08. Regulatory map (obligation → artefact → layer)

New to the field? Start with AI governance explained.

Every other chapter maps forward: a capability, then the obligations it touches. This chapter maps backward: an obligation, then the artefact and the layer that answer it. The unit of the map is a row: an obligation, the engineering artefact that produces the evidence for it, and one of the five stack layers (chapter 04): 1 Govern-as-Code · 2 Inventory & Transparency · 3 Evals & Red Teaming as Evidence · 4 Runtime Controls & Observability · 5 Assurance & Continuous Compliance. The map is a crosswalk for finding the artefact that answers a question, not a certificate that the artefact makes you compliant. It indexes the AI-specific instruments first and then the data-protection, cyber-security, liability, consumer and sector law an AI system meets on day one, so that every row of the obligation register resolves to a table in this chapter. Chapter 19 teaches the data-protection rows in its obligation to artefact map, and chapter 20 the other law that already applies to AI (copyright, anti-discrimination, consumer protection and product liability).

How to read this map

Read each row as a sentence: this obligation is answered by this artefact, which lives in this layer. Three cautions apply throughout.

• Mappings are illustrative, not a claim of conformity. No artefact in this book guarantees compliance, and no standard cited here confers a presumption of conformity (see “What is NOT harmonised yet”). An artefact supports and evidences an obligation; the legal judgement of conformity stays with lawyers, notified bodies and authorities.
• Dates are the post-Omnibus dates. Every EU AI Act date below reflects Regulation (EU) 2026/1744, the Digital Omnibus on AI, of 8 July 2026 (OJ L, 24 July 2026), in force 27 July 2026¹²²². Where the Omnibus moved a date, the moved date is shown; where it did not, the row says so.
• The competent authority differs by regime. For general-purpose AI (GPAI) the supervisor is the AI Office, and GPAI fines are formal Commission decisions under Article 101³⁴. For high-risk systems the supervisors are national market-surveillance authorities, whose penalties run under Article 99⁴. Member States choose their own; Spain, for example, set up a dedicated agency, AESIA (Agencia Española de Supervisión de Inteligencia Artificial), whose statute was approved by Royal Decree 729/2023⁵². The map states the authority per row so the reader knows who asks.

EU AI Act, post-Omnibus

For a teaching walk-through of the Act before this reverse index maps it (scope, the risk ladder, value-chain roles, deployer duties, enforcement and the full post-Omnibus timeline), see chapter 18, which walks the Act end to end.

High-risk obligations (Articles 9–15, 17, 25, 26, 27, 49, 71, 72, 73) apply to Annex III systems from 2 December 2027 and to Annex I embedded systems from 2 August 2028, both deferred by the Omnibus from 2 August 2026 and 2 August 2027 respectively¹². The amended Art. 113(c) defers Chapter III, Sections 1 to 3: classification, the requirements of Articles 8 to 15 and the duties of providers, deployers and other operators in Articles 16 to 27²². The high-risk duties that sit elsewhere (conformity assessment, the declaration, CE marking and registration in Articles 43 to 49 and 71, post-market monitoring and incident reporting in Articles 72, 73 and 75(1a), and the right to explanation in Article 86) belong to chapters that formally apply from 2 August 2026, but they have work to do only once a system is classified high-risk, so the map dates them by the classification date. That is this map’s reading; confirm it with counsel (verify)²²⁶³. Articles 26(11), 27, 49, 71 and 86 concern Annex III systems only, so the 2 August 2028 date does not reach them. The GPAI obligations (Articles 53, 55) have applied since 2 August 2025, with Commission enforcement powers live since 2 August 2026³; providers of models placed on the market before 2 August 2025 comply by 2 August 2027 (Art. 111(3))⁶³. Transparency (Article 50) went live on 2 August 2026 and was not moved⁵. Legacy public-authority high-risk systems keep their original 2 August 2030 date².

The Duty holder column names who the obligation binds, which is a different axis from who enforces it. The high-risk design-and-build duties (Articles 9 to 15 and 17) fall on the provider; Articles 26 and 27 fall on the deployer; Articles 4, 5 and 50 bind both; and Articles 53 and 55 bind the GPAI provider. Article 25 sits across the value chain: it sets the conditions under which a distributor, importer or deployer itself becomes a provider and inherits the provider duties. Articles 22 to 24 bind the authorised representative of a non-EU provider, the importer and the distributor, and Article 54 the authorised representative of a non-EU GPAI provider. This matters for the engineer because the artefacts you can produce depend on which hat your organisation wears: a deployer cannot draw up the provider’s technical documentation, but it must run the Article 26 monitoring and the Article 27 FRIA; and when the model is procured, most of the provider-side evidence becomes something you collect rather than produce (see the Vendor / Model Due-Diligence Gate in chapter 05).

Article 16 is the umbrella for the provider’s high-risk duties. Apart from point (l), accessibility, which has its own row, it adds no artefact of its own; it gathers the obligations the rows below break out article by article: the requirements of Articles 9 to 15 and 17, documentation keeping and logs (Articles 18 and 19), corrective action (Article 20), the conformity assessment (Article 43), the EU declaration of conformity and CE marking (Articles 47 and 48), registration (Articles 49 and 71), post-market monitoring (Article 72) and serious-incident reporting (Article 73). So it reads here as a cross-reference, not a row²²⁶³.

