AI governance frameworks, and what answers them.
The horizontal law, the management standards, the voluntary codes and the control sets the discipline is built against, then the reverse index that names, for each obligation, the engineering artefact that evidences it and the layer it lives in.
Frameworks
78 instruments: laws, standards, codes and control sets. External links open the canonical source.
Compared side by side, clause by clause: ISO 42001 vs EU AI Act · NIST AI RMF vs ISO 42001 · NIST AI RMF vs EU AI Act.
Text description
In rough order of force, weakest first, as of 2026-09-24. Principles and soft law (the OECD AI Principles, the UNESCO Recommendation, the G7 Hiroshima Code of Conduct and the EU High-Level Expert Group guidelines) set the target and the shared vocabulary. Standards and frameworks (the NIST AI RMF, the ISO/IEC 42001 family and the IEEE 7000 series) are voluntary; ISO/IEC 42001 is certifiable, evidences a management system and confers no AI Act presumption of conformity. The NIST AI RMF adapts the OECD lifecycle and dimensions. Harmonised standards, written by CEN-CENELEC JTC 21 on a Commission request, give a presumption of conformity under Article 40 only once their reference is cited in the Official Journal; the chapter found none cited. The Council of Europe Framework Convention (CETS No. 225) binds the Parties that ratify it and is not in force; the EU AI Act is binding law and recalls the HLEG principles in recital 27. The OECD definition of an AI system, the Convention's Article 2 and the AI Act's Article 3(1) use near-identical wording. The engineering rule: build each control once, tag it with every instrument it serves and generate each instrument's view from the tags.
Data table
| Type | Instrument | Obligations |
|---|---|---|
| Laws | EU AI Act (post-Omnibus) | 50 |
| Laws | General Data Protection Regulation (EU) 2016/679 | 15 |
| Laws | NIS2 Directive (EU) 2022/2555 | 2 |
| Laws | Digital Operational Resilience Act (EU) 2022/2554 | 2 |
| Laws | Cyber Resilience Act (EU) 2024/2847 | 1 |
| Laws | Product Liability Directive (EU) 2024/2853 | 3 |
| Laws | Directive (EU) 2019/790 on copyright in the Digital Single Market | 1 |
| Laws | Digital Services Act (EU) 2022/2065 | 2 |
| Laws | Unfair Commercial Practices Directive 2005/29/EC | 1 |
| Laws | Platform Work Directive (EU) 2024/2831 | 1 |
| Laws | Consumer Credit Directive (EU) 2023/2225 | 1 |
| Laws | California SB 53 (TFAIA) | 2 |
| Laws | New York RAISE Act | 1 |
| Laws | California AB 2013 (training-data transparency) | 1 |
| Laws | California AI Transparency Act (SB 942 as amended by AB 853) | 1 |
| Laws | California SB 243 (companion chatbots) | 1 |
| Laws | California CPPA regulations (ADMT, risk assessments, cybersecurity audits) | 2 |
| Laws | New York General Business Law Article 47 (AI companion models) | 1 |
| Laws | Illinois HB 3773 (AI in employment) | 1 |
| Laws | New York City Local Law 144 (automated employment decision tools) | 1 |
| Laws | Utah AI disclosure law (AI Policy Act as amended by SB 226) | 1 |
| Laws | Colorado SB21-169 (insurers' use of external consumer data) | 1 |
| Laws | Virginia Consumer Data Protection Act | 1 |
| Laws | Colorado Privacy Act (SB21-190) | 1 |
| Laws | Minnesota Consumer Data Privacy Act | 1 |
| Laws | Illinois Biometric Information Privacy Act (BIPA) | 1 |
| Laws | Washington My Health My Data Act | 1 |
| Laws | Equal Credit Opportunity Act, Regulation B (12 CFR 1002) | 1 |
| Laws | Fair Credit Reporting Act (15 U.S.C. 1681m) | 1 |
| Laws | Title VII and the Uniform Guidelines on Employee Selection Procedures | 1 |
| Laws | FTC Act s. 5 (15 U.S.C. 45) | 1 |
| Laws | TAKE IT DOWN Act (Public Law 119-12) | 1 |
| Laws | Texas TRAIGA (HB 149) | 1 |
| Laws | Colorado ADMT law (SB 26-189, replacing SB 24-205) | 1 |
| Laws | South Korea AI Basic Act | 8 |
| Laws | UK Data (Use and Access) Act 2025 | 1 |
| Laws | UK Digital Markets, Competition and Consumers Act 2024 | 1 |
| Laws | China Provisions on Algorithmic Recommendation (2022) | 1 |
| Laws | China Provisions on Deep Synthesis (2023) | 1 |
| Laws | China Interim Measures for Generative AI Services (2023) | 1 |
| Laws | China Measures for Labelling AI-Generated Synthetic Content (2025) | 1 |
| Laws | China Personal Information Protection Law (PIPL) | 1 |
| Laws | China Interim Measures for Anthropomorphic Interaction Services (2026) | 1 |
| Laws | Canada Directive on Automated Decision-Making | 1 |
| Laws | Brazil General Data Protection Law (LGPD, Lei 13.709/2018) | 1 |
| Laws | Council of Europe Framework Convention on AI (CETS No. 225) | 4 |
| Frameworks | NIST AI RMF | 4 |
| Frameworks | NIST AI Agent Standards Initiative | 1 |
| Frameworks | NIST IR 8596 Cyber AI Profile (draft) | 1 |
| Frameworks | NIST AI 800-1 (draft) | 1 |
| Frameworks | NIST AI 600-1 Generative AI Profile | 1 |
| Frameworks | CSA STAR for AI | 1 |
| Frameworks | OWASP Top 10 for Agentic Applications 2026 | 1 |
| Frameworks | OWASP Top 10 for LLM Applications 2026 | 1 |
| Frameworks | OMB Memorandum M-25-21 (federal use of AI) | 1 |
| Frameworks | OMB Memorandum M-26-04 (unbiased AI principles in procurement) | 1 |
| Frameworks | GAO AI Accountability Framework (GAO-21-519SP) | 4 |
| Frameworks | Singapore Model AI Governance Framework for Generative AI | 1 |
| Frameworks | Singapore Model AI Governance Framework for Agentic AI | 2 |
| Frameworks | TC260 AI Safety Governance Framework 3.0 | 2 |
| Frameworks | OECD AI Principles (Recommendation on AI, OECD/LEGAL/0449) | 2 |
| Standards | ISO/IEC 42001 | 9 |
| Standards | ISO/IEC 42005 | 1 |
| Standards | ISO/IEC 42006:2025 | 1 |
| Standards | ISO/IEC 23894:2023 | 1 |
| Standards | ISO/IEC 22989:2022 | 1 |
| Standards | OWASP Agent Control Standard (ACS) | 1 |
| Standards | OWASP AIBOM | 1 |
| Standards | GB/T 45654-2025 Basic security requirements for generative AI services | 1 |
| Standards | ETSI EN 304 223 | 1 |
| Standards | EN 18286:2026 | 1 |
| Standards | prEN 18228 AI risk management (draft) | 1 |
| Standards | prEN 18229-1 AI trustworthiness framework, Part 1: logging (draft) | 1 |
| Codes | GPAI Code of Practice | 5 |
| Codes | G7 Hiroshima Process International Code of Conduct for Advanced AI Systems | 4 |
| Control sets | CSA AI Controls Matrix (AICM) v1.1 | 3 |
No obligation rows yet, so not drawn: UK ATRS (standard), AI Verify (framework).
