm, protected knowledge disclosure, infrastructure disclosure, cyber disclosure, public authority confusion, finance overclaim, certification overclaim, recognition overclaim, procurement overclaim, public warning implication, or emergency command implication.

302.12.2 Unsafe Output Incidents may require publication hold, correction, withdrawal, retraction, takedown, controlled notice, public-safe clarification, or downstream notification.

302.13 Bias or Civil Rights Incident.

302.13.1 A Bias or Civil Rights Incident means any suspected or confirmed event in which data, AI, evidence classification, mapping, dashboarding, research design, publication, controlled-room access, public authority learning material, or technical baseline creates or may create discriminatory impact, inaccessible participation, unequal treatment, stigmatization, retaliation risk, or civil rights harm.

302.13.2 Such incidents shall be reviewed for affected persons, affected communities, protected classes, accessibility, language access, public authority misuse, safeguards, correction, remedy, and recurrence prevention.

302.14 Data Leakage Incident.

302.14.1 A Data Leakage Incident means unauthorized, unintended, excessive, or unsafe disclosure or exposure of data through prompts, AI outputs, embeddings, logs, repositories, public links, dashboards, maps, messages, email, vendor systems, screenshots, exports, downloads, backups, or inference.

302.14.2 Data Leakage Incidents shall be classified according to data class, exposure scope, recipients, access duration, sensitivity, retrieval risk, re-identification risk, downstream use, and containment feasibility.

302.15 Unauthorized Agentic Action Incident.

302.15.1 An Unauthorized Agentic Action Incident means any AI agent, automation, workflow, script, bot, tool, or system taking or attempting an action beyond approved permission, including sending communications, changing records, accessing restricted data, altering repositories, triggering releases, granting access, deleting data, rotating secrets, publishing materials, or initiating external transactions.

302.15.2 Unauthorized Agentic Action Incidents shall trigger tool-permission review, access revocation, log review, system isolation where required, correction, and recurrence-prevention controls.

302.16 Cybersecurity Incident.

302.16.1 A Cybersecurity Incident means any suspected or confirmed event affecting confidentiality, integrity, availability, resilience, authentication, authorization, systems, networks, endpoints, repositories, cloud environments, AI systems, controlled rooms, public-good software, financial systems, or data security.

302.16.2 Cybersecurity Incidents may include unauthorized access, malware, phishing, credential compromise, key compromise, token compromise, secret exposure, repository compromise, supply-chain compromise, vulnerability exploitation, denial of service, service disruption, data exfiltration, or system manipulation.

302.17 Unauthorized Access Incident.

302.17.1 An Unauthorized Access Incident means access to data, systems, rooms, repositories, accounts, credentials, logs, models, dashboards, maps, financial systems, public authority materials, protected knowledge, or records by a person, system, service account, AI agent, vendor, provider, or process without sufficient authorization.

302.17.2 Unauthorized Access Incidents may be accidental, negligent, malicious, excessive, expired-access, misconfigured-access, role-conflict, or privilege-abuse incidents.

302.18 Malware, Phishing, Credential, Key, Token, Secret, Repository, Supply Chain, Vulnerability, Denial of Service, or Service Disruption Incident.

302.18.1 The incident taxonomy shall include specific categories for malware, ransomware, phishing, social engineering, credential compromise, key compromise, token compromise, secret exposure, repository compromise, supply-chain compromise, dependency compromise, vulnerability exploitation, denial of service, service outage, cloud incident, AI-provider incident, vendor incident, and platform disruption.

302.18.2 Such incidents shall be escalated according to severity, affected systems, data classes, public authority implications, operational impact, public-safe publication impact, and legal obligations.

302.19 Public-Safe Publication Incident.

302.19.1 A Public-Safe Publication Incident means any publication, dashboard, map, report, technical baseline, software documentation, public authority learning material, donor report, sponsor report, public web content, social content, or Nexus-facing output that releases restricted, inaccurate, misleading, unsafe, confidential, privacy-sensitive, cyber-sensitive, infrastructure-sensitive, protected knowledge, or boundary-defective material.

302.19.2 Public-Safe Publication Incidents shall be reviewed for correction, withdrawal, retraction, takedown, public-safe notice, controlled notice, downstream dependency notification, and recurrence prevention.

302.20 Severity Levels.

302.20.1 Incidents shall be classified by severity according to harm, likelihood, data class, affected persons, affected communities, affected public authorities, affected systems, affected outputs, exposure scope, legal notification risk, public-safe risk, cyber risk, protected knowledge risk, boundary risk, and operational impact.

302.20.2 Severity levels may include informational, low, moderate, high, severe, critical, public-safe critical, protected-knowledge critical, public-authority critical, cyber critical, or other categories approved by policy.

302.20.3 Severity classification shall determine escalation, containment urgency, legal review, notification assessment, Board reporting, public-safe communication, and closeout requirements.

302.21 Incident Taxonomy Records.

302.21.1 The Corporation shall maintain Incident Taxonomy Records, including taxonomy definitions, data incident records, privacy incident records, personal information breach records, public authority data incident records, health-sensitive data incident records, cyber-sensitive data incident records, infrastructure-sensitive data incident records, community-protected and protected knowledge incident records, AI incident records, hallucination incident records, unsafe output incident records, bias and civil rights incident records, data leakage incident records, unauthorized agentic action incident records, cybersecurity incident records, unauthorized access incident records, malware / phishing / credential / key / token / secret / repository / supply-chain / vulnerability / denial-of-service / service-disruption incident records, public-safe publication incident records, severity classifications, corrections, restrictions, and archive records.

---

Section 303. Incident Intake, Triage, Containment, Investigation, Notification, Correction, and Closeout

303.1 Incident Intake.

303.1.1 The Corporation shall maintain incident intake pathways for directors, officers, employees, contractors, fellows, advisors, volunteers, committee members, reviewers, collaborators, vendors, providers, public authority participants, community participants, protected knowledge holders, funders, sponsors, users, and other persons to report suspected or confirmed incidents.

303.1.2 Incident intake shall permit prompt reporting of data incidents, privacy incidents, AI incidents, cybersecurity incidents, public authority data incidents, protected knowledge incidents, public-safe publication incidents, repository incidents, controlled-room incidents, and technical incidents.

303.1.3 Good-faith incident reporting shall be protected from retaliation.

303.2 Incident Case ID.

303.2.1 Each material incident shall be assigned an Incident Case ID or equivalent identifier.

303.2.2 The Incident Case ID shall be used to track intake, triage, severity, containment, evidence, affected materials, affected persons, legal review, notification assessment, correction, remediation, root cause, closeout, and recurrence prevention.

303.3 Triage.

303.3.1 Incidents shall be triaged promptly to determine incident type, data classes, affected systems, affected records, affected outputs, affected persons, affected communities, affected public authorities, exposure scope, containment needs, legal review, notification assessment, and public-safe communication needs.

303.3.2 Triage may be revised as new facts emerge.

303.4 Severity Classification.

303.4.1 Each material incident shall receive a severity classification according to the incident taxonomy, applicable policy, and the actual or potential impact of the incident.

303.4.2 Severity classification shall determine escalation to officers, counsel, Board, public authority contacts, vendors, insurers, funders, affected persons, communities, or other parties where required or appropriate.

303.5 Immediate Containment.

303.5.1 Immediate containment shall be taken where necessary to prevent further access, disclosure, harm, publication, exfiltration, model contamination, repository compromise, public authority confusion, protected knowledge exposure, or unsafe reliance.

303.5.2 Containment may occur before full investigation where risk requires prompt protective action.

303.6 Access Restriction.

303.6.1 Access restriction may include revoking user access, suspending accounts, limiting roles, disabling service accounts, restricting repository permissions, locking controlled rooms, restricting dashboards, disabling public links, limiting vendor access, and suspending AI-tool access.

303.6.2 Access restriction shall be proportionate and shall preserve evidence where needed for investigation.

303.7 Credential, Token, Key, or Secret Rotation.

303.7.1 Where credentials, tokens, keys, secrets, certificates, recovery codes, API keys, signing keys, or authentication materials are compromised or suspected to be compromised, the Corporation shall rotate, revoke, replace, or disable them promptly.