Several rows are built out in full elsewhere in the book:

• Art. 4: the literacy programme as code is AI literacy as code in chapter 12, and the role-based training records have a literacy curriculum template.
• Art. 4a and Art. 10: the lawful basis for special-category data for bias detection is in chapter 19; the data needed to test for bias, and where bias enters the lifecycle, are in chapter 16 (protected characteristics, proxies and the data you need to test).
• Art. 6: the Article 6(3) filter and the profiling override are walked through in chapter 18.
• Art. 9: the risk register as code is the risk register as an evidence record in chapter 13; the same artefact answers the ISO/IEC 23894 row below.
• Art. 11, Art. 43 and Art. 47: Annex IV, element by element and conformity, in order are in chapter 14.
• Art. 13: instructions for use as code have a JSON Schema and filled example; what they must say to deployers and affected people is in the EU AI Act, Articles 13 and 86 (chapter 16).
• Art. 25: when a value-chain actor becomes a provider is set out in chapter 18.
• Art. 26: the deployer duties in operation are in chapter 15.
• Art. 27: the FRIA’s cross-reference to a GDPR Art. 35 DPIA is developed in the DPIA for AI systems (chapter 19).
• Art. 72: the monitoring plan has a post-market monitoring plan schema.
• Art. 3(1): the definitional decision record is built field by field in from definition element to registry field (chapter 11).
• Art. 15(4): feedback loops are monitored in monitoring fairness in production (chapter 16), and for agents in memory and context governance (chapter 23); Art. 16(l) has accessible explanations.
• Art. 17(1)(m): the accountability framework is a lifecycle RACI (chapter 12); Art. 43(4) is substantial modification in chapter 14; Arts. 22 to 24 are the EU operator roles in chapter 18.
• Art. 26(2) to 26(11): the deployer’s paragraphs run through chapter 15, from checking the data and the people to records retention and external communications; chapter 17 covers informing the provider and suspending use (Art. 26(5)) and freezing before fixing (Art. 73(6)), and chapter 23 approval design for agent checkpoints.
• Art. 52: the compute ledger and the two-week notification are in systemic risk: threshold, notification, designation (chapter 18); Art. 53(1)(c) is copyright and training data (chapter 20).
• Art. 86: the explanation record is specified in the EU AI Act, Articles 13 and 86 (chapter 16) and explanation and notice to affected people (chapter 18); Art. 87 is met by a channel for raising concerns (chapter 12).

The Omnibus also changed rows the table compresses to one line²². For Art. 6, a safety component must now have the purpose of preventing or mitigating risks to health and safety, or be one whose failure endangers them (Art. 3(14), Art. 6(1a) to 6(1c)); that narrowing applies to the Annex I route from 2 August 2028, and machinery moved from Section A to Section B of Annex I. The Annex VIII, Section B registration record for systems that rely on the Art. 6(3) filter lost the summary of grounds and the list of Member States. For Art. 25, the initial provider’s duty to cooperate now names documentation sufficient to assess compliance, known limitations and failure modes, and targeted technical access for testing (Art. 25(2)); the information and access must be fixed in a written agreement (Art. 25(4)), and breaches of both paragraphs are fined under Art. 99(4)(da).

The Article 73 clock runs by incident class. The provider notifies the market-surveillance authority of the Member State where the incident occurred immediately once it establishes a causal link between the system and the incident, or the reasonable likelihood of one, and in any event within an outer deadline that runs from the moment the provider or, where applicable, the deployer becomes aware of the incident, not from the causal finding⁵⁷. A deployer that identifies a serious incident informs the provider first, then the importer or distributor and the market-surveillance authority; if it cannot reach the provider, Article 73 applies to the deployer mutatis mutandis (Art. 26(5))⁵⁷:

The Omnibus left these deadlines unchanged. For high-risk systems under the AI Office’s exclusive competence, its new Art. 75(1a) sends the report to the AI Office instead, with Art. 73(2) to (9) applying mutatis mutandis, and the AI Office passes it on to the market-surveillance authority of the Member State where the provider is established²². An incident rarely starts only one clock: chapter 17 lays the Article 73 clock beside the GDPR, NIS2, DORA, CRA and GPAI clocks in the overlapping clocks, and the incident record schema holds the timestamps each one needs.

Two cross-cutting provisions frame the penalties. Under Art. 101, the Commission may fine GPAI providers up to 3% of worldwide annual turnover or EUR 15 million, whichever is higher³⁴. Under the Omnibus’s new Art. 75a–75d, the AI Office gains investigation powers, binding commitments and non-compliance decisions over the AI systems for which the amended Art. 75(1) makes it exclusively competent (systems built on a GPAI model by the same provider or undertaking, and systems in designated very large online platforms or search engines), not over GPAI models in general. Periodic penalty payments reach up to 5% of average daily income or worldwide annual turnover per day for a continuing breach (Art. 75c(5)). The Omnibus entered into force on 27 July 2026, and Chapter IX, where these articles sit, applies from 2 August 2026 under Art. 113⁸²². High-risk administrative fines run under Art. 99 (ceilings of 7%, 3% and 1% of turnover depending on the breach), imposed by national authorities⁴. Chapter 18 sets out who supervises what and the AI Office’s direct powers.