| Framework | Type | Issuer | Summary |
|---|---|---|---|
| EU AI Act (post-Omnibus) | law | European Union | The EU's horizontal, risk-tiered law for AI (Regulation (EU) 2024/1689), amended by the Digital Omnibus (Regulation (EU) 2026/1744). High-risk Annex III obligations apply from 2 December 2027; Annex I embedded systems from 2 August 2028. |
| GPAI Code of Practice | code | European Commission | The voluntary instrument (published 10 July 2025) providers use to demonstrate compliance with the GPAI obligations until harmonised standards exist. Three chapters: Safety and Security, Transparency and Copyright. |
| General Data Protection Regulation (EU) 2016/679 | law | European Union | The EU data-protection regulation, applying since 25 May 2018. It applies alongside the AI Act whenever an AI system processes personal data: lawful basis, minimisation, DPIA, automated decisions, data subject rights and breach notification. |
| NIS2 Directive (EU) 2022/2555 | law | European Union | The EU cyber-security directive for essential and important entities: risk-management measures (incl. business continuity and supply-chain security) and significant-incident reporting on a 24-hour, 72-hour and one-month clock. Member States apply their measures from 18 October 2024. |
| Digital Operational Resilience Act (EU) 2022/2554 | law | European Union | The EU regulation on the digital operational resilience of financial entities, applying since 17 January 2025: ICT risk management, major ICT-related incident reporting and the management of ICT third-party risk, including AI and model services. |
| Cyber Resilience Act (EU) 2024/2847 | law | European Union | The EU regulation on cyber-security requirements for products with digital elements. Its reporting of actively exploited vulnerabilities and severe incidents (Article 14) applies from 11 September 2026; the rest from 11 December 2027. |
| Product Liability Directive (EU) 2024/2853 | law | European Union | The revised EU strict-liability regime for defective products. Software, including AI systems, is a product; it applies to products placed on the market or put into service after 9 December 2026, the transposition deadline. |
| Directive (EU) 2019/790 on copyright in the Digital Single Market | law | European Union | The EU copyright directive whose Article 4 text-and-data-mining exception applies only where rightholders have not reserved their works, for online content by machine-readable means. The AI Act points GPAI providers to it (Art. 53(1)(c)). |
| Digital Services Act (EU) 2022/2065 | law | European Union | The EU regulation on intermediary services, applying in full since 17 February 2024. For online platforms it bans deceptive or manipulative interface design and requires the main parameters of recommender systems to be explained. |
| Unfair Commercial Practices Directive 2005/29/EC | law | European Union | The EU consumer-protection directive against unfair, misleading and aggressive commercial practices, whose blacklist was extended to fake and unverified consumer reviews by Directive (EU) 2019/2161. It reaches AI-generated claims, reviews and chatbot answers. |
| Platform Work Directive (EU) 2024/2831 | law | European Union | The EU directive on platform work, including the first EU rules on algorithmic management: limits on data processing, transparency, human oversight and human review of automated monitoring and decision-making systems. Transposition by 2 December 2026. |
| Consumer Credit Directive (EU) 2023/2225 | law | European Union | The revised EU consumer credit directive. Its creditworthiness rules give consumers a right to human intervention, an explanation and a review where the assessment involves automated processing; Member States apply it from 20 November 2026. |
| ISO/IEC 42001 | standard | ISO/IEC | The AI management-system (AIMS) standard (2023); Annex A groups control objectives into nine areas (A.2–A.10). It is a management-system standard, not the Article 17 QMS, and its European adoption confers no presumption of conformity. |
| ISO/IEC 42005 | standard | ISO/IEC | Guidance for AI system impact assessment (2025); the natural companion to EU AI Act Article 27 (FRIA) and ISO/IEC 42001 Annex A.5. |
| ISO/IEC 42006:2025 | standard | ISO/IEC | Requirements for bodies providing audit and certification of AI management systems (2025). It builds on ISO/IEC 17021-1 and sets the competence and consistency a certification body must meet to credibly certify an organisation to ISO/IEC 42001: in short, who may credibly certify you to 42001. |
| ISO/IEC 23894:2023 | standard | ISO/IEC | Guidance on AI risk management (2023), adapting ISO 31000 to AI. It gives organisations a process for identifying, analysing and treating AI-specific risk; a companion to EU AI Act Article 9 and the NIST AI RMF. |
| ISO/IEC 22989:2022 | standard | ISO/IEC | AI concepts and terminology (2022): the shared vocabulary for AI systems, the AI system life cycle and AI stakeholder roles that the other SC 42 standards and many registries reuse. |
| NIST AI RMF | framework | NIST | The AI Risk Management Framework 1.0 (January 2023; there is no 2.0). Voluntary and US-origin, it organises risk work into four functions (Govern, Map, Measure, Manage) that map cleanly onto the five-layer stack. |
| NIST AI Agent Standards Initiative | framework | NIST (CAISI) | A NIST Center for AI Standards and Innovation initiative (launched 17 February 2026) to develop interoperable, secure standards for AI agents, spanning agent identity, authentication, authorisation and agent security. |
| NIST IR 8596 Cyber AI Profile (draft) | framework | NIST | A draft Cybersecurity Framework (CSF 2.0) profile for AI, the Cyber AI Profile, organised around Secure, Defend and Thwart. Initial preliminary draft released 16 December 2025 (comments closed 30 January 2026); still in draft as of 2026-09-24. |
| NIST AI 800-1 (draft) | framework | NIST | Draft voluntary guidance (Managing Misuse Risk for Dual-Use Foundation Models) for identifying, measuring and mitigating the misuse risk of dual-use foundation models across the AI lifecycle. Second public draft January 2025; still in draft as of 2026-09-24. |
| NIST AI 600-1 Generative AI Profile | framework | NIST | The AI RMF profile for generative AI (26 July 2024): 12 risks that generative AI creates or exacerbates and suggested actions coded to the Govern, Map, Measure and Manage functions. Voluntary. |
| CSA AI Controls Matrix (AICM) v1.1 | controls | Cloud Security Alliance | A control framework (published 22 June 2026) defining 247 control objectives across 18 domains, spanning governance, data, model and runtime, with crosswalks to ISO 42001 and NIST AI RMF. |
| CSA STAR for AI | framework | Cloud Security Alliance | The assurance and certification programme built on the AICM, with a self-assessment tier, an automated "Valid-AI-ted" tier and a Level 2 combining third-party ISO/IEC 42001 certification with the validated assessment. |
| OWASP Top 10 for Agentic Applications 2026 | framework | OWASP GenAI Security Project | The agent threat catalogue (ASI01 Agent Goal Hijack … ASI10 Rogue Agents) that the runtime controls are built against. |
| OWASP Top 10 for LLM Applications 2026 | framework | OWASP GenAI Security Project | The LLM threat catalogue, including Excessive Agency at #3, answered by prompt-injection and output-handling controls and an eval gate. |
| OWASP Agent Control Standard (ACS) | standard | OWASP GenAI Security Project | A standard for expressing agent controls as machine-readable control definitions the stack consumes directly. |
| OWASP AIBOM | standard | OWASP GenAI Security Project | The AI bill-of-materials format and generator, producing CycloneDX ML-BOM and SPDX 3.0 AI output at build. |
| California SB 53 (TFAIA) | law | State of California | A frontier-AI transparency law in force 1 January 2026, binding large frontier developers (models above ~10^26 FLOP; developer revenue over USD 500M) to publish a frontier AI framework, and every frontier developer to report critical safety incidents to the California Office of Emergency Services within 15 days. |
| New York RAISE Act | law | State of New York | A frontier-AI safety law (S6953B) binding large frontier developers to publish a safety framework and every frontier developer to report critical safety incidents. Signed 19 December 2025 and effective 1 January 2027 after a March 2026 chapter amendment, with oversight in an office within the New York Department of Financial Services (DFS). |
| California AB 2013 (training-data transparency) | law | State of California | A California law (chaptered 28 September 2024) requiring developers of generative AI systems made available to Californians to post a summary of their training datasets on or before 1 January 2026 and on each substantial modification. |
| California AI Transparency Act (SB 942 as amended by AB 853) | law | State of California | A California provenance law: covered providers of public generative AI systems offer a free detection tool and embed latent disclosures in generated image, video and audio from 2 August 2026; large online platforms follow from 1 January 2027 and capture devices from 1 January 2028. |
| California SB 243 (companion chatbots) | law | State of California | A California law on companion chatbots (chaptered 13 October 2025): AI disclosure, three-hourly reminders and content limits for known minors, a suicide and self-harm protocol, annual reports from 1 July 2027 and a private right of action. |