303.7.2 Rotation shall include review of dependent systems, logs, releases, repositories, vendors, integrations, backups, and downstream access.

303.8 System Isolation.

303.8.1 Systems, repositories, endpoints, cloud resources, AI tools, controlled rooms, datasets, dashboards, maps, or applications may be isolated where necessary to prevent further compromise, exposure, corruption, publication, or unauthorized processing.

303.8.2 Isolation shall be recorded and reviewed for business continuity, evidence preservation, public authority trust, and restoration planning.

303.9 Publication Hold.

303.9.1 A Publication Hold shall suspend release, circulation, indexing, posting, distribution, or external reliance on publications, public-safe reports, dashboards, maps, technical baselines, donor reports, sponsor reports, Academy materials, public authority learning materials, or Nexus-facing outputs affected by an incident.

303.9.2 Publication Holds shall identify affected materials, reason, review authority, release conditions, notice requirements, and correction pathway.

303.10 Repository Hold.

303.10.1 A Repository Hold shall suspend or restrict commits, releases, issues, downloads, public access, package publication, branch changes, secret exposure, or repository access where a repository incident, supply-chain risk, license risk, security risk, or data exposure risk exists.

303.10.2 Repository Holds shall preserve logs, commits, artifacts, release history, issue history, and evidence needed for investigation.

303.11 Data Hold.

303.11.1 A Data Hold shall suspend use, processing, sharing, publication, deletion, or transfer of affected data pending investigation or review.

303.11.2 Data Holds may apply to source data, derived data, synthetic data, embeddings, retrieval indexes, model outputs, dashboards, maps, evidence records, public authority data, health-sensitive data, cyber-sensitive data, infrastructure-sensitive data, protected knowledge, or controlled-room materials.

303.12 Controlled-Room Hold.

303.12.1 A Controlled-Room Hold shall suspend or restrict access to a controlled room, clean room, data room, evidence room, public authority room, no-download room, or related outputs where incident risk exists.

303.12.2 Controlled-Room Holds shall identify affected participants, exhibits, logs, outputs, restrictions, review authority, and closeout conditions.

303.13 Investigation.

303.13.1 Incident investigation shall determine facts, timeline, affected data, affected systems, affected persons, affected communities, affected public authorities, cause, scope, severity, containment sufficiency, legal obligations, correction needs, notification needs, and recurrence-prevention actions.

303.13.2 Investigation shall preserve relevant evidence, logs, access records, repository history, model records, inference records, vendor records, communications, controlled-room records, and publication records.

303.13.3 Investigations shall be conducted by competent persons with appropriate independence, confidentiality, privilege protection, and conflict management.

303.14 Legal and Privilege Review.

303.14.1 Incidents shall receive legal and privilege review where required or appropriate, including where personal information, public authority data, health-sensitive data, cyber-sensitive data, protected knowledge, controlled technology, contracts, grants, vendors, law enforcement, regulators, litigation, insurance, public statements, or notification obligations are implicated.

303.14.2 Legal review shall not be used to suppress necessary correction, conceal misconduct, avoid lawful notification, or prevent public-safe remediation.

303.15 Notification Assessment.

303.15.1 The Corporation shall assess whether an incident triggers notification to regulators, public authorities, data subjects, participants, communities, Tribal or Indigenous authorities where applicable, funders, grantors, sponsors, providers, hosts, partners, vendors, insurers, law enforcement, or the public.

303.15.2 Notification assessment shall consider applicable law, contracts, public authority terms, grants, research ethics, privacy requirements, cybersecurity obligations, community safeguards, protected knowledge protocols, public-safe communication, and institutional commitments.

303.16 Regulatory, Public Authority, Data Subject, Funder, Partner, Provider, Community, or Public Notice Where Required or Appropriate.

303.16.1 Notices shall be issued where required by law, contract, grant, public authority terms, ethics review, institutional policy, or competent determination.

303.16.2 Notices shall be accurate, timely, public-safe, non-misleading, and limited to information appropriate for the recipient and purpose.

303.16.3 Notices shall avoid exposing sensitive infrastructure, cyber details, protected knowledge, personal information, health-sensitive information, privileged information, or public authority restricted information beyond lawful and necessary scope.

303.17 Correction, Supersession, Withdrawal, Retraction, or Takedown.

303.17.1 Incidents affecting outputs, data, AI systems, dashboards, maps, repositories, public-good software, technical baselines, public authority learning materials, controlled-room outputs, donor reports, sponsor reports, or Nexus-facing materials may require correction, supersession, withdrawal, retraction, takedown, restriction, deletion, or archive update.

303.17.2 Corrective action shall include downstream dependency notification where affected materials have been reused, cited, routed, published, indexed, embedded, trained on, or externally shared.

303.18 Remediation.

303.18.1 Remediation shall address the immediate weakness, affected data or systems, affected outputs, user impacts, access controls, vendor controls, AI controls, repository controls, public-safe claims, protected knowledge safeguards, and policy gaps.

303.18.2 Remediation may include patching, configuration change, access reduction, model restriction, index deletion, data deletion, key rotation, vendor contract change, training, policy amendment, publication correction, controlled-room redesign, or replacement of systems.

303.19 Root Cause Review.

303.19.1 Material incidents shall receive root cause review proportionate to severity.

303.19.2 Root cause review shall identify technical, procedural, contractual, organizational, human, AI, vendor, access-control, documentation, training, design, or governance failures contributing to the incident.

303.19.3 Root cause review shall produce recurrence-prevention actions and accountable owners where appropriate.

303.20 Closeout.

303.20.1 Incident closeout shall occur only after containment, investigation, notification assessment, required notices, correction, remediation, root cause review, access restoration or revocation, records completion, and recurrence-prevention actions are completed or assigned.

303.20.2 Closeout records shall identify final severity, final facts, final scope, actions taken, notices issued, unresolved residual risk, owner of remaining actions, archive status, and Board reporting where required.

303.21 Incident Response Records.

303.21.1 The Corporation shall maintain Incident Response Records, including intake records, Incident Case IDs, triage records, severity classifications, containment records, access restrictions, credential / token / key / secret rotation records, system isolation records, Publication Holds, Repository Holds, Data Holds, Controlled-Room Holds, investigation records, legal and privilege review records, notification assessments, notices, corrections, supersessions, withdrawals, retractions, takedowns, remediation records, root cause reviews, closeout records, recurrence-prevention records, and archive records.

---

Section 304. Breach Notification, Public-Safe Communication, Coordinated Vulnerability Disclosure, and Law Enforcement Interface

304.1 Breach Notification Purpose.

304.1.1 Breach notification governance shall ensure that suspected or confirmed breaches, cybersecurity incidents, privacy incidents, data incidents, public authority data incidents, AI incidents, protected knowledge incidents, vulnerability incidents, and public-safe publication incidents are assessed for lawful, contractual, ethical, public authority, funder, community, and institutional notice obligations.

304.1.2 Breach notification shall protect affected persons, affected communities, public authority trust, privacy, cybersecurity, protected knowledge, institutional integrity, and public-safe communication while avoiding overdisclosure that increases harm.

304.2 Legal Notification Assessment.

304.2.1 The Corporation shall conduct legal notification assessment where an incident may trigger notification obligations under federal law, state law, territorial law, sectoral law, privacy law, cybersecurity law, health law, education law, public authority terms, contracts, grants, vendor agreements, research ethics, community protocols, or other applicable obligations.

304.2.2 Legal notification assessment shall consider incident facts, data classes, affected persons, affected jurisdictions, unauthorized acquisition or access, likelihood of harm, encryption or other protections, containment, misuse evidence, regulatory thresholds, timing requirements, notice content, and recipient categories.

304.3 Federal Notification Requirements Where Applicable.

304.3.1 Federal notification requirements shall be reviewed where incidents involve federally regulated data, federal public authority data, federal grants, federal contracts, health data, education data, financial data, communications data, cybersecurity obligations, controlled technology, sanctions, export-control, or other federal requirements.

304.3.2 Federal notices shall be issued only by competent authority and shall be accurate, timely, privilege-aware, public-safe, and consistent with legal obligations.

304.4 State and Territorial Notification Requirements Where Applicable.

304.4.1 State and territorial notification requirements shall be reviewed for incidents involving personal information, sensitive personal information, health data, biometric data, education data, public authority data, cybersecurity events, breach notification laws, consumer protection laws, and other applicable state or territorial obligations.