GPAI Code of Practice

The GPAI Code of Practice, published 10 July 2025, is the voluntary instrument providers use to demonstrate compliance with the GPAI obligations until harmonised standards exist. It has three chapters⁹¹⁰.

The Code is voluntary; signing it is one route to demonstrating compliance, not a legal presumption of conformity⁹. Signatory status is dynamic and counted per the Commission’s official list¹⁰.

For serious-incident reporting specifically, the Commission published a reporting template on 4 November 2025 for serious incidents involving GPAI models with systemic risk. It aligns to the Article 55 reporting duty and to Commitment 9 of the Code’s Safety and Security chapter, and is the concrete artefact the incident pipeline emits, the same template named in the Article 55 row above²⁶. One internal incident record can feed this template and every other regime’s report. Commitment 9 of the Safety and Security chapter sets the clocks (2, 5, 10 or 15 days by incident class, intermediate reports at least every four weeks, a final report within 60 days of resolution) and a retention of at least five years, and its Appendix 1 lists the capability to operate autonomously and access to tools among the sources of systemic risk and loss of control among the specified risks, which is why the two rows above matter to anyone who deploys agents on a GPAI model⁶⁵. Chapter 17 sets Commitment 9 beside the other regimes in the overlapping clocks, and chapter 23 reads Appendix 1 for agents in EU AI Act hooks for agents.

Data protection and other EU law

An AI system placed on the EU market meets more law than the AI Act on day one. The rows below index the obligations the v0.5.0 chapters teach: the GDPR, the cyber-security and incident-reporting regimes, and the liability, copyright, consumer and sector law that already reaches AI. They answer to their own supervisors (data protection authorities, CSIRTs and financial supervisors, courts and consumer authorities), not to the AI Act’s market-surveillance authorities, and several apply through national transposition, so the “Applies” column says so. Mappings are illustrative, not a claim of conformity.

The GDPR

The GDPR has applied since 25 May 2018⁶⁶. Chapter 19 walks every row below, with the EDPB’s anonymity test for trained models⁶⁷, in privacy and data protection law applied to AI; chapter 16 builds the explanation and contest records of Arts. 15(1)(h) and 22 in data protection: the GDPR and the UK regime.

Cyber-security and incident-reporting law

NIS2 binds essential and important entities through national law, which Member States apply from 18 October 2024⁶⁸; DORA has applied to financial entities since 17 January 2025⁶⁹, with the incident time limits fixed in Delegated Regulation (EU) 2025/301⁷⁰; and the Cyber Resilience Act’s reporting duty applies from 11 September 2026, ahead of the rest of that Regulation on 11 December 2027⁷¹. An AI incident can start several of these clocks at once: chapter 17 lays them side by side in the overlapping clocks, and chapter 15 covers continuity when the provider fails.

Liability, copyright, consumer and sector law

The revised Product Liability Directive treats software, AI systems included, as a product and applies to products placed on the market after 9 December 2026⁷². The text-and-data-mining exception of the DSM Directive yields to a rightholder’s reservation⁷³, which is the rule Art. 53(1)(c) of the AI Act points to. The Digital Services Act, the Unfair Commercial Practices Directive, the Platform Work Directive and the revised Consumer Credit Directive add duties for platforms, traders, digital labour platforms and creditors^74757677. Chapter 20 teaches them: copyright and training data, the EU Product Liability Directive, duty to warn after updates, the EU: UCPD, DSA and the AI Act, employment and credit and lending.

ISO/IEC 42001, 42005 and 42006

ISO/IEC 42001:2023 is the AI management-system (AIMS) standard; its 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 (EN ISO/IEC 42001:2026) confers no presumption of conformity¹¹¹².

ISO/IEC 42005:2025 provides guidance for AI system impact assessment and is the natural companion to Article 27 (FRIA) and Annex A.5¹³. The operating model and RACI of the A.3 row are built in a lifecycle RACI (chapter 12), and chapter 22 places these standards in the ISO/IEC family as a whole.

Two further ISO/IEC standards sit alongside the AIMS. ISO/IEC 42006:2025 sets the requirements for the bodies that audit and certify AI management systems: building on ISO/IEC 17021-1, it is the answer to “who may credibly certify you to 42001”, since it fixes the competence and consistency a certification body must show. ISO/IEC 23894:2023 gives guidance on AI risk management, adapting ISO 31000 to AI; it is the risk-process companion to Article 9 and to the NIST AI RMF²⁷²⁸. The table also carries ISO/IEC 42005:2025, the impact-assessment guidance¹³, and ISO/IEC 22989:2022, the concepts and terminology standard whose AI stakeholder roles and life-cycle vocabulary a registry can reuse field for field⁷⁸.

Maps to: these controls are realised through layers 1, 2, 3 and 5 of the stack; the impact-assessment control (A.5 / ISO 42005) supports EU AI Act Art. 27. Mappings are illustrative, not a claim of conformity.