| California CPPA regulations (ADMT, risk assessments, cybersecurity audits) | law | State of California | Regulations under the California Consumer Privacy Act, approved on 23 September 2025 and effective 1 January 2026: automated decisionmaking technology duties for significant decisions from 1 January 2027, risk assessments, and cybersecurity audits phased from 2028. |
| New York General Business Law Article 47 (AI companion models) | law | State of New York | A New York law on AI companion models: a protocol to detect suicidal ideation and self-harm and refer users to crisis services, and a notice that the user is not talking to a human at the start and at least every three hours, enforced by the Attorney General. |
| Illinois HB 3773 (AI in employment) | law | State of Illinois | An amendment to the Illinois Human Rights Act (Public Act 103-0804), effective 1 January 2026: employers may not use AI with a discriminatory effect on protected classes, nor ZIP codes as a proxy, and must notify employees and applicants. |
| New York City Local Law 144 (automated employment decision tools) | law | City of New York | A New York City law, enforced since 5 July 2023: an independent bias audit of an automated employment decision tool within one year before use, a published summary of the results and notices to candidates and employees. |
| Utah AI disclosure law (AI Policy Act as amended by SB 226) | law | State of Utah | Utah's generative-AI disclosure duties, re-enacted by SB 226 with effect from 7 May 2025: disclosure when a person clearly asks, prominent disclosure in high-risk interactions by regulated occupations, and a safe harbour for clear disclosure; the AI Policy Act itself is repealed on 1 July 2027. |
| Colorado SB21-169 (insurers' use of external consumer data) | law | State of Colorado | A Colorado law (effective 7 September 2021) barring insurers from unfair discrimination through external consumer data, algorithms and predictive models, with a risk-management framework, testing and a chief-risk-officer attestation under rules adopted per line of insurance. |
| Virginia Consumer Data Protection Act | law | Commonwealth of Virginia | Virginia's consumer privacy law. Controllers document data protection assessments for targeted advertising, sale, risky profiling and sensitive data for processing created after 1 January 2023, available to the Attorney General on request. |
| Colorado Privacy Act (SB21-190) | law | State of Colorado | Colorado's consumer privacy law, effective 1 July 2023: opt-outs incl. profiling in furtherance of significant decisions, data protection assessments and a universal opt-out mechanism. |
| Minnesota Consumer Data Privacy Act | law | State of Minnesota | Minnesota's consumer privacy law, effective 31 July 2025, whose consumer rights include questioning the result of profiling, learning the reason, reviewing and correcting the data and having the decision re-evaluated. |
| Illinois Biometric Information Privacy Act (BIPA) | law | State of Illinois | The Illinois biometric privacy statute (Public Act 95-994, effective 3 October 2008): informed written consent before collecting biometric identifiers, a retention and destruction schedule and a private right of action with statutory damages. |
| Washington My Health My Data Act | law | State of Washington | Washington's consumer health data law (from 31 March 2024): consent to collect and separate consent to share consumer health data, including data derived by algorithms or machine learning, and a signed authorisation for any sale. |
| OMB Memorandum M-25-21 (federal use of AI) | framework | United States (federal) | The OMB memorandum of 3 April 2025 that binds US federal agencies: a definition of high-impact AI and minimum risk-management practices for it, documented within 365 days. It rescinded M-24-10. |
| OMB Memorandum M-26-04 (unbiased AI principles in procurement) | framework | United States (federal) | The OMB memorandum of 11 December 2025 implementing EO 14319: agencies updated procurement policies by 11 March 2026, and solicitations for large language models request minimum transparency from vendors. |
| Equal Credit Opportunity Act, Regulation B (12 CFR 1002) | law | United States (federal) | The US fair-lending rule that requires a creditor taking adverse action to give the specific principal reasons, whatever the complexity of the model behind the decision. |
| Fair Credit Reporting Act (15 U.S.C. 1681m) | law | United States (federal) | The US consumer-report statute whose adverse-action notice includes, since 21 July 2011, the numerical credit score used and its key factors. |
| Title VII and the Uniform Guidelines on Employee Selection Procedures | law | United States (federal) | US employment-discrimination law: disparate impact under Title VII s. 703(k) (Civil Rights Act of 1991) and the four-fifths rule of 29 CFR 1607.4(D), which apply to AI selection tools as to any other selection procedure. |
| FTC Act s. 5 (15 U.S.C. 45) | law | United States (federal) | The US prohibition of unfair or deceptive acts or practices, under which the Federal Trade Commission requires competent and reliable evidence for claims about an AI system's accuracy or performance. |
| TAKE IT DOWN Act (Public Law 119-12) | law | United States (federal) | A US federal law enacted on 19 May 2025 that criminalises knowing publication of non-consensual intimate images, including digital forgeries, and requires covered platforms to run a notice-and-removal process by 19 May 2026, removing reported images within 48 hours. |
| GAO AI Accountability Framework (GAO-21-519SP) | framework | U.S. Government Accountability Office | An audit framework published on 30 June 2021: four principles (governance, data, performance, monitoring) and 31 key practices, each with questions for the entity and procedures for auditors and third-party assessors. Written for federal agencies and other entities; binds no one. |
| Texas TRAIGA (HB 149) | law | State of Texas | The Texas Responsible Artificial Intelligence Governance Act (HB 149), in force 1 January 2026. Intent-based prohibitions (e.g. behaviour manipulation, unlawful discrimination, and social scoring by governmental entities), AI-use disclosure by government agencies and health care providers, and a regulatory sandbox, enforced by the Attorney General and preempting local AI rules. |
| Colorado ADMT law (SB 26-189, replacing SB 24-205) | law | State of Colorado | Colorado's law on automated decision-making technology in consequential decisions, SB 26-189 (signed 14 May 2026, effective 1 January 2027): developer documentation, deployer notice, an explanation within 30 days of an adverse outcome, human review and three-year records. It repealed and re-enacted SB 24-205, the Colorado AI Act, whose duty of care against algorithmic discrimination had been delayed to 30 June 2026. |
| South Korea AI Basic Act | law | Republic of Korea | South Korea's framework Act on AI (Basic Act on the Development of AI and the Establishment of a Foundation for Trust, Act No. 20676), in force 22 January 2026 with its Enforcement Decree. It sets baseline duties for AI operators and heightened obligations for "high-impact" AI in sensitive sectors; the ministry (MSIT) announced a guidance period of at least one year in which fact-finding and fines are held back except in exceptional cases, while the duties apply. |
| UK Data (Use and Access) Act 2025 | law | United Kingdom | The UK has no horizontal AI act; it governs AI through sector regulators and the AI Security Institute. For automated decision-making, the Data (Use and Access) Act 2025 replaced UK GDPR Article 22 with Articles 22A–22D (in force 5 February 2026): a permission-plus-safeguards regime for significant, solely automated decisions. |
| UK Digital Markets, Competition and Consumers Act 2024 | law | United Kingdom | The UK consumer-protection regime in force since 6 April 2025: unfair commercial practices are prohibited, and Schedule 20 bans fake and concealed-incentive consumer reviews, however they are produced. |
| UK Algorithmic Transparency Recording Standard (ATRS) v4.0 | standard | UK Government Digital Service | A two-tier record template for public-sector algorithmic tools, mandatory for government departments and for arm's-length bodies that deliver public or frontline services under the December 2024 scope policy, and recommended across the wider public sector (as of 2026-09-24). |
| Singapore Model AI Governance Framework for Generative AI | framework | IMDA / AI Verify Foundation | Singapore's voluntary Model AI Governance Framework for Generative AI (IMDA and the AI Verify Foundation, May 2024). It sets out governance dimensions (testing, transparency, incident reporting, security and content provenance) as guidance, not law. |
| Singapore Model AI Governance Framework for Agentic AI | framework | IMDA | Voluntary guidance launched on 22 Jan 2026; version 1.5 (20 May 2026, updated 5 Jun 2026) has four dimensions: assess and bound the risks upfront, make humans meaningfully accountable, implement technical controls and processes, enable end-user responsibility. |
| AI Verify Testing Framework (Singapore, IMDA / AI Verify Foundation) | framework | IMDA / AI Verify Foundation | Singapore's voluntary testing framework for traditional and generative AI (updated 29 May 2025): process checks and technical tests against 11 governance principles. IMDA publishes official crosswalks to the NIST AI RMF, the NIST Generative AI Profile, the G7 Hiroshima Process Code of Conduct and ISO/IEC 42001. |