304.4.2 State-by-state and territory-by-territory differences in timing, content, attorney general notice, consumer notice, credit monitoring, regulator notice, and substitute notice shall be reviewed where applicable.

304.5 Contractual Notification Requirements.

304.5.1 Contractual notification requirements shall be reviewed for vendor agreements, processor agreements, cloud agreements, AI-provider agreements, public authority agreements, grants, sponsorships, donor agreements, research agreements, collaboration agreements, licenses, data-sharing agreements, controlled-room agreements, and repository agreements.

304.5.2 Contractual notice shall be coordinated with legal notification, public-safe communication, confidentiality obligations, privilege, and incident containment.

304.6 Public Authority Notification Requirements.

304.6.1 Public authority notification requirements shall be reviewed where incidents involve Public Authority Data, public authority systems, public authority participants, public authority learning materials, public authority rooms, procurement-related data, grant-related data, public finance data, public health data, emergency management data, public safety data, public works data, cyber-sensitive public-sector data, or infrastructure-sensitive data.

304.6.2 Public authority notices shall preserve capacity classification, official-status accuracy, public records implications, confidentiality, public-safe content, and non-execution boundaries.

304.7 Funder, Grantor, Sponsor, Provider, Host, Partner, or Community Notification Requirements.

304.7.1 The Corporation shall review whether funders, grantors, sponsors, providers, hosts, partners, universities, laboratories, communities, Tribal or Indigenous authorities where applicable, protected knowledge holders, collaborators, or other stakeholders must or should receive notice of an incident.

304.7.2 Such notices shall be limited to what is required or appropriate and shall not disclose personal information, protected knowledge, cyber-sensitive details, infrastructure-sensitive details, privileged material, public authority restricted information, or confidential information beyond lawful and necessary scope.

304.7.3 Sponsor, provider, donor, funder, or host notice shall not grant those actors control over investigation, correction, publication, public authority notice, regulatory notice, or public-safe communication.

304.8 Public-Safe Communication Discipline.

304.8.1 Incident communications, breach notices, public statements, controlled notices, website updates, public authority communications, donor communications, sponsor communications, provider communications, community communications, media responses, and internal updates shall be accurate, timely, proportionate, public-safe, non-misleading, and limitation-aware.

304.8.2 Public-safe communication shall avoid speculation, false reassurance, unsupported blame, public authority overclaim, finance overclaim, certification overclaim, recognition overclaim, procurement overclaim, public warning implication, emergency command implication, and disclosure of sensitive details that could increase harm.

304.8.3 Communications shall identify what is known, what is not yet known where appropriate, what actions are being taken, what affected persons or recipients should do where appropriate, and how correction or follow-up will occur.

304.9 No Overdisclosure of Sensitive Infrastructure, Cyber, Personal, Health, Public Authority, or Protected Knowledge Information.

304.9.1 The Corporation shall not overdisclose incident details where disclosure would expose vulnerabilities, credentials, keys, tokens, system architecture, critical infrastructure, sensitive locations, public authority restricted information, personal information, health-sensitive data, children’s or youth data, community-protected data, Tribal / Indigenous knowledge, protected knowledge, privileged information, or confidential materials beyond lawful and necessary scope.

304.9.2 Where disclosure is legally required but sensitive, the Corporation shall use lawful minimization, redaction, controlled notice, staggered disclosure, recipient-specific notice, coordinated disclosure, or counsel-guided communication.

304.10 Coordinated Vulnerability Disclosure.

304.10.1 The Corporation may use coordinated vulnerability disclosure where it identifies, receives, validates, or is affected by a vulnerability in Corporation systems, public-good software, third-party systems, public authority systems, repositories, AI systems, controlled-room systems, cloud systems, or infrastructure-sensitive systems.

304.10.2 Coordinated vulnerability disclosure shall be scoped to reduce harm, protect researchers acting in good faith, notify affected maintainers or owners where appropriate, avoid premature publication of exploit details, preserve evidence, coordinate remediation, and support public-safe disclosure.

304.10.3 Coordinated vulnerability disclosure shall not be represented as certification, cyber compliance approval, managed security service, public authority cyber command, emergency command, public warning, or legal guarantee.

304.11 Law Enforcement Interface Where Required or Appropriate.

304.11.1 The Corporation may interface with law enforcement where required or appropriate in relation to cybercrime, theft, extortion, fraud, unauthorized access, threats, violence, exploitation, unlawful disclosure, sanctions concerns, export-control concerns, public safety threats, or other matters requiring lawful law enforcement engagement.

304.11.2 Law enforcement interface shall be managed by competent authority, with legal review where appropriate, and shall preserve privilege, confidentiality, data minimization, public authority trust, protected knowledge, affected-person safety, and public-safe communication.

304.11.3 Law enforcement interface shall not convert the Corporation into a public authority, investigative agency, emergency command body, public warning body, regulator, or enforcement agency.

304.12 Regulator Interface Where Required or Appropriate.

304.12.1 The Corporation may interface with regulators, attorneys general, privacy authorities, cybersecurity authorities, charitable regulators, tax authorities, public health authorities, public records authorities, grant authorities, or other competent bodies where required or appropriate.

304.12.2 Regulator interface shall be accurate, timely, records-supported, privilege-aware, public-safe, and coordinated through competent authority.

304.12.3 Regulatory communication shall not be used to imply approval, endorsement, certification, public authority adoption, procurement approval, public finance approval, recognition, finance-readiness, or legal compliance approval beyond the record.

304.13 Privilege and Confidentiality Protection.

304.13.1 Incident investigation, notification assessment, legal review, regulator interface, law enforcement interface, public authority communication, vendor communication, and public-safe communication shall preserve attorney-client privilege, work product, confidentiality, research confidentiality, public authority restrictions, data subject privacy, protected knowledge, and security-sensitive information where applicable.

304.13.2 Privilege and confidentiality protections shall not be used to avoid required notification, suppress necessary correction, conceal misconduct, or continue unsafe practices.

304.14 Notification and Disclosure Records.

304.14.1 The Corporation shall maintain Notification and Disclosure Records, including breach notification purpose records, legal notification assessments, federal notification reviews, state and territorial notification reviews, contractual notification reviews, public authority notification reviews, funder / grantor / sponsor / provider / host / partner / community notification reviews, public-safe communication records, overdisclosure risk reviews, coordinated vulnerability disclosure records, law enforcement interface records, regulator interface records, privilege and confidentiality protection records, notices issued, notices withheld with rationale, corrections, public statements, controlled notices, closeout records, and archive records.

Section 305. Backup, Disaster Recovery, Business Continuity, RTO / RPO, and Resilience Testing

305.1 Backup Purpose.

305.1.1 The Corporation shall maintain backup controls sufficient to preserve the availability, integrity, recoverability, and continuity of its mission-critical records, systems, repositories, datasets, evidence records, research records, governance records, fiscal records, public authority data, controlled-room materials, AI governance records, model records, inference records, compute workload records, cybersecurity logs, public-good software, technical baselines, dashboards, maps, publications, correction records, and other institutional assets necessary to sustain the Corporation’s public-benefit purposes.

305.1.2 Backup controls shall support lawful continuity, evidence preservation, technical memory, correctionability, public authority trust, privacy, cybersecurity, protected knowledge safeguards, and resilience against accidental deletion, malicious deletion, ransomware, system failure, cloud-service failure, repository compromise, human error, vendor failure, natural disaster, cyber incident, infrastructure disruption, legal hold failure, and operational interruption.

305.1.3 Backup shall not be treated as mere file duplication. Backup shall be a governed institutional control subject to classification, encryption where appropriate, access restriction, retention alignment, restoration testing, incident response, legal hold, privacy obligations, protected knowledge restrictions, and secure disposal.

305.2 Disaster Recovery Purpose.

305.2.1 Disaster recovery shall enable the Corporation to restore critical systems, records, repositories, controlled rooms, public-good technical assets, publications, financial systems, communication channels, public authority interfaces, and technical operations after disruption.