NIST AI RMF

The NIST AI Risk Management Framework 1.0 (January 2023; there is no 2.0) organises risk work into four functions. It is voluntary and US-origin, and it maps cleanly onto the five-layer stack¹⁴. As of 2026-09-24, NIST’s framework page states that AI RMF 1.0 “is being revised as part of the White House AI Action Plan”; no revised text had been published, so 1.0 remains the version to cite and to pin in control metadata⁵⁹. Chapter 22 covers the NIST AI RMF in depth: the seven trustworthy characteristics, the 19 categories of the Core, the Playbook and the profiles.

Newer NIST AI work

Beyond the RMF, three newer NIST efforts bear on the stack. Two are still in draft, and the copy says so. The AI Agent Standards Initiative, launched by NIST’s Center for AI Standards and Innovation (CAISI) on 17 February 2026, targets interoperable, secure standards for AI agents: identity, authentication, authorisation and agent security²⁹. The draft IR 8596 Cyber AI Profile (initial preliminary draft, 16 December 2025; comments closed 30 January 2026, and still the current version as of 2026-09-24) is a Cybersecurity Framework (CSF 2.0) profile for AI, organised around Secure, Defend and Thwart³⁰. The draft AI 800-1 (Managing Misuse Risk for Dual-Use Foundation Models; second public draft, January 2025; no final version published as of 2026-09-24) is voluntary guidance for identifying, measuring and mitigating misuse risk across the AI lifecycle³¹. The older Generative AI Profile, NIST AI 600-1 (26 July 2024), is the RMF’s companion for generative AI: 12 risks and suggested actions coded to the four functions, and the table carries it too⁷⁹.

CSA AICM and STAR for AI

The Cloud Security Alliance’s AI Controls Matrix (AICM) v1.1, published 22 June 2026, defines 247 control objectives across 18 domains, and the STAR for AI programme provides the assurance scheme around it, in three levels: a Level 1 self-assessment, a Level 1 “Valid-AI-ted” automated validation, and a Level 2 that adds ISO/IEC 42001 certification¹⁵⁸³.

Two lines of work push the AICM toward agents and frontier risk. For agents, the controls sit inside the matrix itself (for example IAM-18 Agent Access Restriction and AIS-11 Agents Security Boundaries)⁸⁰, and the CSA’s agentic control-plane programme adds two published specifications, the Agentic Trust Framework (zero trust for agents with earned autonomy tiers) and AARM (interception of agent actions before they execute)³²⁸¹⁸². Earlier editions of this map listed a proposed “Agentic Control Supplement” to the AICM; it could not be matched to a CSA primary document as of 2026-09-24, so the row now names what is published (verify). For frontier risk, the Catastrophic Risk Annex adds a set of enhanced controls for high-autonomy systems, meant to be proven through pilot audits rather than asserted³³.

OWASP GenAI Security Project

OWASP’s GenAI Security Project supplies the threat vocabulary the controls are built against, plus two formats (the Agent Control Standard and an AIBOM) that the stack consumes directly¹⁶¹⁷⁵⁴. Chapter 23 maps each agentic entry to its controls in threats mapped to controls.

US federal and state laws

The United States has no horizontal federal AI statute; the binding AI-specific rules are state laws, and they differ in scope. Two bind only frontier developers, with the heaviest duties on the large ones; the rest reach ordinary developers, deployers and operators, and federal law that predates AI (fair lending, credit reporting, employment discrimination, the FTC Act) already reaches AI decisions. The tables below carry them all as rows. Chapter 21 teaches them: the federal layer (executive orders, the OMB memoranda that bind agencies and their vendors, and the federal push against state AI laws) and the state laws that bind private organisations, each with its scope, dates, duties and enforcement.

Frontier-developer laws

Two US state laws bind only frontier developers, with the heaviest duties on the large ones, not general deployers: a narrower scope than the EU AI Act’s risk-tiering. They ask large frontier developers to publish safety frameworks and every frontier developer to report critical safety incidents to the state¹⁸¹⁹²⁵. California’s whistleblower protection is met in practice by a channel for raising concerns (chapter 12). The frontier safety frameworks crosswalk maps three published frameworks to the NIST AI RMF and ISO/IEC 42001, as a coverage map rather than a conformity claim.

Other state AI laws

These state laws reach beyond frontier developers to ordinary developers, deployers and operators: Texas and Colorado; California’s training-data, provenance and companion-chatbot laws; New York’s companion rules; Illinois; New York City’s Local Law 144; and Utah^{5584858688899091}. Chapter 21 sets each one out with its enforcement in its table of state laws, and compares the hiring rules in one hiring tool, four regimes.

State privacy and sector laws

State privacy statutes reach AI through profiling opt-outs, assessments, consent and review rights, and one sector law reaches insurers’ models. Chapter 19 compares them in the United States and chapter 20 covers housing, insurance and public services.