| China Provisions on Algorithmic Recommendation (2022) | law | CAC, MIIT, MPS and SAMR (China) | China's binding Provisions on Algorithmic Recommendation (CAC, MIIT, MPS and SAMR, Order No. 9), in force 1 March 2022. Algorithm filing for services with public-opinion or social-mobilisation capacity, a security assessment, display of the filing number, and a user option to switch off personalised recommendation. |
| China Provisions on Deep Synthesis (2023) | law | CAC, MIIT and MPS (China) | China's binding Provisions on Deep Synthesis (CAC, MIIT and MPS, Order No. 12), in force 10 January 2023. Conspicuous labels where synthetic content could mislead the public and non-removable technical marks, training-data management, separate consent for face and voice editing, and filing plus a security assessment for opinion-shaping functions. |
| China Interim Measures for Generative AI Services (2023) | law | CAC and six other bodies (China) | China's binding Interim Measures for Generative AI Services (CAC and six other bodies, Order No. 15), in force 15 August 2023, for services offered to the public within the PRC. Lawful-source training data and foundation models, content labelling under the deep-synthesis rules, a security assessment and algorithm filing for opinion-shaping services, and a duty to stop, remove, retrain and report on illegal content. |
| China Measures for Labelling AI-Generated Synthetic Content (2025) | law | CAC, MIIT, MPS and NRTA (China) | China's binding Measures for Labelling AI-Generated Synthetic Content (CAC, MIIT, MPS and NRTA), in force 1 September 2025 alongside the mandatory standard GB 45438-2025. Explicit labels (text, audio or graphic) and implicit metadata labels carrying the provider's name or code and a content number; distribution platforms verify the metadata and flag suspected AI content. |
| GB/T 45654-2025 Basic security requirements for generative AI services | standard | SAMR / SAC, drafted by TC260 (China) | GB/T 45654-2025 Basic security requirements for generative AI services (SAMR / SAC, drafted by TC260), a recommended (voluntary) national standard implemented 1 November 2025. Training-corpus source and content screening, model-safety requirements and the evaluation methods that underpin the security assessment. |
| TC260 AI Safety Governance Framework 3.0 | framework | TC260 under CAC guidance (China) | China's voluntary AI safety governance framework (14 September 2026, building on 1.0 in 2024 and 2.0 in 2025): a three-block risk taxonomy, technological and governance countermeasures and role-based guidelines. Appendix 2 is an agentic AI risk-management framework covering identity, human checkpoints, tool control, runtime guardrails, memory, auditing and decommissioning. |
| China Personal Information Protection Law (PIPL) | law | NPC Standing Committee (China) | China's personal-information law, in force 1 November 2021: legal bases, separate consent for sensitive data, rules for automated decision-making (Art. 24) and a personal information protection impact assessment kept at least three years (Arts. 55 and 56). |
| China Interim Measures for Anthropomorphic Interaction Services (2026) | law | CAC, NDRC, MIIT, MPS and SAMR (China) | China's binding measures for AI services that offer sustained emotional interaction, in force 15 July 2026: a minors' mode, AI signals and a two-hour reminder, an easy exit, a security assessment at 1 million registered or 100,000 monthly active users, and filing. |
| Canada Directive on Automated Decision-Making | law | Government of Canada (Treasury Board) | The Treasury Board directive that binds Canadian federal institutions' automated decision systems (in effect since 1 April 2019; modified 24 June 2025): a published algorithmic impact assessment, impact-level requirements, notice, explanation, peer review and recourse. |
| Brazil General Data Protection Law (LGPD, Lei 13.709/2018) | law | Federative Republic of Brazil | Brazil's data-protection law. Its Article 20 gives data subjects a right to request review of decisions taken solely on automated processing, with clear information on the criteria and procedures used. |
| Council of Europe Framework Convention on AI (CETS No. 225) | law | Council of Europe | A treaty on AI and human rights, democracy and the rule of law, binding on Parties once in force; not in force as of 2026-09-24. The EU implements it through the AI Act. |
| OECD AI Principles (Recommendation on AI, OECD/LEGAL/0449) | framework | OECD | Five values-based principles for AI actors and five recommendations to governments, revised on 3 May 2024; a non-binding intergovernmental standard. |
| G7 Hiroshima Process International Code of Conduct for Advanced AI Systems | code | G7 | Eleven voluntary actions for organisations developing advanced AI systems, from lifecycle risk management to content provenance and data protection. |
| ETSI EN 304 223 | standard | ETSI | Securing Artificial Intelligence (SAI); Baseline Cyber Security Requirements for AI Models and Systems (V2.1.1, December 2025). A cross-border European standard setting 13 security principles across the five stages of the AI lifecycle: a cyber-security baseline, not an EU AI Act harmonised standard. |
| EN 18286:2026 | standard | CEN-CENELEC | The Article 17 QMS standard, published July 2026 (the first JTC 21 AI Act standard to reach publication) but not yet cited in the Official Journal, so it carries no presumption of conformity. |
| prEN 18228 AI risk management (draft) | standard | CEN-CENELEC JTC 21 | Draft harmonised standard supporting AI Act Art. 9; its Enquiry vote closed on 30 Jul 2026, as reported by Genorma on 2026-09-24. |
| prEN 18229-1 AI trustworthiness framework, Part 1: logging (draft) | standard | CEN-CENELEC JTC 21 | Draft harmonised standard supporting AI Act Art. 12; its Enquiry vote closed on 20 Aug 2026, as reported by Genorma on 2026-09-24. |
Obligation → artefact → layer
182 rows, grouped by framework. Read each as a sentence: this obligation is answered by this artefact, which lives in this layer. Filter by the layer you are building, or by EU AI Act duty holder.
Cross-reference: the obligation register (one page per row, with a stable id) · tools by layer · the patterns · topic crosswalk.
Obligation heat matrix
| Framework | |||||
|---|---|---|---|---|---|
EU AI Act
50 rows| Obligation | Duty holder | Artefact | Layers | Applies from | Chapter |
|---|---|---|---|---|---|
| EU AI Act Art. 3(1) AI system definition (scope of the Act) AIGE-OBL-EUAIA-ART3-1 | Provider + deployer (scoping) | Definitional decision record in the registry: definition applied, elements found, excluded family if any, reason, decider, date | L2 | 2025-02-02 (Chapter I) | ch. 08 → |
| EU AI Act Art. 4 AI literacy AIGE-OBL-EUAIA-ART4 | Provider + deployer | Literacy programme as code; role-based training records; onboarding gates | L1 | 2025-02-02; reworded 2026-07-27 (in force) | ch. 08 → |
| EU AI Act Art. 4a lawful basis for special-category data in bias detection AIGE-OBL-EUAIA-ART4A | Provider | Data governance controls; pseudonymisation and retention-as-code; data card noting basis and deletion | L2 | 2026-07-27 (new, in force) | ch. 08 → |
| EU AI Act Art. 5 prohibited practices (incl. new NCII and CSAM bans) AIGE-OBL-EUAIA-ART5 | Provider + deployer | Policy-as-code blocklist; input/output guardrails; refusal and abuse detection | L1 L4 | 2026-12-02 (new bans); earlier prohibitions from 2025-02-02 | ch. 08 → |
| EU AI Act Art. 6 classification of high-risk AI systems (incl. the Annex III route) AIGE-OBL-EUAIA-ART6 | Provider | Risk-tiering as code; high-risk classification decision record; register entry flagging Annex III status | L1 L2 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 6(3)–(4) documented non-high-risk assessment and registration AIGE-OBL-EUAIA-ART6-3 | Provider | Classification decision record (Annex III point, Art. 6(3) condition, explicit profiling flag); Art. 49(2) registration entry pushed from the registry | L1 L2 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 9 risk management system AIGE-OBL-EUAIA-ART9 | Provider | Risk register as code; threat models; linkage to FRIA and eval results | L1 L3 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 10 data and data governance AIGE-OBL-EUAIA-ART10 | Provider | Data cards; lineage; bias and quality tests in CI | L2 L3 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 11 technical documentation (Annex IV) AIGE-OBL-EUAIA-ART11 | Provider | AIBOM (CycloneDX ML-BOM, SPDX 3.0 AI); auto-generated technical documentation; model cards | L2 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 12 record-keeping and logging AIGE-OBL-EUAIA-ART12 | Provider | Structured, signed logs; OpenTelemetry traces; tamper-evident event store | L4 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 13 transparency and information to deployers AIGE-OBL-EUAIA-ART13 | Provider | Instructions for use as code; model and data cards; capability and limitation notes | L2 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 14 human oversight AIGE-OBL-EUAIA-ART14 | Provider | Human-in-the-loop checkpoints; kill switch; override and escalation paths | L4 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 15 accuracy, robustness and cybersecurity AIGE-OBL-EUAIA-ART15 | Provider | Eval gate; adversarial red-team