305.2.2 Disaster recovery planning shall address cyberattack, ransomware, credential compromise, vendor outage, cloud outage, repository outage, website outage, data corruption, accidental deletion, natural disaster, infrastructure failure, telecommunications failure, public authority interface disruption, controlled-room disruption, key-person unavailability, and material continuity risk.

305.2.3 Disaster recovery shall preserve data integrity, confidentiality, public-safe publication status, legal holds, audit trails, access controls, encryption, key management, public authority restrictions, protected knowledge restrictions, and correction history during restoration.

305.3 Business Continuity Purpose.

305.3.1 Business continuity shall ensure that the Corporation can continue its essential governance, legal, financial, research, evidence, methods, data, AI, cybersecurity, publication, public authority learning, controlled-room, public-good technical asset, and incident-response functions during and after disruption.

305.3.2 Business continuity shall identify essential functions, continuity owners, minimum operating capabilities, alternate communication channels, delegated authorities, emergency access rules, public-safe communication rules, vendor dependencies, manual fallback procedures, and restoration priorities.

305.3.3 Continuity planning shall preserve non-execution boundaries. Continuity activity shall not convert the Corporation into an emergency command body, public warning authority, managed security provider, infrastructure operator, public authority, finance-readiness authority, certification body, procurement body, or enterprise execution actor.

305.4 Critical Systems Inventory.

305.4.1 The Corporation shall maintain a Critical Systems Inventory identifying systems whose disruption would materially impair governance, research, evidence integrity, data protection, cybersecurity, public authority trust, public-safe publication, technical memory, financial administration, controlled rooms, or public-good technical assets.

305.4.2 Critical systems may include identity systems, email and communications systems, cloud environments, repositories, evidence systems, document systems, financial systems, backup systems, AI systems, model registers, data registers, processing registers, controlled-room systems, dashboards, maps, public websites, incident-response systems, security systems, and publication systems.

305.4.3 The Critical Systems Inventory shall identify system owner, custodian, purpose, data classes, dependencies, vendor, access controls, backup status, recovery priority, Recovery Time Objective, Recovery Point Objective, incident escalation path, and retirement status where applicable.

305.5 Critical Data Inventory.

305.5.1 The Corporation shall maintain a Critical Data Inventory identifying data and records whose loss, corruption, unauthorized alteration, unauthorized disclosure, or unavailability would materially affect legal compliance, governance, research integrity, evidence integrity, public authority trust, technical memory, public-good technical assets, financial administration, privacy, cybersecurity, protected knowledge safeguards, or correctionability.

305.5.2 Critical data may include Board records, bylaws, policies, contracts, grants, fiscal records, tax records, data registers, processing registers, public authority data, evidence records, methods records, ontology records, model registers, inference records, compute workload records, technical baselines, software repositories, controlled-room exhibits, publication records, correction records, incident records, access logs, audit logs, and protected knowledge records.

305.5.3 Critical Data Inventory entries shall identify data owner, custodian, classification, source, location, backup frequency, retention status, encryption status, access restrictions, restoration priority, legal hold status, public-safe status, and correction path.

305.6 Backup Frequency.

305.6.1 Backup frequency shall be established according to data criticality, change rate, legal requirements, public authority restrictions, research integrity needs, publication reliance, system dependency, Recovery Point Objective, cybersecurity risk, and institutional tolerance for loss.

305.6.2 Critical records and systems shall be backed up on a schedule sufficient to meet the Corporation’s continuity requirements and applicable obligations.

305.6.3 Backup frequency may differ by data class, including more frequent backup for active repositories, financial systems, evidence systems, model registers, access logs, incident records, controlled-room exhibits, public authority materials, and actively maintained public-good technical assets.

305.7 Backup Security.

305.7.1 Backups shall be protected according to the classification and sensitivity of the underlying data.

305.7.2 Backup security controls may include encryption, access restriction, multi-factor authentication, separate administrative credentials, immutability where appropriate, offline or logically separated copies where appropriate, geographic redundancy where lawful, logging, monitoring, malware protection, integrity checking, and vendor-security review.

305.7.3 Backups containing public authority data, personal information, health-sensitive data, cyber-sensitive data, infrastructure-sensitive data, controlled technology, finance-sensitive data, confidential data, community-protected data, Tribal / Indigenous data, or protected knowledge shall not be restored, exported, transferred, AI-processed, or disclosed without applicable review and authority.

305.8 Backup Restoration Testing.

305.8.1 The Corporation shall test restoration of backups on a schedule proportionate to system criticality, data sensitivity, operational dependence, public authority relevance, cybersecurity risk, and continuity need.

305.8.2 Restoration testing shall verify that backups are complete, readable, secure, timely, uncorrupted, access-controlled, retention-aligned, and capable of restoring essential functions.

305.8.3 Failed restoration tests shall be logged, escalated, corrected, retested where appropriate, and reviewed for continuity impact.

305.9 Recovery Time Objectives.

305.9.1 Recovery Time Objectives shall identify the target time within which critical systems, records, repositories, controlled rooms, public authority interfaces, public-good technical assets, publications, communication channels, or financial systems should be restored after disruption.

305.9.2 Recovery Time Objectives shall be determined according to public-benefit priority, legal obligations, public authority trust, incident-response needs, publication exposure, financial controls, research continuity, evidence integrity, data sensitivity, and resource capacity.

305.9.3 Failure to meet a Recovery Time Objective shall be reviewed for corrective action, but Recovery Time Objectives shall not constitute external guarantees, warranties, service-level commitments, public authority commitments, emergency-response commitments, or operational assurances unless separately and lawfully agreed by competent authority.

305.10 Recovery Point Objectives.

305.10.1 Recovery Point Objectives shall identify the maximum tolerable amount of data loss, measured by time or record state, for critical systems and data assets after disruption.

305.10.2 Recovery Point Objectives shall inform backup frequency, replication strategy, repository mirroring, snapshot policy, log preservation, controlled-room continuity, evidence-system continuity, and public-good software continuity.

305.10.3 Recovery Point Objectives shall be reviewed after incidents, restoration failures, system migrations, vendor changes, public authority interface changes, and material changes in data sensitivity or operational dependence.

305.11 Continuity Tabletop Exercises.

305.11.1 The Corporation may conduct continuity tabletop exercises to test response to cyber incidents, ransomware, cloud outages, repository compromise, public website compromise, public authority data incidents, controlled-room disruptions, AI incidents, protected knowledge incidents, financial system disruption, key-person unavailability, and publication incidents.

305.11.2 Tabletop exercises shall test roles, escalation, communication, legal review, public-safe communication, access restriction, restoration, vendor coordination, public authority notification, data subject notification where applicable, correction, and Board reporting.

305.11.3 Tabletop findings shall be recorded and used to improve policies, systems, training, contracts, backups, disaster recovery, business continuity, incident response, and public-safe communication.

305.12 Repository Continuity.

305.12.1 Repository continuity shall preserve access to, integrity of, and recoverability of repositories containing public-good software, technical baselines, schemas, methods, documentation, datasets, model cards, system cards, benchmark cards, proof receipts, issues, releases, and archive records.

305.12.2 Repository continuity controls may include repository backups, mirrors, export rights, branch protections, access reviews, signed releases where appropriate, dependency records, SBOM records where appropriate, release notes, issue history, and archive copies.

305.12.3 Repository continuity shall protect against account loss, provider outage, malicious deletion, secret exposure, unauthorized public release, supply-chain compromise, dependency compromise, and loss of technical memory.

305.13 Register Continuity.

305.13.1 Register continuity shall preserve the availability and integrity of the Corporation’s Data Register, Processing Register, Model Register, Methods Register, Evidence Register where used, Controlled-Room Register where used, Incident Register where used, Vendor Register where used, and related governance registers.

305.13.2 Registers shall be backed up, access-controlled, versioned where appropriate, restorable, and protected against unauthorized alteration, deletion, corruption, or silent supersession.

305.13.3 Register continuity shall ensure that restoration does not erase correction history, legal hold status, access history, public authority restrictions, protected knowledge restrictions, or archive status.

305.14 Controlled-Room Continuity.

305.14.1 Controlled-room continuity shall ensure that controlled rooms, clean rooms, data rooms, evidence rooms, public authority rooms, no-download rooms, and restricted processing environments can be suspended, restored, closed, or migrated without compromising restricted materials, logs, access controls, exhibit integrity, public authority trust, protected knowledge, or correctionability.