Federal law that already reaches AI

Federal agencies are bound by OMB memoranda⁹⁸⁹⁹. For private organisations the federal rules that reach AI are older, technology-neutral law: adverse-action notices in lending and credit reporting, disparate impact in employment, the FTC Act’s ban on deceptive practices, and the TAKE IT DOWN Act’s removal duty^{100101102103104105106}. Chapter 21 covers the federal layer; chapter 16 builds the reason codes of credit adverse-action notices and the four-fifths rule; chapter 20 covers claims substantiation and deepfakes and synthetic media.

Federal audit and oversight

Auditors bring their own lens. The GAO AI Accountability Framework (GAO-21-519SP, 30 June 2021) binds no one, but it is what the US Government Accountability Office, inspectors general and third-party assessors use to audit an AI system, and it holds “other entities” to the same practices as federal agencies¹²⁰. It has four principles (governance, data, performance, monitoring) and 31 key practices, each with questions for the entity and audit procedures for the assessor. An entity that can answer those questions from its own records is ready for the audit. It does not replace the NIST AI RMF (the operational risk process) or a binding law such as the EU AI Act; it is the check applied to both.

Other jurisdictions

The map’s spine is the EU AI Act, but a governance function working across borders answers to more than one regime. These rows are stamped as of 2026-09-24; where a rule is still moving, the copy says so.

South Korea, article by article

The AI Basic Act’s operator duties sit in Articles 31 to 36, with the mechanics in the Enforcement Decree³⁴⁶². All of them have applied since 22 January 2026; what MSIT holds back during its guidance period of at least one year is fact-finding and fines, not the duties⁵⁸. Chapter 21 walks each article in South Korea: the AI Basic Act.

United Kingdom

The UK has no horizontal AI act. It governs AI through existing sector regulators (the ICO, the FCA, the MHRA and others), coordinated centrally, plus the AI Security Institute (renamed from the AI Safety Institute in February 2025) for frontier-model evaluation³⁸. For automated decision-making, the Data (Use and Access) Act 2025 replaced UK GDPR Article 22 with new Articles 22A–22D (in force 5 February 2026): a permission-plus-safeguards model for significant, solely automated decisions, with tighter conditions where special-category data is used. The safeguards (a meaningful-human-review path, a channel to make representations and to contest, and a decision notice) are the artefact the engineer builds³⁷. Chapter 19 sets Articles 22A–22D beside the GDPR and the US regimes in the regimes side by side, and chapter 16 turns the safeguards into explanation and contest records in data protection: the GDPR and the UK regime. UK consumer law reaches AI-generated reviews: the Digital Markets, Competition and Consumers Act 2024 bans fake and concealed-incentive reviews from 6 April 2025¹¹¹, taught in chapter 20 under the United Kingdom: DMCC Act.

China

China governs AI in two tiers, and this map keeps them apart. The binding tier is a set of departmental rules issued by the Cyberspace Administration of China (CAC) with co-issuers, on algorithmic recommendation (2022), deep synthesis (2023), generative AI services (2023) and the labelling of AI-generated synthetic content (2025); several are territorial, and the Interim Measures for Generative AI Services apply only to services offered “to the public within the PRC”⁴³⁴⁴⁴⁵. The labelling duty is backed by a mandatory national standard, GB 45438-2025, that carries the metadata fields the measure requires⁴⁶⁴⁷. The voluntary tier is the recommended standard GB/T 45654-2025 ⁴⁸ and the TC260 AI Safety Governance Framework⁴¹⁴². The Cybersecurity Law, amended by the NPC Standing Committee on 2025-10-28 and in force 2026-01-01, adds a programmatic Article 20 on AI that does not by itself create operator duties⁴⁹; the most recent binding rule, the Interim Measures for Anthropomorphic Interaction Services (in force 2026-07-15), is narrow in scope (services that offer sustained emotional interaction); it has a row below and chapter 21 sets it out under China: what chapter 08 does not already cover⁵⁰. Personal information is governed by the Personal Information Protection Law, whose Article 24 on automated decision-making also has a row¹¹⁰. The framework 3.0 (14 September 2026) is a TC260 technical document published under CAC guidance and described as “a reference for developers, providers and users”: it sets a three-block risk taxonomy (inherent, application and secondary), grades risk qualitatively by scenario, intelligence level and scale with no compute or parameter threshold, treats open-source models as a distinct risk profile and names computing-power security as a risk category, and its Appendix 2 walks the agent lifecycle from design to decommissioning⁴². Practitioner commentary reported agents and physically interactive systems as the headline change in the new version⁵¹.

The instruments above resolve to obligation-to-artefact rows in the same shape as the other jurisdictions:

Appendix 2’s agent controls line up with the two agentic references this chapter already carries, the OWASP Top 10 for Agentic Applications ¹⁶ and the NIST AI Agent Standards Initiative²⁹:

Chapter 23 compares these agent-control sources with the CSA’s and Singapore’s in frameworks written for agents.

Treaty and international soft law

Three international instruments carry duties an engineer can evidence. The Council of Europe Framework Convention (CETS No. 225) binds Parties, not companies, and is not yet in force; inside the EU it is implemented through the AI Act¹¹²¹¹³. The OECD AI Principles, revised on 3 May 2024, and the G7 Hiroshima Code of Conduct of 30 October 2023 are voluntary¹¹⁴¹¹⁵. Chapter 22 teaches them in what the Convention asks for, the five principles and five recommendations and the G7 Hiroshima Process.