suite; robustness and security controls; regression evals | L3 L4 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 15(4) feedback loops in systems that continue to learn AIGE-OBL-EUAIA-ART15-4 | Provider | Feedback-loop fairness monitor; retraining-data bias check; agent memory write gate with provenance and rollback to a known-good snapshot | L3 L4 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 16(l) accessibility requirements for high-risk AI systems AIGE-OBL-EUAIA-ART16-L | Provider | Accessibility test results for every notice, instruction and explanation shown to people, run in the pipeline; accessible explanation templates | L2 L3 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 17 quality management system AIGE-OBL-EUAIA-ART17 | Provider | QMS-as-code; versioned policies; pipeline controls and change management | L1 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 17(1)(m) accountability framework within the quality management system AIGE-OBL-EUAIA-ART17-1M | Provider | RACI as code compiled into registry owner fields and code-owner rules; committee decision records | L1 L2 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 18 documentation keeping AIGE-OBL-EUAIA-ART18 | Provider | Retention-as-code for the technical file, QMS records, notified-body decisions and the EU declaration (10 years); write-once evidence store | L2 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 19 automatically generated logs kept by the provider AIGE-OBL-EUAIA-ART19 | Provider | Log-retention policy as code (at least six months, set by intended purpose); tamper-evident log store | L4 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 20 corrective actions and duty of information AIGE-OBL-EUAIA-ART20 | Provider | CAPA record; withdraw, disable or recall runbook; notification of distributors and deployers | L1 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 22 authorised representative of non-EU high-risk providers AIGE-OBL-EUAIA-ART22 | Non-EU provider + authorised representative | Written mandate; 10-year copies of the declaration, technical documentation and certificate; authority contact route | L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 23 obligations of importers AIGE-OBL-EUAIA-ART23 | Importer | Import verification record against each check; 10-year document copies | L2 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 24 obligations of distributors AIGE-OBL-EUAIA-ART24 | Distributor | Distribution check record; hold, withdraw or recall workflow | L2 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 25 responsibilities along the AI value chain AIGE-OBL-EUAIA-ART25 | Provider + value-chain actors | Value-chain due-diligence gate; provider/deployer responsibility allocation; AIBOM and model/data cards collected from upstream providers | L2 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 26 deployer obligations for high-risk systems AIGE-OBL-EUAIA-ART26 | Deployer | Deployment registry; monitoring hooks; assigned oversight and logging retention | L2 L4 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 26(2) human oversight assigned to persons with competence, training and authority AIGE-OBL-EUAIA-ART26-2 | Deployer | Oversight assignment in the registry; role-based training records with expiry; approver roster per checkpoint class with authority to pause or refuse | L1 L4 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 26(4) input data relevant and sufficiently representative AIGE-OBL-EUAIA-ART26-4 | Deployer | Input-data checks against the population in the deployment record; drift monitors on inputs | L2 L3 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 26(5) deployer monitoring, suspension and informing the provider AIGE-OBL-EUAIA-ART26-5 | Deployer | Monitoring plan with signal owners; tested suspension path (feature flag, traffic switch); provider incident contact in the registry; pre-filled risk and serious-incident notices | L2 L4 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 26(6) deployer retention of automatically generated logs AIGE-OBL-EUAIA-ART26-6 | Deployer | Retention-as-code schedule; tamper-evident log store; legal hold while an incident is open | L4 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 26(7) informing workers before workplace use AIGE-OBL-EUAIA-ART26-7 | Deployer (employer) | Worker-information record dated before first use and linked to the registry entry; HR intake trigger for Annex III point 4 systems | L2 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 26(11) informing people subject to Annex III decisions AIGE-OBL-EUAIA-ART26-11 | Deployer | AI-use notice at the decision point; notice templates versioned as code; notice delivery log | L2 L4 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 27 Fundamental Rights Impact Assessment (FRIA) AIGE-OBL-EUAIA-ART27 | Deployer | FRIA-as-code from a template; cross-reference to a GDPR Art. 35 DPIA | L1 L2 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 43 conformity assessment AIGE-OBL-EUAIA-ART43 | Provider | Conformity-assessment workflow; internal-control or notified-body evidence pack; traceability to Annex IV documentation | L1 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 43(4) new conformity assessment on substantial modification AIGE-OBL-EUAIA-ART43-4 | Provider | Change-classification policy on merge; pre-determined change envelope as code; reassessment trigger in the registry | L1 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 47 EU declaration of conformity AIGE-OBL-EUAIA-ART47 | Provider | Auto-generated EU declaration of conformity from the evidence; CE-marking record | L2 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 48 CE marking AIGE-OBL-EUAIA-ART48 | Provider | CE-marking record (physical or digital) generated with the EU declaration; notified-body number where applicable | L2 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 49/71 registration of high-risk systems in the EU database AIGE-OBL-EUAIA-ART49-71 | Provider; public-authority deployer | Agent/model registry with an API that feeds registration; owner and status per entry | L2 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 50 transparency for certain AI systems AIGE-OBL-EUAIA-ART50 | Provider + deployer | Content labelling and machine-readable marking (e.g. C2PA-style); chatbot disclosure banner | L4 L2 | 2026-08-02; marking grace for existing systems to 2026-12-02 | ch. 08 → |
| EU AI Act Art. 52 notification of a GPAI model meeting the systemic-risk threshold AIGE-OBL-EUAIA-ART52 | GPAI provider | Compute ledger per model lineage with a planned-compute threshold alert; notification record filed within two weeks | L2 L5 | Obligations from 2025-08-02; enforcement from 2026-08-02 | ch. 08 → |
| EU AI Act Art. 53 GPAI provider obligations AIGE-OBL-EUAIA-ART53 | GPAI provider | Model cards; training-content summary; AIBOM and dataset provenance | L2 | Obligations from 2025-08-02; enforcement from 2026-08-02 | ch. 08 → |
| EU AI Act Art. 53(1)(c) copyright policy honouring text-and-data-mining reservations AIGE-OBL-EUAIA-ART53-1C | GPAI provider | Versioned copyright policy; crawler decision logs; training-data rights ledger with the reservation-check result | L1 L2 L5 | Obligations from 2025-08-02; enforcement from 2026-08-02 | ch. 08 → |
| EU AI Act Art. 54 authorised representative of non-EU GPAI providers AIGE-OBL-EUAIA-ART54 | Non-EU GPAI provider + authorised representative | Written mandate; Annex XI documentation copy kept 10 years; AI Office contact route | L5 | Obligations from 2025-08-02; enforcement from 2026-08-02 | ch. 08 → |
| EU AI Act Art. 55 GPAI models with systemic risk AIGE-OBL-EUAIA-ART55 | GPAI provider (systemic risk) | Eval and red-team suite; incident pipeline on the Commission serious-incident reporting template; weight-security controls; threat model | L3 L4 L5 | Obligations from 2025-08-02; enforcement from 2026-08-02 | ch. 08 → |
| EU AI Act Art. 60 testing in real-world conditions outside sandboxes AIGE-OBL-EUAIA-ART60 | Provider / prospective provider | Real-world testing plan; Art. 61 informed-consent records; test monitoring, logging and incident hooks | L3 L4 | 2026-08-02 | ch. 08 → |
| EU AI Act Art. 72 post-market monitoring AIGE-OBL-EUAIA-ART72 | Provider | Continuous assurance telemetry; monitoring plan; drift and performance signals | L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 73 serious-incident reporting AIGE-OBL-EUAIA-ART73 | Provider | Incident detection and triage pipeline; reporting-clock automation; evidence capture | L5 L4 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 73(6) incident investigation without altering the system AIGE-OBL-EUAIA-ART73-6 | Provider | Evidence-preservation step: model, prompt, policy and retrieval-index snapshots; sealed traces; fix shipped on a new version | L4 L5 | 2027-12-02 (Annex III) | ch. 08 → |
| EU AI Act Art. 75(1a) serious incidents reported to the AI Office AIGE-OBL-EUAIA-ART75-1A | Provider (systems under AI Office competence) | Incident pipeline routing by a registry competence flag (national MSA or AI Office) | L2 L5 | 2027-12-02 (Annex III); new by the Omnibus (reading, verify) | ch. 08 → |
| EU AI Act Art. 86 right to explanation of individual decision-making AIGE-OBL-EUAIA-ART86 | Deployer | Explanation record per decision (system and model version, reason codes, determinative or advisory role, human decider); request-handling workflow and response log | L2 L4 L5 | 2027-12-02 (Annex III); Chapter IX applies from 2026-08-02, but the right needs a classified Annex III system (reading, verify) | ch. 08 → |