305.14.2 Controlled-room continuity shall include access revocation procedures, exhibit backup, log preservation, output hold, participant communication, closeout procedures, and restoration conditions.

305.14.3 Controlled-room restoration shall not reopen access until room authority, participant eligibility, data classification, AI-use restrictions, public-safe restrictions, and incident status are reviewed.

305.15 Public Authority Interface Continuity.

305.15.1 Public authority interface continuity shall preserve lawful, accurate, capacity-classified, public-safe, and non-executing communication with public authorities during disruptions.

305.15.2 Continuity planning shall identify authorized public authority contacts, alternate communication methods, public authority data restrictions, notification obligations, public-safe language, and boundaries preventing public warning, emergency command, procurement approval, grant approval, public finance approval, or official adoption overclaim.

305.16 Public-Good Technical Asset Continuity.

305.16.1 Public-good technical asset continuity shall preserve access to and integrity of public-good software, open technical baselines, schemas, APIs, SDKs, documentation, model cards, system cards, benchmark cards, proof-receipt formats, observability tools, public-safe publication templates, and related public-good assets.

305.16.2 Continuity controls shall protect against private enclosure, loss of maintainers, repository compromise, license loss, missing provenance, dependency collapse, unpatched vulnerabilities, unsupported releases, and uncorrected public-safe claims.

305.16.3 Public-good technical asset continuity shall include custodian records, repository controls, release archives, license records, dependency records, security records, correction paths, and succession planning where appropriate.

305.17 Post-Incident Continuity Review.

305.17.1 After a material incident or disruption, the Corporation shall review whether backup, disaster recovery, business continuity, Recovery Time Objectives, Recovery Point Objectives, restoration, communication, vendor coordination, public authority interface, controlled-room continuity, and public-good technical asset continuity performed as intended.

305.17.2 Post-incident continuity review shall identify gaps, delays, failed assumptions, unavailable records, failed backups, excessive dependencies, vendor failures, access-control issues, public-safe communication issues, and training needs.

305.17.3 Corrective action shall be assigned, tracked, and closed through competent records.

305.18 Backup, Disaster Recovery, and Continuity Records.

305.18.1 The Corporation shall maintain Backup, Disaster Recovery, and Continuity Records, including backup purpose records, disaster recovery records, business continuity records, Critical Systems Inventory records, Critical Data Inventory records, backup frequency records, backup security records, restoration testing records, Recovery Time Objective records, Recovery Point Objective records, tabletop exercise records, repository continuity records, register continuity records, controlled-room continuity records, public authority interface continuity records, public-good technical asset continuity records, post-incident continuity reviews, corrective actions, restrictions, and archive records.

---

Section 306. Data, AI, Cybersecurity, and Privacy Training

306.1 Training Purpose.

306.1.1 The Corporation shall maintain training sufficient to ensure that persons performing, governing, supporting, reviewing, funding, contributing to, or accessing Corporation activities understand their responsibilities for data governance, privacy, AI governance, cybersecurity, controlled rooms, secure development, incident response, public-safe publication, protected knowledge, public authority data, and non-execution boundaries.

306.1.2 Training shall be proportionate to role, access, risk, data class, system access, public authority interface, controlled-room participation, AI use, repository role, publication role, and technical contribution.

306.1.3 Training shall support prevention, early detection, accurate reporting, public-safe communication, correctionability, and institutional culture of care rather than mere formal compliance.

306.2 Director Training.

306.2.1 Directors shall receive training appropriate to their oversight role concerning data governance, privacy, cybersecurity, AI governance, incident escalation, public authority data, protected knowledge, controlled rooms, legal holds, public-safe publication, vendor oversight, and Board reporting.

306.2.2 Director training shall emphasize fiduciary oversight, risk appetite, policy approval, incident oversight, public-benefit purpose, non-execution boundaries, conflicts, confidentiality, and avoidance of unsupported public statements.

306.3 Officer Training.

306.3.1 Officers shall receive training appropriate to operational responsibility for data governance, AI governance, cybersecurity, public authority interfaces, privacy compliance, controlled rooms, vendor management, incident response, repository continuity, publication review, and enforcement.

306.3.2 Officer training shall include escalation thresholds, authority limits, records obligations, legal review triggers, public-safe communication, and correction pathways.

306.4 Staff and Contractor Training.

306.4.1 Staff and contractors shall receive training before or promptly after being granted access to Corporation systems or data and periodically thereafter.

306.4.2 Training shall address acceptable use, data classification, access control, confidentiality, privacy, AI-use restrictions, cybersecurity hygiene, incident reporting, public authority data handling, protected knowledge handling, secure sharing, records retention, and publication limits.

306.4.3 Contractors shall receive role-specific training and shall not receive access to sensitive systems or data until confidentiality obligations, access approvals, and required training are complete unless a documented exception applies.

306.5 Fellow, Advisor, Volunteer, Contributor, Maintainer, and Repository Participant Training.

306.5.1 Fellows, advisors, volunteers, contributors, maintainers, repository participants, open-source contributors with elevated access, and technical collaborators shall receive training appropriate to their access and role.

306.5.2 Training may include repository security, contribution rules, license obligations, secret handling, public-good technical asset boundaries, AI-use restrictions, data classification, protected knowledge, public authority data limits, issue and pull request discipline, secure development, and public-safe documentation.

306.5.3 External contribution shall not excuse compliance with secure development, repository, data, AI, privacy, cybersecurity, or protected knowledge controls.

306.6 Committee and Council Training.

306.6.1 Committee members, council participants, advisory group participants, review bodies, and working group participants shall receive training appropriate to the data, records, public authority interfaces, conflicts, controlled rooms, and publication pathways relevant to their mandate.

306.6.2 Training shall include confidentiality, capacity classification, conflicts, public authority boundaries, finance boundaries, certification boundaries, recognition boundaries, procurement neutrality, provider neutrality, public-safe publication, and correctionability.

306.7 Public Authority Participant Training Where Appropriate.

306.7.1 Public authority participants may receive training or orientation where appropriate concerning capacity classification, public authority data handling, public records considerations, controlled-room rules, public-safe interpretation, non-execution boundaries, publication limits, and correction pathways.

306.7.2 Public authority participant training shall not imply official adoption, public authority delegation, public warning authority, emergency command, procurement approval, grant approval, regulatory action, public finance approval, or sovereign decision.

306.8 Provider, Sponsor, Host, and Partner Training Where Appropriate.

306.8.1 Providers, sponsors, hosts, partners, vendors, funders, donors, collaborators, and enterprise actors may receive training or orientation where their participation involves data access, technical contribution, controlled-room access, public-safe claims, public-good technical assets, public authority interfaces, or publication.

306.8.2 Training shall address sponsor non-control, provider neutrality, no endorsement, no certification, no recognition, no finance-readiness, no procurement approval, no public authority meaning, no public warning, no emergency command, protected knowledge restrictions, data restrictions, and misuse of Corporation name or materials.

306.9 Privacy Training.

306.9.1 Privacy training shall address personal information, sensitive personal information, rights-bearing data, health-sensitive data, children’s or youth data, education data, public authority data, notice, consent, minimization, purpose limitation, rights requests, complaint pathways, privacy incidents, breach reporting, and public-safe publication.

306.9.2 Privacy training shall emphasize that public availability, technical access, informal permission, or AI tool convenience does not eliminate privacy obligations.

306.10 Data Governance Training.

306.10.1 Data governance training shall address data inventory, data classification, lawful basis, authority records, permission records, consent, notice, access classes, data quality, provenance, source lineage, retention, deletion, restricted reuse, AI restrictions, public-safe status, and correction paths.

306.10.2 Training shall distinguish raw data, evidence data, public-safe data, restricted data, public authority data, community-protected data, protected knowledge, derived data, synthetic data, embeddings, model outputs, and publication outputs.

306.11 AI Governance Training.

306.11.1 AI governance training shall address approved AI tools, prohibited AI tools, approved uses, prohibited uses, data-input restrictions, no-training requirements, prompt and inference records, hallucination risk, bias risk, prompt injection, data leakage, human review, model registers, agentic AI boundaries, and AI incident reporting.