Mappings are illustrative, not a claim of conformity.

What is NOT harmonised yet

The map has a hole, and it is important to state it plainly rather than paper over it.

• No harmonised standard is cited in the Official Journal. As of 2026-09-24, Article 40’s presumption of conformity is available to no one, because no harmonised standard has been OJ-cited²⁰.
• EN 18286 is published but not cited. The Article 17 QMS standard EN 18286:2026 was published in July 2026 (the first JTC 21 AI Act standard to reach publication), but it is not yet cited in the Official Journal, so it carries no presumption of conformity²⁰²¹.
• The other JTC 21 drafts are at or before Enquiry. As of 2026-09-24, as reported by a pan-European standards information point, the Enquiry votes on prEN 18228 (Art. 9, risk management) and prEN 18282 (Art. 15, cybersecurity) closed on 30 Jul 2026 and on prEN 18229-1 (Art. 12, logging) on 20 Aug 2026; prEN 18229-3 (Art. 14, human oversight) entered Enquiry on 30 Jul 2026; and prEN 18229-4 and -5 (Art. 15, accuracy and robustness) were approved as new projects on 24 Jun 2026⁶⁰, consistent with the public tracker that had them at Enquiry in mid-2026⁵³. CEN and CENELEC may publish a priority deliverable directly after a positive Enquiry vote and target them for Q4 2026⁶¹; publication would still not be an OJ citation. Chapter 22 tracks every deliverable and the status of the JTC 21 harmonised standards, and explains how the presumption of conformity works.
• ISO/IEC 42001 is not the Article 17 QMS. Certifying to EN ISO/IEC 42001:2026 evidences an AI management system; it does not confer an AI Act presumption of conformity, because it is not a harmonised standard and its scope differs from the Article 17 QMS¹¹¹².
• The Code of Practice is voluntary. Signing the GPAI Code is a way to demonstrate compliance with GPAI obligations; it is not a legal presumption of conformity⁹.
• China’s framework does not cross-reference the Western instruments. The TC260 AI Safety Governance Framework 3.0 cites no ISO/IEC 42001, ISO/IEC 23894, NIST AI RMF or EU AI Act (its named reference points are the Global AI Governance Initiative and UN-centred channels), and it does not name China’s own binding rules either⁴². A crosswalk between the two stacks is something the engineer builds; neither side’s documents supply it.

The register carries the three JTC 21 deliverables the book relies on, with their status as of 2026-09-24²⁰²¹⁶⁰:

The practical reading: for the period this edition covers, you cannot buy a presumption of conformity off the shelf. The obligation-to-artefact rows above are how a governance function evidences the obligation on its own merits while the harmonised standards are still being written.

Maps to: this chapter is the reverse index for the whole book; every EU AI Act article, the GPAI Code of Practice, the GDPR, NIS2, DORA, the CRA, the Product Liability, DSM, Unfair Commercial Practices, Platform Work and Consumer Credit Directives, the DSA, ISO/IEC 42001, 42005, 42006, 23894 and 22989, NIST AI RMF and the newer NIST AI work, CSA AICM, OWASP GenAI/Agentic, the CEN-CENELEC JTC 21 deliverables, the US federal and state laws, the other jurisdictions and the treaty and soft-law instruments named above map onto the five-layer stack (chapter 04) and the pattern catalogue (chapter 05). Mappings are illustrative, not a claim of conformity.

What you can do this week

1. Find your rows. For one system, list the rows of this map that apply to it, by role and risk class, with the date from which each applies.
2. Name one artefact per row. Next to each row, write the artefact that evidences it today, or mark the gap.
3. Put the dates in the pipeline. Store each applicable date as data that your policy checks read, so an obligation that starts to apply shows up as a failing check, not a surprise.
4. Evidence one obligation on its merits. For the obligation you most expected a harmonised standard to cover, write down how you evidence it today without a presumption of conformity.
5. Re-read on change. When a date or a row in this map changes, re-run the first step for your systems and record the difference.