| EU AI Act Art. 87 reporting of infringements and protection of reporting persons AIGE-OBL-EUAIA-ART87 | Legal entities with internal reporting channels under Directive (EU) 2019/1937 | Internal reporting channel run as a case system with the directive's clocks encoded; sealed reporter identity; retaliation monitoring; routing to the incident pipeline | L1 L5 | 2026-08-02 (general application date) | ch. 08 → |
GPAI Code of Practice
5 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| Safety and Security (systemic-risk models only) AIGE-OBL-GPAICOP-SAFETY | Eval and red-team suite; adversarial testing harness; incident pipeline; weight-security controls | L3 L4 L5 | ch. 08 → |
| Transparency AIGE-OBL-GPAICOP-TRANSPARENCY | Model cards; structured model documentation; AIBOM | L2 | ch. 08 → |
| Copyright AIGE-OBL-GPAICOP-COPYRIGHT | Training-data provenance and licence records; policy-as-code for source filtering | L1 L2 | ch. 08 → |
| Safety and Security Commitment 9: serious-incident reporting AIGE-OBL-GPAICOP-SAFETY-C9 | Incident pipeline on the Commission template fields; per-class clocks; five-year retention policy; downstream reporting channel | L4 L5 | ch. 08 → |
| Safety and Security Appendix 1.3 and 1.4: autonomy, tool use and loss of control as systemic risks AIGE-OBL-GPAICOP-SAFETY-APP1 | Provider evaluations of autonomy and tool use, requested at the vendor due-diligence gate and filed with the agent registry entry | L3 L5 | ch. 08 → |
GDPR
15 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| GDPR Art. 5(1)(b) and 6(4) purpose limitation AIGE-OBL-GDPR-ART5-1B | Purpose tags on datasets; purpose-match policy in training and indexing pipelines; compatibility assessment record | L1 L2 | ch. 08 → |
| GDPR Art. 6 lawful basis per processing moment AIGE-OBL-GDPR-ART6 | Basis registry per dataset and stage; versioned legitimate-interest assessment | L2 | ch. 08 → |
| GDPR Art. 7 conditions for consent and its withdrawal AIGE-OBL-GDPR-ART7 | Consent-purpose log joined to datasets and model versions; withdrawal propagated to the pipelines | L2 L5 | ch. 08 → |
| GDPR Art. 9 special categories, incl. inferred sensitive data AIGE-OBL-GDPR-ART9 | Proxy test in CI; inference policy with a runtime output classifier; Art. 9(2) condition record | L1 L3 L4 | ch. 08 → |
| GDPR Arts. 13–14 transparency to data subjects AIGE-OBL-GDPR-ART13-14 | Notice generated from the registry entry; model card; automated-decision notice per system | L2 | ch. 08 → |
| GDPR Art. 15(1)(h) access to meaningful information about the logic involved AIGE-OBL-GDPR-ART15-1H | Per-request explanation generated from the decision record; system-level automated-decision notice | L4 L5 | ch. 08 → |
| GDPR Arts. 15–17 and 21 data subject rights against trained models AIGE-OBL-GDPR-ART15-17-21 | Request workflow across corpus, snapshots, retrieval index, logs and weights; fulfilment record | L4 L5 | ch. 08 → |
| GDPR Art. 22 solely automated decisions and their safeguards AIGE-OBL-GDPR-ART22 | Decision record with reason codes; contest channel and human review log; appeal outcomes by group | L4 L5 | ch. 08 → |
| GDPR Art. 5(1)(c) and 25 minimisation and data protection by design and by default AIGE-OBL-GDPR-ART25 | Feature justification record; PII and special-category filters; retention as code | L1 L3 | ch. 08 → |
| GDPR Art. 5(2) accountability for a model anonymity claim AIGE-OBL-GDPR-ART5-2 | Anonymity evidence pack; membership-inference and extraction evals in the eval gate | L3 L5 | ch. 08 → |
| GDPR Art. 28 processors, incl. AI vendors AIGE-OBL-GDPR-ART28 | AI vendor clause checklist in the due-diligence gate (no training, retention, region, sub-processors, change notice) | L2 L5 | ch. 08 → |
| GDPR Art. 30 records of processing activities AIGE-OBL-GDPR-ART30 | Records generated per processing moment from the registry and data cards | L2 L5 | ch. 08 → |
| GDPR Arts. 33–34 personal data breach notification AIGE-OBL-GDPR-ART33-34 | Personal-data-breach branch of the incident pipeline with its own 72-hour timer; data-subject notice template | L4 L5 | ch. 08 → |
| GDPR Arts. 35–36 DPIA and prior consultation AIGE-OBL-GDPR-ART35-36 | AI DPIA template with AI-specific fields, cross-referenced by the FRIA; recorded decision when no DPIA is needed | L1 L2 | ch. 08 → |
| GDPR Arts. 44–49 international transfers, incl. remote inference AIGE-OBL-GDPR-ART44-49 | Transfer register; residency and routing policy as code; transfer impact assessment | L1 L4 L5 | ch. 08 → |
NIS2
2 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| NIS2 Art. 21(2)(c)–(d) business continuity and supply-chain security AIGE-OBL-NIS2-ART21-2 | Continuity plan for AI dependencies with tested fallbacks; supplier assessments for model and platform providers | L4 L5 | ch. 08 → |
| NIS2 Art. 23 significant-incident reporting AIGE-OBL-NIS2-ART23 | Per-regime clock on the incident record; significance determination with an owner; cause coding reused in the final report | L5 | ch. 08 → |
DORA
2 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| DORA Art. 19 major ICT-related incident reporting AIGE-OBL-DORA-ART19 | Classification record with a timestamp; per-regime clock; consistent cause coding for recurring-incident aggregation | L5 | ch. 08 → |
| DORA Art. 28(3) and 28(8) register of ICT third-party arrangements and exit strategies AIGE-OBL-DORA-ART28 | Register entries for AI and model services; exit plan and exit-drill record | L2 L5 | ch. 08 → |
Cyber Resilience Act
1 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| Cyber Resilience Act Art. 14 reporting of actively exploited vulnerabilities and severe incidents AIGE-OBL-CRA-ART14 | Vulnerability and incident clocks on the incident record; submission through the single reporting platform | L4 L5 | ch. 08 → |
EU Product Liability Directive
3 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| EU Product Liability Directive Art. 4(1) software, incl. AI systems, as a product AIGE-OBL-PLD-ART4 | Defect-liability review at design; contractual recourse; residual risk in the register | L1 L5 | ch. 08 → |
| EU Product Liability Directive Arts. 9–10 disclosure of evidence and presumption of defect AIGE-OBL-PLD-ART9-10 | Defence file per release: AIBOM with hashes, eval history, failure-mode analysis, signed logs, instructions for use, kept for the liability period | L2 L3 L5 | ch. 08 → |
| EU Product Liability Directive Art. 11(2) no later-defect defence for software, its updates or missing safety updates in the manufacturer's control AIGE-OBL-PLD-ART11-2 | Change log; regression evals per release; patch decision records; versioned warnings | L3 L4 L5 | ch. 08 → |
DSM Directive
1 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| DSM Directive Art. 4(3) text-and-data-mining reservations AIGE-OBL-DSM-ART4-3 | Crawler policy-as-code honouring reservations; training-data rights ledger with the reservation-check result, method and date | L1 L2 | ch. 08 → |
Digital Services Act
2 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| Digital Services Act Art. 25 no deceptive or manipulative interface design AIGE-OBL-DSA-ART25 | Interface review record; red-team eval for manipulative outputs | L3 L5 | ch. 08 → |
| Digital Services Act Art. 27 recommender system transparency AIGE-OBL-DSA-ART27 | Recommender parameter card generated from the ranking configuration; log of the options offered to users | L2 L4 | ch. 08 → |
UCPD
1 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| UCPD Arts. 5–7 and Annex I unfair and misleading commercial practices, incl. fake reviews AIGE-OBL-UCPD-ART5-7 | Claims register linked to eval results; review-provenance checks; chatbot answer evals on product claims | L1 L3 L4 | ch. 08 → |
Platform Work Directive
1 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| Platform Work Directive Arts. 7 and 9–11 automated monitoring and decision-making systems AIGE-OBL-PWD-ART7-11 | Registry of automated systems with their main parameters; data-category deny list; two-yearly impact evaluation; explanation and human review log | L1 L2 L3 L5 | ch. 08 → |
Consumer Credit Directive
1 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| Consumer Credit Directive Art. 18(8) human intervention in automated creditworthiness assessment AIGE-OBL-CCD2-ART18-8 | Explanation artefact per model version; review path and log | L3 L4 L5 | ch. 08 → |
ISO/IEC 42001
9 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| A.2 Policies related to AI AIGE-OBL-ISO42001-A2 | Policy-as-code library; versioned policy repository | L1 | ch. 08 → |
| A.3 Internal organization AIGE-OBL-ISO42001-A3 | Operating model; RACI; ownership in the registry | L1 L2 | ch. 08 → |
| A.4 Resources for AI systems AIGE-OBL-ISO42001-A4 | Resource inventory; AIBOM; environment manifests | L2 | ch. 08 → |
| A.5 Assessing impacts of AI systems AIGE-OBL-ISO42001-A5 | Impact assessment as code; FRIA/DPIA linkage (ISO/IEC 42005) | L1 L3 | ch. 08 → |
| A.6 AI system life cycle AIGE-OBL-ISO42001-A6 | Pipeline controls; eval gates; change management | L1 L3 L4 | ch. 08 → |
| A.7 Data for AI systems AIGE-OBL-ISO42001-A7 | Data cards; lineage; data quality tests | L2 L3 | ch. 08 → |
| A.8 Information for interested parties AIGE-OBL-ISO42001-A8 | Model/data cards; machine-readable disclosures | L2 | ch. 08 → |
| A.9 Use of AI systems AIGE-OBL-ISO42001-A9 | Runtime guardrails; usage telemetry | L4 | ch. 08 → |