306.11.2 AI training shall make clear that AI outputs are not evidence, technical truth, public authority decisions, finance-readiness determinations, certifications, recognitions, procurement recommendations, legal advice, clinical advice, public warnings, or emergency commands unless separate lawful authority and review support the relevant status.

306.12 Cybersecurity Training.

306.12.1 Cybersecurity training shall address authentication, multi-factor authentication, password security, phishing, social engineering, device security, secure sharing, repository security, secrets management, malware, incident reporting, vendor risk, controlled-room security, public authority data protection, and sensitive data handling.

306.12.2 Persons with privileged access, repository maintainer access, controlled-room access, financial system access, AI tool administrative access, or public authority data access shall receive heightened cybersecurity training.

306.13 Controlled-Room Training.

306.13.1 Controlled-room training shall address room purpose, room charter, access conditions, no-download rules, exhibit handling, output review, logging, monitoring, prohibited copying, prohibited AI use where applicable, public authority capacity, finance boundaries, protected knowledge, and closeout.

306.13.2 Controlled-room participants shall acknowledge room rules before access where appropriate.

306.14 Secure Development Training.

306.14.1 Secure development training shall address secure coding, code review, dependency review, license review, vulnerability review, SBOM discipline where appropriate, artifact signing where appropriate, build integrity, release controls, secrets management, repository access, and rollback.

306.14.2 Maintainers and technical contributors shall understand that public-good software and technical baselines require public-safe documentation, provenance, license discipline, correction paths, and secure release controls.

306.15 Incident Response Training.

306.15.1 Incident response training shall address how to identify, report, escalate, contain, preserve evidence for, communicate about, and correct data, AI, cybersecurity, privacy, public authority, protected knowledge, repository, controlled-room, and publication incidents.

306.15.2 Training shall emphasize prompt reporting, non-retaliation for good-faith reports, avoidance of unauthorized deletion or self-remediation that destroys evidence, and use of approved communication channels.

306.16 Public-Safe Publication and Boundary Training.

306.16.1 Public-safe publication and boundary training shall address limitation language, evidence support, uncertainty disclosure, sponsor and provider disclosure, public authority references, finance-boundary discipline, certification-boundary discipline, recognition-boundary discipline, procurement neutrality, provider neutrality, public warning avoidance, emergency-command avoidance, and correction.

306.16.2 Training shall address misuse of terms including proof receipt, validated, verified, Nexus-compatible, Docket, Grid, recognized, finance-ready, insurance-ready, certified, public authority-approved, public-safe, and emergency-ready.

306.17 Training Renewal.

306.17.1 Training shall be renewed periodically and when roles, access, systems, data classes, AI tools, law, public authority interfaces, controlled rooms, incident patterns, or policies materially change.

306.17.2 Failure to complete required training may result in access restriction, role limitation, controlled-room exclusion, repository access suspension, AI tool restriction, publication approval restriction, or other corrective action.

306.18 Training Records.

306.18.1 The Corporation shall maintain Training Records, including training purpose records, director training records, officer training records, staff and contractor training records, fellow / advisor / volunteer / contributor / maintainer / repository participant training records, committee and council training records, public authority participant training records where appropriate, provider / sponsor / host / partner training records where appropriate, privacy training records, data governance training records, AI governance training records, cybersecurity training records, controlled-room training records, secure development training records, incident response training records, public-safe publication and boundary training records, renewal records, acknowledgments, exceptions, corrective actions, and archive records.

---

Section 307. Monitoring, Audit, Assurance, Control Testing, and Continuous Improvement

307.1 Monitoring Purpose.

307.1.1 Monitoring shall provide ongoing visibility into the effectiveness of controls governing data, AI, cybersecurity, privacy, controlled rooms, public authority data, protected knowledge, public-safe publication, vendors, repositories, technical baselines, public-good software, and incident response.

307.1.2 Monitoring shall support early detection of control failure, unauthorized access, unsafe AI use, data quality defects, privacy risk, cyber risk, public authority boundary risk, protected knowledge exposure, vendor risk, repository risk, and publication risk.

307.1.3 Monitoring shall be proportionate, lawful, privacy-aware, non-discriminatory, non-retaliatory, and limited to legitimate governance, security, compliance, integrity, and public-benefit purposes.

307.2 Access Monitoring.

307.2.1 Access monitoring shall review material access to systems, repositories, controlled rooms, datasets, AI tools, model registers, inference records, public authority data, protected knowledge, financial systems, dashboards, maps, and administrative platforms.

307.2.2 Access monitoring may identify unusual access, excessive access, expired access, privilege creep, unauthorized download, unapproved export, failed authentication, after-hours activity, service-account anomaly, agentic AI action, or access inconsistent with role.

307.2.3 Access monitoring findings shall be escalated, investigated, corrected, and recorded where material.

307.3 Model Monitoring.

307.3.1 Model monitoring shall assess approved AI systems and models for output quality, hallucination, bias, drift, unsafe outputs, prompt injection, data leakage, retrieval defects, unauthorized use, vendor changes, model version changes, incident history, and continued suitability.

307.3.2 Model monitoring shall be proportionate to risk class and shall include human review where necessary.

307.3.3 Model monitoring findings may result in limitation updates, user training, model restriction, prompt update, retrieval correction, vendor escalation, public-safe correction, suspension, retirement, or replacement.

307.4 Cybersecurity Monitoring.

307.4.1 Cybersecurity monitoring shall review systems, cloud environments, endpoints, repositories, identity systems, logs, network activity, vulnerability status, backup status, vendor alerts, public websites, controlled-room systems, AI systems, and financial systems for security events and control failure.

307.4.2 Cybersecurity monitoring shall support vulnerability management, incident detection, credential compromise detection, secret exposure detection, repository security, cloud security, and business continuity.

307.5 Data Quality Monitoring.

307.5.1 Data quality monitoring shall review data for accuracy, completeness, timeliness, relevance, integrity, provenance, source lineage, classification, permissions, staleness, duplication, missingness, transformation errors, labeling errors, and correction needs.

307.5.2 Monitoring shall be prioritized for high-reliance datasets, public authority data, evidence records, dashboards, maps, technical baselines, AI retrieval indexes, model evaluation datasets, public-safe publications, and Nexus-facing materials.

307.6 Privacy Compliance Monitoring.

307.6.1 Privacy compliance monitoring shall review notice, consent, lawful basis, data subject rights, minimization, purpose limitation, retention, deletion, access controls, privacy impact assessments, data protection impact assessments, breach response, AI-use restrictions, and cross-border transfer controls.

307.6.2 Privacy compliance monitoring shall identify privacy drift, unauthorized reuse, excessive retention, unapproved AI processing, rights-request failure, and publication risk.

307.7 Public Authority Data Monitoring.

307.7.1 Public Authority Data monitoring shall review public authority capacity records, permitted use, prohibited use, public records implications, confidentiality, AI-use restrictions, publication restrictions, transfer restrictions, retention, deletion, and public-safe language.

307.7.2 Monitoring shall ensure that Public Authority Data is not converted into public warning, emergency command, procurement approval, regulatory approval, grant approval, public finance approval, official adoption, recognition, finance-readiness, certification, or enterprise execution meaning.

307.8 Protected Knowledge Monitoring.

307.8.1 Protected Knowledge monitoring shall review access, use, AI processing, mapping, publication, transfer, attribution, non-attribution, withdrawal, restriction, correction, grievance, and safeguard controls for community-protected, Tribal / Indigenous, local, territorial, cultural, environmental, sacred, ecological, and protected knowledge.

307.8.2 Monitoring shall identify extraction risk, unauthorized generalization, semantic drift, public-safe mapping risk, community exposure, AI embedding risk, and sponsor or provider misuse.

307.9 Vendor Monitoring.

307.9.1 Vendor monitoring shall review material third-party technology providers, data processors, AI providers, cloud providers, cybersecurity providers, repository providers, hosting providers, and software vendors for continued compliance with contractual, security, privacy, AI-use, subprocessor, incident notification, deletion, portability, and public-safe requirements.

307.9.2 Vendor monitoring may include review of updated terms, AI training defaults, subprocessor notices, security reports, incidents, service availability, audit materials, access records, and exit risk.