Sources

1. [1] “AI Omnibus enters into force” (Reg. (EU) 2026/1744, in force 2026-07-27; Annex III high-risk → 2 Dec 2027; Annex I → 2 Aug 2028; legacy public-authority → 2 Aug 2030). European Commission. 2026-07-27. https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force (verified: primary)
2. [2] Regulation (EU) 2026/1744 (Digital Omnibus on AI), Art. 1 amendments to Reg. (EU) 2024/1689: new Art. 4a (special-category data for bias detection, pseudonymisation, deletion once bias is corrected); new Art. 5(1)(ba)–(bb) NCII and CSAM bans from 2 Dec 2026; Art. 111(2) public-authority systems by 2 Aug 2030; new Art. 111(4) Art. 50(2) marking by 2 Dec 2026 for systems placed on the market before 2 Aug 2026; Art. 113 dates. Publications Office of the EU (EUR-Lex). 2026-07-24. https://eur-lex.europa.eu/eli/reg/2026/1744/oj/eng (verified: primary)
3. [3] Commission enforcement powers over GPAI providers apply from 2 August 2026; obligations since 2 August 2025 (fines up to 3% of worldwide turnover or EUR 15M under Art. 101). European Commission, AI Act Service Desk. 2026-08-02. https://ai-act-service-desk.ec.europa.eu/en/ai-act/faq/commissions-enforcement-powers-related-ai-act-obligations-providers-most-advanced-models (verified: primary)
4. [4] Regulation (EU) 2024/1689 (AI Act), Art. 101 (Commission fines for GPAI providers: up to 3% or EUR 15M) and Art. 99 (penalties by national authorities: 7% / 3% / 1% ceilings). Publications Office of the EU (EUR-Lex). 2024-07-12. https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng (verified: primary)
5. [5] “Safer and more transparent AI” (Art. 50 transparency live 2 Aug 2026). European Commission. 2026-08-02. https://commission.europa.eu/news-and-media/news/safer-and-more-transparent-ai-2026-08-02_en (verified: primary)
6. [6] “AI literacy, the Digital Omnibus and Article 4 of the AI Act” (Art. 4 reworded to “support the development of” AI literacy; the reworded text applies from 27 Jul 2026). Law & Technology. 2026. https://lawandtechnology.eu/en/ai-literacy-digital-omnibus-article-4-ai-act/ (verified: secondary)
7. [7] Regulation (EU) 2024/1689 (AI Act), Art. 27 (FRIA for deployers of Annex III high-risk; Art. 27(4) cross-reference to a GDPR Art. 35 DPIA). Publications Office of the EU (EUR-Lex). 2024-07-12. https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng (verified: primary)
8. [8] Regulation (EU) 2026/1744 (Digital Omnibus on AI), new Arts. 75a–75d of the AI Act: AI Office investigation powers, binding commitments, non-compliance decisions and periodic penalty payments up to 5% of average daily income or worldwide annual turnover per day (Art. 75c(5)). Publications Office of the EU (EUR-Lex). 2026-07-24. https://eur-lex.europa.eu/eli/reg/2026/1744/oj/eng (verified: primary)
9. [9] The General-Purpose AI Code of Practice (published 10 Jul 2025; a voluntary tool; three chapters: Transparency, Copyright, Safety and Security). European Commission. 2025-07-10. https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai (verified: primary)
10. [10] GPAI Code of Practice: contents and signatories (Safety & Security applies to systemic-risk models; official signatory list). European Commission. 2026. https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai (verified: primary)
11. [11] “ISO/IEC 42001 and the AI Act: why certification is not yet a presumption of conformity” (ISO 42001 AIMS ≠ Art. 17 QMS). Law & Technology. 2026. https://lawandtechnology.eu/en/iso-iec-42001-and-the-ai-act-why-certification-is-not-yet-a-presumption-of-conformity/ (verified: secondary)
12. [12] CSA research note on EU AI Act, prEN 18286 and ISO/IEC 42001 (scope difference; EN ISO/IEC 42001:2026 not a harmonised standard). Cloud Security Alliance. 2026-04-28. https://labs.cloudsecurityalliance.org/research/csa-research-note-eu-ai-act-pren-18286-iso-42001-20260428-cs/ (verified: secondary)
13. [13] ISO/IEC 42005:2025, AI system impact assessment (companion to Art. 27 and ISO 42001 Annex A.5). ISO/IEC. 2025-05. https://www.iso.org/standard/44545.html (verified: secondary)
14. [14] AI Risk Management Framework 1.0 (functions: Govern, Map, Measure, Manage). NIST. 2023-01-26. https://www.nist.gov/itl/ai-risk-management-framework (verified: primary)
15. [15] AI Controls Matrix (AICM) v1.1 (247 control objectives, 18 domains) and STAR for AI. Cloud Security Alliance. 2026-06-22. https://cloudsecurityalliance.org/artifacts/ai-controls-matrix-v1-1 (verified: primary)
16. [16] OWASP Top 10 for Agentic Applications for 2026 (ASI01 Agent Goal Hijack; ASI02 Tool Misuse and Exploitation; ASI03 Identity and Privilege Abuse; ASI04 Agentic Supply Chain Vulnerabilities; ASI05 Unexpected Code Execution (RCE); ASI06 Memory & Context Poisoning; ASI07 Insecure Inter-Agent Communication; ASI08 Cascading Failures; ASI09 Human-Agent Trust Exploitation; ASI10 Rogue Agents; names as in the document’s contents, read 2026-09-24). OWASP GenAI Security Project. 2025-12-09. https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/ (verified: primary)