| A.10 Third-party and customer relationships AIGE-OBL-ISO42001-A10 | Supplier AIBOM; contractual and technical control mapping | L2 L5 | ch. 08 → |
ISO/IEC 42006
1 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| ISO/IEC 42006:2025 requirements for AIMS certification bodies AIGE-OBL-ISO42006-CB | Accredited certification scope; auditor-competence evidence; certificate register | L5 | ch. 08 → |
ISO/IEC 23894
1 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| ISO/IEC 23894:2023 guidance on AI risk management AIGE-OBL-ISO23894-RISK | Risk register as code; AI risk taxonomy; linkage to EU AI Act Art. 9 and the NIST AI RMF | L1 L3 | ch. 08 → |
ISO/IEC 42005
1 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| ISO/IEC 42005:2025 guidance for AI system impact assessment AIGE-OBL-ISO42005-IA | Impact assessment as code from a template; FRIA and DPIA cross-references; re-assessment triggers | L1 L3 | ch. 08 → |
ISO/IEC 22989
1 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| ISO/IEC 22989:2022 AI concepts, terminology and stakeholder roles AIGE-OBL-ISO22989-CONCEPTS | Registry field names and role vocabulary aligned to the standard's terms; glossary cross-references | L2 | ch. 08 → |
NIST AI RMF
5 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| GOVERN AIGE-OBL-NISTRMF-GOVERN | Policy-as-code; operating model; registry ownership | L1 L2 | ch. 08 → |
| MAP AIGE-OBL-NISTRMF-MAP | Threat models; use-case and impact mapping; data/model cards | L2 L3 | ch. 08 → |
| MEASURE AIGE-OBL-NISTRMF-MEASURE | Eval gates; adversarial red-team suite; metrics per failure mode | L3 | ch. 08 → |
| MANAGE AIGE-OBL-NISTRMF-MANAGE | Runtime guardrails; incident pipeline; continuous assurance | L4 L5 | ch. 08 → |
| NIST AI 600-1 Generative AI Profile AIGE-OBL-NIST-AI600-1 | Generative-AI eval suites named after the profile's action ids (e.g. confabulation, information integrity); a risk-register entry per profile risk | L1 L3 | ch. 08 → |
NIST (agent, cyber and misuse work)
3 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| NIST AI Agent Standards Initiative (2026) AIGE-OBL-NIST-AGENTS | Agent registry; non-human-identity controls; agent authentication and authorisation; adversarial agent evals | L3 L4 | ch. 08 → |
| NIST IR 8596 Cyber AI Profile (draft) AIGE-OBL-NIST-IR8596 | AI-system security controls; runtime observability; threat detection mapped to CSF 2.0 | L3 L4 | ch. 08 → |
| NIST AI 800-1 misuse risk for dual-use foundation models (draft) AIGE-OBL-NIST-AI800-1 | Misuse red-team suite; capability and dangerous-capability evals; safety framework | L3 | ch. 08 → |
CSA AICM / STAR for AI
4 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| AICM v1.1: 247 control objectives across 18 domains AIGE-OBL-CSA-AICM | Control catalogue mapped to policy-as-code and evals; crosswalk to ISO 42001 / NIST AI RMF | L1 L3 L5 | ch. 08 → |
| STAR for AI assurance and certification programme AIGE-OBL-CSA-STAR | Machine-readable evidence submission; continuous assurance telemetry | L5 | ch. 08 → |
| AICM agent controls with the CSA Agentic Trust Framework and AARM specification AIGE-OBL-CSA-AICM-AGENTIC | Agent-specific control definitions; policy-as-code for agent scope and tools; runtime guardrails | L1 L4 | ch. 08 → |
| AICM Catastrophic Risk Annex (enhanced controls for high-autonomy systems) AIGE-OBL-CSA-AICM-CATASTROPHIC | Enhanced controls for high-autonomy systems; kill switch and oversight controls; pilot-audit evidence | L4 L5 | ch. 08 → |
OWASP GenAI Security Project
4 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| Top 10 for Agentic Applications 2026 AIGE-OBL-OWASP-AGENTIC | Agent threat model; adversarial evals; runtime guardrails; kill switch | L3 L4 | ch. 08 → |
| Top 10 for LLM Applications 2026 AIGE-OBL-OWASP-LLM | Prompt-injection and output-handling controls; eval gate | L3 L4 | ch. 08 → |
| Agent Control Standard (ACS) AIGE-OBL-OWASP-ACS | Machine-readable control definitions for agents | L1 L4 | ch. 08 → |
| AIBOM AIGE-OBL-OWASP-AIBOM | AIBOM at build (CycloneDX ML-BOM, SPDX 3.0 AI) | L2 | ch. 08 → |
US frontier-developer laws
3 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| California SB 53 (TFAIA) AIGE-OBL-USCA-SB53 | Published safety framework; incident pipeline reporting to the state; transparency artefacts | L5 L4 | ch. 08 → |
| New York RAISE Act (signed 2025-12-19; effective 2027-01-01) AIGE-OBL-USNY-RAISE | Published frontier AI safety framework; 72-hour incident and disclosure pipeline reporting to the DFS oversight office | L5 L4 | ch. 08 → |
| California SB 53 whistleblower protections for covered employees AIGE-OBL-USCA-SB53-WHISTLE | Anonymous internal reporting channel with status updates; notice acknowledgment records; quarterly summary to officers and directors | L1 L5 | ch. 08 → |
US state AI laws
9 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| Texas TRAIGA (HB 149; in force 2026-01-01) AIGE-OBL-USTX-TRAIGA | Prohibited-use policy-as-code; AI-use disclosure controls; complaint and incident handling | L1 L4 | ch. 08 → |
| Colorado SB 26-189 automated decision-making technology (replaces the Colorado AI Act, SB 24-205; effective 2027-01-01) AIGE-OBL-USCO-AIACT | ADMT inventory; developer documentation pack; notice and adverse-outcome explanation templates; human-review queue; three-year record store | L2 L4 L5 | ch. 08 → |
| California AB 2013 training-data transparency for generative AI AIGE-OBL-USCA-AB2013 | Data card per dataset, published at release; training-data rights ledger | L2 | ch. 08 → |
| California AI Transparency Act (SB 942 as amended by AB 853) AIGE-OBL-USCA-SB942 | Provenance pipeline writing latent metadata; public detection endpoint; platform-side provenance display | L3 L4 | ch. 08 → |
| California SB 243 companion chatbots AIGE-OBL-USCA-SB243 | Companion-mode policy; reminder timer; crisis-referral classifier and log; annual report | L1 L4 L5 | ch. 08 → |
| New York GBL Article 47 AI companion models AIGE-OBL-USNY-GBL47 | Crisis-referral classifier and log; notice timer | L4 L5 | ch. 08 → |
| Illinois HB 3773 AI in employment decisions AIGE-OBL-USIL-HB3773 | AI-in-HR inventory; adverse-impact eval per protected class; notice record | L2 L3 L4 | ch. 08 → |
| NYC Local Law 144 automated employment decision tools AIGE-OBL-USNYC-LL144 | Impact-ratio eval by sex, race/ethnicity and intersectional category; published audit summary; notice record | L3 L5 | ch. 08 → |
| Utah AI disclosure duties (SB 226 amendments to the AI Policy Act) AIGE-OBL-USUT-SB226 | Disclosure component with an interaction-risk flag; conversation log showing the disclosure | L4 | ch. 08 → |
US state privacy and sector laws
8 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| California CPPA regulations on automated decisionmaking technology AIGE-OBL-USCA-CPPA-ADMT | ADMT register; pre-use notice; opt-out or appeal workflow; ADMT access response | L2 L4 L5 | ch. 08 → |
| California CPPA regulations on risk assessments AIGE-OBL-USCA-CPPA-RA | Risk assessment per triggering activity; submission record | L5 | ch. 08 → |
| Virginia CDPA data protection assessments, incl. risky profiling AIGE-OBL-USVA-CDPA | Assessment template per processing activity; profiling register | L1 L5 | ch. 08 → |
| Colorado Privacy Act profiling opt-out and data protection assessments AIGE-OBL-USCO-CPA | Opt-out flag honoured at inference; assessment template; intake data-class flags | L1 L2 L4 L5 | ch. 08 → |
| Minnesota CDPA right to question the result of profiling AIGE-OBL-USMN-MCDPA | Reason-and-review workflow with re-evaluation on corrected data | L4 L5 | ch. 08 → |
| Illinois BIPA consent and retention for biometric identifiers AIGE-OBL-USIL-BIPA | Written-consent capture; retention schedule as code; destruction log | L1 L2 | ch. 08 → |
| Washington My Health My Data Act consent for consumer health data AIGE-OBL-USWA-MHMDA | Separate consent records for collection and sharing; signed sale authorisation; intake flags for derived health data | L1 L2 | ch. 08 → |
| Colorado SB21-169 insurers' use of external consumer data and predictive models AIGE-OBL-USCO-SB21-169 | Inventory of external data sources and models; disparity testing; chief-risk-officer attestation record | L2 L3 L5 | ch. 08 → |
US federal law
7 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| OMB M-25-21 minimum practices for high-impact AI AIGE-OBL-USFED-OMB-M25-21 | Use-case inventory entry; pre-deployment test report; AI impact assessment; monitoring plan; appeal path | L2 L3 L4 L5 | ch. 08 → |
| OMB M-26-04 minimum LLM transparency in federal procurement AIGE-OBL-USFED-OMB-M26-04 | Acceptable use policy; model, system or data cards; end-user resources; feedback channel | L2 L5 | ch. 08 → |
| ECOA Regulation B adverse-action notice with specific principal reasons AIGE-OBL-USFED-REGB-1002-9 | Reason-code service versioned with the model; reason-code fidelity eval; notice template | L3 L4 L5 | ch. 08 → |
| FCRA adverse-action notice with the credit score used AIGE-OBL-USFED-FCRA-1681M | Score and key-factor record per adverse decision; notice template | L4 | ch. 08 → |
| Title VII disparate impact and the UGESP four-fifths rule AIGE-OBL-USFED-TITLE7-703K | Adverse-impact-ratio eval per group with counts and confidence intervals; job-relatedness validation; alternatives search log | L3 L5 | ch. 08 → |
| FTC Act s. 5 substantiation of AI performance claims AIGE-OBL-USFED-FTC-S5 | Claims register linked to current eval runs; substantiation gate on release copy | L1 L3 L5 | ch. 08 → |