307.10 Control Testing.

307.10.1 The Corporation shall test material controls periodically or event-triggered according to risk, including access controls, backup restoration, incident response, AI-use restrictions, no-training controls, controlled-room rules, repository protections, secure release controls, privacy rights workflows, deletion workflows, vendor controls, and public-safe publication controls.

307.10.2 Control testing shall identify design effectiveness, operating effectiveness, gaps, exceptions, failures, and improvement actions.

307.10.3 Control testing may be internal, independent, automated, manual, tabletop-based, sample-based, evidence-based, or external where appropriate.

307.11 Internal Audit or Assurance Review.

307.11.1 The Corporation may conduct internal audit, internal assurance, management review, committee review, or independent internal review of Article X controls.

307.11.2 Internal assurance may review data governance, privacy, AI governance, cybersecurity, controlled rooms, vendor security, incident response, backup, disaster recovery, secure development, public authority data, protected knowledge, and public-safe publication.

307.11.3 Internal audit or assurance shall preserve independence appropriate to the review scope and shall report material findings to competent authority.

307.12 External Review Where Appropriate.

307.12.1 The Corporation may use external review where appropriate for cybersecurity assessment, privacy review, AI governance review, penetration testing, secure development review, repository security, cloud security, controlled-room controls, public authority data handling, protected knowledge safeguards, or incident response.

307.12.2 External reviewers shall be subject to confidentiality, conflict review, data access restrictions, public authority restrictions, protected knowledge restrictions, and contract terms appropriate to the review.

307.12.3 External review shall not be represented as certification, accreditation, legal compliance approval, public authority approval, procurement approval, finance-readiness, recognition, rating, guarantee, or immunity from correction unless a competent independent authority lawfully issues such status and the Corporation accurately describes it.

307.13 Findings.

307.13.1 Monitoring, audit, assurance, control testing, and external review findings shall be recorded, classified, assigned severity, assigned owner, assigned due date where appropriate, and tracked to resolution.

307.13.2 Findings may include control deficiency, policy gap, training gap, access issue, vendor risk, AI risk, privacy risk, cyber risk, publication risk, protected knowledge risk, public authority boundary risk, repository risk, retention issue, continuity issue, or correction need.

307.14 Corrective Action.

307.14.1 Corrective action shall address findings, incidents, control failures, audit results, monitoring anomalies, vendor deficiencies, privacy gaps, AI defects, cybersecurity weaknesses, protected knowledge exposure, public authority data issues, repository risks, and public-safe publication defects.

307.14.2 Corrective action may include policy change, access restriction, system configuration change, data correction, model restriction, vendor restriction, training, incident response, repository hold, publication correction, controlled-room redesign, technical update, contract amendment, or termination.

307.15 Control Updates.

307.15.1 Controls shall be updated when monitoring, audit, assurance, incidents, legal changes, vendor changes, technology changes, AI model changes, public authority requirements, protected knowledge obligations, or operational changes reveal that existing controls are insufficient.

307.15.2 Control updates shall be documented, communicated to affected persons, implemented, and tested where appropriate.

307.16 Policy Updates.

307.16.1 Policies under Article X shall be reviewed and updated periodically and when material changes occur in law, technology, operations, public authority interfaces, AI systems, cybersecurity threats, data classes, vendor terms, controlled-room activities, incident patterns, or public-safe publication risk.

307.16.2 Policy updates shall preserve consistency with the Corporation’s public-benefit purpose, non-execution boundaries, role separation, data governance, privacy, cybersecurity, AI governance, protected knowledge safeguards, public authority boundary discipline, and correctionability.

307.17 Technical Updates.

307.17.1 Technical updates shall be implemented where necessary to address vulnerabilities, control gaps, system drift, model drift, repository risk, backup weakness, cloud misconfiguration, access-control weakness, logging gaps, vendor changes, or public-safe publication risks.

307.17.2 Technical updates shall be tested, documented, released through appropriate controls, and monitored for unintended effects.

307.18 Continuous Improvement Records.

307.18.1 The Corporation shall maintain Continuous Improvement Records, including monitoring purpose records, access monitoring records, model monitoring records, cybersecurity monitoring records, data quality monitoring records, privacy compliance monitoring records, public authority data monitoring records, protected knowledge monitoring records, vendor monitoring records, control testing records, internal audit or assurance review records, external review records where appropriate, findings, corrective actions, control updates, policy updates, technical updates, closure records, and archive records.

---

Section 308. Enforcement of Article X

308.1 Article X Enforcement Purpose.

308.1.1 Enforcement of Article X shall protect data governance, privacy, AI governance, cybersecurity, controlled rooms, public authority data, health-sensitive data, cyber-sensitive data, infrastructure-sensitive data, community-protected data, Tribal / Indigenous data, protected knowledge, cross-border transfer discipline, secure development, repository security, incident response, public-safe publication, evidence integrity, research integrity, technical memory, and public-benefit trust.

308.1.2 Enforcement shall be proportionate, records-based, non-retaliatory, conflict-reviewed, privacy-aware, public-safe, and capable of immediate protective action where necessary.

308.1.3 Enforcement shall preserve the Corporation’s non-executing role and shall not convert enforcement activity into public authority action, public warning, emergency command, certification, recognition, finance-readiness, procurement approval, provider preference, managed security service, or enterprise execution.

308.2 Data Governance Violation Review.

308.2.1 Data Governance Violation Review shall be initiated where data may have been collected, received, processed, classified, stored, accessed, shared, published, retained, deleted, archived, AI-processed, or reused without lawful basis, authority, permission, proper classification, purpose limitation, access control, minimization, quality review, retention compliance, or correction path.

308.2.2 Review may require data hold, access restriction, lawful basis review, permission review, classification review, deletion where appropriate, correction of affected outputs, downstream notification, training, policy amendment, or disciplinary action.

308.3 Privacy Violation Review.

308.3.1 Privacy Violation Review shall be initiated where personal information, sensitive personal information, rights-bearing data, children’s or youth data, education data, health-sensitive data, public authority data, or privacy-relevant data may have been processed, disclosed, retained, linked, inferred, published, AI-processed, or transferred unlawfully or inconsistently with notice, consent, rights, minimization, or purpose limitation.

308.3.2 Review may require privacy impact review, rights-request review, data deletion, access restriction, breach notification assessment, affected-person notice where required, public-safe correction, vendor action, training, or policy change.

308.4 AI Governance Violation Review.

308.4.1 AI Governance Violation Review shall be initiated where AI systems may have been used without approval, used with prohibited data, used for unauthorized training, used without human review, used to generate material hallucination, used to issue boundary-defective outputs, used to process protected knowledge improperly, used to communicate autonomously without authorization, or used outside permitted purpose.

308.4.2 Review may require AI hold, model restriction, prompt review where safe, inference record review, deletion or unlearning request where available, vendor notice, output correction, publication withdrawal, access restriction, retraining of users, or model retirement.

308.5 Cybersecurity Violation Review.

308.5.1 Cybersecurity Violation Review shall be initiated where systems, credentials, keys, tokens, secrets, repositories, cloud environments, endpoints, networks, AI systems, controlled rooms, logs, or data may have been accessed, configured, shared, exposed, altered, or used insecurely.

308.5.2 Review may require credential rotation, key revocation, token revocation, system isolation, repository hold, vulnerability remediation, access revocation, incident response, vendor escalation, user training, or legal review.

308.6 Controlled-Room Violation Review.

308.6.1 Controlled-Room Violation Review shall be initiated where a controlled room, clean room, data room, evidence room, public authority room, no-download room, or restricted processing environment may have been accessed, used, copied from, exported from, AI-processed, disclosed, monitored, or closed inconsistently with its Room Charter or applicable restrictions.

308.6.2 Review may require room hold, access revocation, exhibit review, output review, log review, participant restriction, public authority notice where required, protected knowledge review, correction, withdrawal, or closeout revision.

308.7 Public Authority Data Violation Review.

308.7.1 Public Authority Data Violation Review shall be initiated where Public Authority Data may have been received, used, published, transferred, AI-processed, mapped, retained, deleted, or represented without lawful authority, capacity record, permitted use, public-safe review, or public authority restrictions.