| TAKE IT DOWN Act notice and removal of intimate images, incl. digital forgeries AIGE-OBL-USFED-TAKEITDOWN | Takedown pipeline with a 48-hour clock, owner and log; matching of identical copies | L4 L5 | ch. 08 → |
US federal audit and oversight
4 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| GAO AI Accountability Framework principle 1: governance AIGE-OBL-USGAO-GOV | Registry entry with owner and intended purpose; RACI; risk register as code; compliance mapping; published system card | L1 L2 L5 | ch. 08 → |
| GAO AI Accountability Framework principle 2: data AIGE-OBL-USGAO-DATA | Dataset datasheet with lineage; data-quality and representativeness evals; data-flow map | L2 L3 | ch. 08 → |
| GAO AI Accountability Framework principle 3: performance AIGE-OBL-USGAO-PERF | Eval suite with versioned metrics per component and system; bias eval; human-oversight procedure | L3 L4 | ch. 08 → |
| GAO AI Accountability Framework principle 4: monitoring AIGE-OBL-USGAO-MON | Monitoring plan with drift thresholds; monitoring log with corrective actions; periodic review record | L4 L5 | ch. 08 → |
Other jurisdictions
9 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| South Korea AI Basic Act (in force 2026-01-22) AIGE-OBL-KR-AIBASIC | Risk register for high-impact AI; AI-use notification; AI-content labelling | L1 L2 L4 | ch. 08 → |
| Singapore IMDA Model AI Governance Framework for Generative AI (voluntary) AIGE-OBL-SG-GENAI | Eval suite; model cards; content provenance and watermarking | L2 L3 L4 | ch. 08 → |
| UK GDPR Arts. 22A–22D permission-plus-safeguards model for significant, solely automated decisions (Data (Use and Access) Act 2025) AIGE-OBL-UK-ADM | ADM safeguards: meaningful-human-review path, contest and representation channel, decision notice | L4 L2 | ch. 08 → |
| ETSI EN 304 223 baseline cyber-security for AI models and systems AIGE-OBL-ETSI-304223 | AI-system security controls across the lifecycle; supply-chain and AIBOM checks; runtime hardening | L4 | ch. 08 → |
| Singapore IMDA Model AI Governance Framework for Agentic AI: agent identity and scoped authorisations (voluntary) AIGE-OBL-SG-AGENTIC-IDENTITY | Agent registry with a workload identity per agent; delegated, short-lived credentials never broader than the user | L2 L4 | ch. 08 → |
| Singapore IMDA Model AI Governance Framework for Agentic AI: human checkpoints for significant actions (voluntary) AIGE-OBL-SG-AGENTIC-CHECKPOINTS | Checkpoint classes in the tool gateway; approval log; oversight metrics | L4 L5 | ch. 08 → |
| Canada Directive on Automated Decision-Making (federal institutions) AIGE-OBL-CAN-DADM | Published AIA rendered from a shared impact fact base; notice and explanation templates; peer-review record; recourse path; re-assessment triggers as code | L1 L2 L5 | ch. 08 → |
| Brazil LGPD Art. 20 review of automated decisions AIGE-OBL-BR-LGPD-ART20 | Review workflow; statement of criteria and procedures per system | L4 L5 | ch. 08 → |
| UK DMCC Act 2024 banned practices: fake and concealed-incentive reviews AIGE-OBL-UK-DMCC-S225 | Policy blocking review generation; review-provenance log | L1 L4 | ch. 08 → |
South Korea AI Basic Act
7 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| Korea AI Basic Act Art. 31(1) prior notice of high-impact or generative AI AIGE-OBL-KR-ART31-1 | Notice component in UI, terms and contracts; notice inventory per user surface | L2 L4 | ch. 08 → |
| Korea AI Basic Act Art. 31(2)–(3) generative-AI output labels and realistic-content notice AIGE-OBL-KR-ART31-2 | Provenance pipeline: visible label or machine-readable mark plus at least one text or voice notice | L3 L4 | ch. 08 → |
| Korea AI Basic Act Art. 32 safety duties for high-compute systems AIGE-OBL-KR-ART32 | Lifecycle risk register; safety-incident monitoring; results report to MSIT | L3 L4 L5 | ch. 08 → |
| Korea AI Basic Act Art. 33 high-impact self-review and confirmation AIGE-OBL-KR-ART33 | Classification decision record per system: Art. 2(4) area, risk rationale, training-data overview, MSIT reply | L1 L2 | ch. 08 → |
| Korea AI Basic Act Art. 34 measures for high-impact AI AIGE-OBL-KR-ART34 | Risk management, explanation and user-protection plans; named human overseer; published summary; five-year evidence store | L1 L2 L4 L5 | ch. 08 → |
| Korea AI Basic Act Art. 35 fundamental-rights impact assessment (best effort) AIGE-OBL-KR-ART35 | Impact assessment carrying the seven decree elements | L1 L5 | ch. 08 → |
| Korea AI Basic Act Art. 36 domestic representative AIGE-OBL-KR-ART36 | Designation filed with MSIT; evidence-access runbook for the representative | L5 | ch. 08 → |
China
9 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| Provisions on Algorithmic Recommendation (in force 2022-03-01) AIGE-OBL-CN-ALGOREC | Algorithm inventory with filing record and number; security-assessment evidence pack; opt-out control at runtime | L1 L2 L4 | ch. 08 → |
| Provisions on Deep Synthesis (in force 2023-01-10) AIGE-OBL-CN-DEEPSYN | Content-provenance pipeline (visible label plus metadata mark); training-data governance record; consent gate; pre-release security assessment | L2 L3 L4 | ch. 08 → |
| Interim Measures for Generative AI Services (in force 2023-08-15) AIGE-OBL-CN-GENAI | Data-lineage and licensing record; eval gate on generated content; incident pipeline with a retraining loop; filing record | L2 L3 L4 L5 | ch. 08 → |
| Measures for Labelling AI-Generated Synthetic Content with GB 45438-2025 (in force 2025-09-01) AIGE-OBL-CN-LABEL | Provenance and watermarking pipeline emitting the GB 45438 metadata fields; platform-side detection and flagging | L3 L4 | ch. 08 → |
| GB/T 45654-2025 Basic security requirements for generative AI services (voluntary; implemented 2025-11-01) AIGE-OBL-CN-GBT45654 | Corpus-screening record; eval question banks; security-assessment report | L3 L5 | ch. 08 → |
| TC260 AI Safety Governance Framework 3.0: operators' guidelines §5.3 (voluntary; 2026-09-14) AIGE-OBL-CN-TC260-OPS | Risk register keyed to the framework's taxonomy; log-retention policy (six months) with audit; real-time risk monitoring; resilience assessment | L1 L4 L5 | ch. 08 → |
| TC260 Framework 3.0 Appendix 2: agentic AI risk management (voluntary; 2026-09-14) AIGE-OBL-CN-TC260-AGENTS | Agent registry with identity and scope; approval-log store; tool allow-list with integrity checks; runtime guardrails and kill switch; memory-scope policy; decommissioning runbook | L2 L3 L4 L5 | ch. 08 → |
| China PIPL Art. 24 automated decision-making and Arts. 55–56 impact assessment AIGE-OBL-CN-PIPL-ART24 | Explanation service and manual-decision route; non-personalised option at runtime; impact assessment record kept three years | L2 L4 L5 | ch. 08 → |
| Interim Measures for Anthropomorphic Interaction Services (in force 2026-07-15) AIGE-OBL-CN-ANTHRO | Minors'-mode configuration; reminder timer; user-count threshold monitor; security-assessment report; filing record | L1 L2 L4 L5 | ch. 08 → |
Treaty and international soft law
10 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| Council of Europe Convention Art. 14(2)(a)–(b) documentation to contest decisions AIGE-OBL-COE-ART14-2 | Decision record per consequential output; contest path with the record attached | L2 L5 | ch. 08 → |
| Council of Europe Convention Art. 15(2) notice of interaction with an AI system AIGE-OBL-COE-ART15-2 | Interaction disclosure enforced at runtime | L4 | ch. 08 → |
| Council of Europe Convention Art. 16 risk and impact management AIGE-OBL-COE-ART16 | Risk register as code; impact assessment linked to the registry; monitoring against a baseline | L1 L2 L4 | ch. 08 → |
| Council of Europe Convention Art. 16(2)(g) testing before first use and on significant modification AIGE-OBL-COE-ART16-2G | Eval gate on release and on material change | L3 | ch. 08 → |
| OECD AI Principle 1.4(b) override, repair or decommission safely AIGE-OBL-OECD-P1-4B | Tested kill switch; decommissioning record in the registry | L2 L4 | ch. 08 → |
| OECD AI Principle 1.5(b)–(c) traceability and systematic risk management AIGE-OBL-OECD-P1-5 | Evidence store keyed to registry ids; risk register as code; supplier records | L1 L5 | ch. 08 → |
| G7 Hiroshima Code action 1: lifecycle risk management and pre-deployment testing AIGE-OBL-G7-A1 | Adversarial red-team suite; eval gate | L3 | ch. 08 → |
| G7 Hiroshima Code actions 2 and 4: post-deployment monitoring and incident sharing AIGE-OBL-G7-A2-4 | Runtime monitoring; incident pipeline with an external-sharing branch | L4 L5 | ch. 08 → |
| G7 Hiroshima Code action 3: public reporting of capabilities and limitations AIGE-OBL-G7-A3 | Model card published from the registry | L2 | ch. 08 → |
| G7 Hiroshima Code action 7: content authentication and provenance AIGE-OBL-G7-A7 | Provenance marking at output; verification test | L4 | ch. 08 → |
CEN-CENELEC JTC 21
3 rows| Obligation | Artefact | Layers | Chapter |
|---|---|---|---|
| EN 18286:2026 quality management system for EU AI Act purposes AIGE-OBL-CEN-EN18286 | QMS processes run as pipeline stages; design and change-control evidence | L1 L5 | ch. 08 → |
| prEN 18228 AI risk management (draft) AIGE-OBL-CEN-PREN18228 | Provider risk file per system; acceptability criteria as code; control monitoring | L1 L3 L5 | ch. 08 → |
| prEN 18229-1 AI trustworthiness framework, Part 1: logging (draft) AIGE-OBL-CEN-PREN18229-1 | Logging specification per system; structured, signed event logs mapped to the draft | L4 | ch. 08 → |