308.7.2 Review shall assess whether the violation created or could create official adoption confusion, public warning implication, emergency command implication, procurement implication, grant implication, regulatory implication, public finance implication, public records issue, confidentiality breach, or public authority trust harm.

308.8 Health-Sensitive Data Violation Review.

308.8.1 Health-Sensitive Data Violation Review shall be initiated where health-sensitive, public health, clinical, medical, youth, vulnerable-population, biosecurity, health-system, or health-adjacent data may have been accessed, processed, disclosed, AI-processed, published, mapped, retained, transferred, or deleted without required lawful basis, consent, ethics review, privacy protection, public authority permission, or safeguards.

308.8.2 Review may require health privacy review, ethics review, de-identification review, public-safe publication review, affected-person notice where required, public authority notice where required, data restriction, deletion where appropriate, correction, or incident response.

308.9 Cyber-Sensitive or Infrastructure-Sensitive Data Violation Review.

308.9.1 Cyber-Sensitive or Infrastructure-Sensitive Data Violation Review shall be initiated where vulnerability data, credentials, keys, tokens, secrets, logs, architecture, critical infrastructure data, geospatial layers, operational dependencies, telecom data, AI-RAN / O-RAN data, utility data, port data, transportation data, digital twin outputs, or emergency-system data may have been exposed, transferred, published, mapped, AI-processed, or mishandled.

308.9.2 Review may require immediate containment, access restriction, public-safe redaction, coordinated vulnerability disclosure, public authority notice where required, key rotation, system isolation, publication withdrawal, legal review, or law enforcement interface where appropriate.

308.10 Community-Protected, Tribal / Indigenous, or Protected Knowledge Violation Review.

308.10.1 Community-Protected, Tribal / Indigenous, or Protected Knowledge Violation Review shall be initiated where community-protected data, Indigenous data, Indigenous knowledge, local knowledge, territorial knowledge, cultural knowledge, environmental knowledge, sacred knowledge, ecological knowledge, vulnerability-related knowledge, or protected knowledge may have been collected, accessed, disclosed, mapped, translated, generalized, embedded, trained on, published, transferred, commercialized, or used without lawful authority, protocol permission, consent where required, safeguards review, attribution control, or public-safe approval.

308.10.2 Review may require immediate hold, access restriction, AI index deletion where feasible, publication withdrawal, mapping correction, non-attribution, community review, Tribal or Indigenous review where applicable, grievance response, remedy, controlled notice, public-safe correction, deletion where lawful, or recurrence-prevention action.

308.11 Cross-Border Transfer Violation Review.

308.11.1 Cross-Border Transfer Violation Review shall be initiated where data may have been transferred, remotely accessed, stored, AI-processed, embedded, published, or made available across jurisdictions without required lawful basis, contractual safeguards, public authority permission, localization review, privacy review, export-control review, sanctions review, protected knowledge review, or public-safe approval.

308.11.2 Review may require transfer hold, access revocation, data return, deletion, localization, vendor restriction, cross-border notice where required, sanctions or export-control review, legal review, correction, or termination of the transfer arrangement.

308.12 Secure Development or Repository Violation Review.

308.12.1 Secure Development or Repository Violation Review shall be initiated where code, repositories, technical baselines, public-good software, datasets, models, documentation, releases, secrets, keys, tokens, dependencies, licenses, SBOM records, signed artifacts, build pipelines, or release controls may have been mishandled.

308.12.2 Review may require repository hold, access restriction, secret rotation, dependency removal, vulnerability remediation, license correction, release withdrawal, artifact revocation, signing-key rotation, SBOM correction, rollback, patch release, or public-safe notice.

308.13 Incident Response Violation Review.

308.13.1 Incident Response Violation Review shall be initiated where an incident may have been unreported, reported late, misclassified, mishandled, concealed, inadequately contained, inadequately investigated, inadequately notified, inadequately corrected, improperly communicated, or closed without adequate remediation.

308.13.2 Review may require reopened investigation, severity reclassification, notification reassessment, additional containment, correction, public-safe communication, root cause review, training, policy amendment, or disciplinary action.

308.14 Corrective Action.

308.14.1 Corrective action under Article X may include correction, supersession, withdrawal, retraction, takedown, limitation language, access restriction, data hold, AI hold, repository hold, publication hold, controlled-room hold, system quarantine, technical isolation, credential rotation, key revocation, token revocation, model restriction, model retirement, vendor restriction, vendor termination, public-safe notice, controlled notice, legal review, regulator notice, public authority notice, funder notice, community notice where appropriate, training, policy amendment, technical update, contract remedy, or disciplinary action.

308.14.2 Corrective action shall be proportionate to severity, intent, recurrence, affected persons, affected communities, affected public authorities, affected systems, affected data classes, affected outputs, legal obligations, public-safe risk, cyber risk, privacy risk, protected knowledge risk, boundary risk, and recurrence-prevention needs.

308.15 Access Restriction.

308.15.1 The Corporation may restrict access to data, systems, AI tools, models, repositories, controlled rooms, dashboards, maps, publications, public authority materials, protected knowledge, financial systems, administrative systems, vendor systems, or technical assets where necessary to protect integrity, privacy, cybersecurity, confidentiality, public authority trust, protected knowledge, public-safe publication, or correctionability.

308.15.2 Access restriction may be temporary or continuing and may include account suspension, role reduction, repository permission removal, controlled-room exclusion, AI-tool restriction, service-account disabling, credential revocation, device restriction, vendor-access suspension, or privileged-access removal.

308.15.3 Access restriction shall be recorded and reviewed for restoration conditions where appropriate.

308.16 Data Hold, AI Hold, Repository Hold, Publication Hold, Controlled-Room Hold, System Quarantine, Technical Isolation, Credential Rotation, Model Restriction, Model Retirement, Vendor Restriction, Suspension, Termination, Public or Controlled Correction, Legal Referral, Regulator Referral, Public Authority Referral, Funder Referral, or Law Enforcement Referral Where Required or Appropriate.

308.16.1 Where required or appropriate, the Corporation may impose or pursue one or more protective, corrective, remedial, or referral actions, including:

308.16.1(a) Data Hold, AI Hold, Repository Hold, Publication Hold, Controlled-Room Hold, dashboard hold, map hold, release hold, or evidence hold; 308.16.1(b) system quarantine, technical isolation, network restriction, cloud-resource isolation, endpoint isolation, controlled-room suspension, or public-link disabling; 308.16.1(c) credential rotation, key revocation, token revocation, certificate revocation, secret replacement, recovery-code replacement, signing-key rotation, or privileged-access reset; 308.16.1(d) model restriction, model-use hold, model configuration change, retrieval-index restriction, embedding deletion where feasible, model retirement, model replacement, or vendor AI restriction; 308.16.1(e) vendor restriction, vendor notice, vendor remediation demand, vendor contract amendment, vendor access suspension, vendor termination, migration, or exit; 308.16.1(f) suspension, role restriction, controlled-room exclusion, repository access removal, publication authority removal, committee recusal, contract termination, or relationship termination; 308.16.1(g) public correction, controlled correction, public-safe clarification, takedown, withdrawal, retraction, supersession, limitation notice, downstream dependency notice, or archive update; and 308.16.1(h) legal referral, regulator referral, public authority referral, funder referral, grantor referral, insurer notice, journal or repository notice, law enforcement referral, or other lawful escalation.

308.16.2 Referral or notification shall preserve privilege, confidentiality, public-safe communication, protected knowledge, public authority trust, data minimization, and non-retaliation.

308.16.3 Remedial action shall not be delayed for reputational reasons where delay would materially increase harm, legal risk, public authority confusion, privacy exposure, cyber exposure, protected knowledge exposure, or misleading reliance.

308.17 Enforcement Records.

308.17.1 The Corporation shall maintain Article X Enforcement Records, including enforcement purpose records, violation reviews, intake records, case identifiers, affected data records, affected systems, affected outputs, affected persons, affected communities, affected public authorities, severity classifications, interim measures, access restrictions, holds, quarantines, technical isolations, credential rotations, model restrictions, model retirements, vendor restrictions, suspensions, terminations, public corrections, controlled corrections, referrals, notices, investigation records, findings, corrective actions, remediation records, root cause records, recurrence-prevention records, closeout records, and archive records.