17. [17] 2026 Top 10 for LLM Applications (released 3 Aug 2026; Excessive Agency is LLM03:2026) and the Agent Control Standard (ACS), donated to the project. OWASP GenAI Security Project. 2026-09-01. https://genai.owasp.org/2026/09/01/owasp-genai-security-project-unveils-2026-top-10-for-llm-applications-new-agent-control-standard-and-sponsors-as-community-tops-30000-members/ (verified: primary)
18. [18] “California’s SB 53: the first frontier AI law explained” (frontier developers: trained a foundation model with more than 10^26 operations; large frontier developers: also more than USD 500M in annual gross revenue; up to USD 1M per violation; AG enforcement). Future of Privacy Forum (Justine Gluck). 2025-10-03. https://fpf.org/blog/californias-sb-53-the-first-frontier-ai-law-explained/ (verified: secondary)
19. [19] “Governor Hochul Signs Nation-Leading Legislation to Require AI Frameworks for AI Frontier Models” (RAISE Act, S6953B/A6453B, signed 19 Dec 2025; agreed chapter amendment; creates an oversight office within the Department of Financial Services; 72-hour incident reporting). Governor Kathy Hochul (New York State). 2025-12-19. https://www.governor.ny.gov/news/governor-hochul-signs-nation-leading-legislation-require-ai-frameworks-ai-frontier-models (verified: primary)
20. [20] Standardisation of the AI Act (no harmonised standard yet referenced in the Official Journal, so no Art. 40 presumption; page last updated 2026-08-03; no Commission implementing decision citing one found in the Publications Office index on 2026-09-24). European Commission. 2026-08-03. https://digital-strategy.ec.europa.eu/en/policies/ai-act-standardisation (verified: primary)
21. [21] “EN 18286 in the Spotlight: Supporting Compliance with the AI Act” (EN 18286:2026, Art. 17 QMS, published; the first standard in support of the AI Act). CEN-CENELEC. 2026-07-31. https://www.cencenelec.eu/news-events/news/2026/en-in-the-spotlight/2026-07-30-ai-quality-management/ (verified: primary)
22. [22] Regulation (EU) 2026/1744 (Digital Omnibus on AI), of 8 July 2026, amending Reg. (EU) 2024/1689 et al.; OJ L, 24 July 2026; in force 27 Jul 2026; amends Art. 3(14), 6(1a)–(1c), 25(2) and (4), 75(1), new 75(1a) (serious incidents of systems under the AI Office’s competence reported to the AI Office), 99(4)(da), Annex I (machinery to Section B) and Annex VIII Section B (points 7 and 9 deleted); Art. 73 not amended; Art. 113(a) Chapters I and II apply from 2 Feb 2025; Art. 113(c) as replaced: Chapter III, Sections 1, 2 and 3, except Art. 6(5), apply from 2 Dec 2027 (Annex III) and 2 Aug 2028 (Annex I). Publications Office of the EU (EUR-Lex). 2026-07-24. https://eur-lex.europa.eu/eli/reg/2026/1744/oj/eng (verified: primary)
23. [23] Regulation (EU) 2024/1689 (AI Act), Art. 25 (responsibilities along the AI value chain; conditions under which a value-chain actor becomes a provider; information flow to downstream actors). Publications Office of the EU (EUR-Lex). 2024-07-12. https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng (verified: primary)
24. [24] NY State Senate Bill 2025-S6953B (RAISE Act; signed 19 Dec 2025 as Chapter 699; frontier model = trained with over 10^26 operations costing over USD 100M; safety protocols and 72-hour incident disclosure; Attorney General and Division of Homeland Security and Emergency Services; superseded in scope by the 2026 chapter amendment, which drops the cost test, adds a USD 500M revenue test and moves oversight to the DFS [25]). New York State Senate. 2025-12-19. https://www.nysenate.gov/legislation/bills/2025/S6953/amendment/B (verified: primary)
25. [25] “New York Finalizes RAISE Act for Frontier AI Models; Law Takes Effect January 1, 2027” (chapter amendment introduced 6 Jan 2026, passed 11 Mar 2026, signed 27 Mar 2026; effective 1 Jan 2027; SB 53’s thresholds: frontier model above 10^26 operations, large frontier developer above USD 500M annual revenue; large frontier developers publish the framework, all frontier developers report critical safety incidents; DFS oversight office). Wiley. 2026-04-03. https://www.wiley.law/alert-New-York-Finalizes-RAISE-Act-for-Frontier-AI-Models-Law-Takes-Effect-January-1-2027 (verified: secondary)
26. [26] “AI Act: Commission publishes a reporting template for serious incidents involving general-purpose AI models with systemic risk” (template for serious-incident reporting under Art. 55; aligned to Commitment 9 of the GPAI Code). European Commission. 2025-11-04. https://digital-strategy.ec.europa.eu/en/library/ai-act-commission-publishes-reporting-template-serious-incidents-involving-general-purpose-ai (verified: primary)
27. [27] ISO/IEC 42006:2025, Requirements for bodies providing audit and certification of AI management systems (builds on ISO/IEC 17021-1; who may credibly certify to 42001). ISO/IEC. 2025. https://www.iso.org/standard/42006 (verified: secondary)
28. [28] ISO/IEC 23894:2023, Guidance on AI risk management (adapts ISO 31000 to AI). ISO/IEC. 2023-02. https://www.iso.org/standard/77304.html (verified: secondary)
29. [29] “Announcing the AI Agent Standards Initiative for Interoperable and Secure Innovation” (CAISI initiative; agent identity, authentication and security). NIST. 2026-02-17. https://www.nist.gov/news-events/news/2026/02/announcing-ai-agent-standards-initiative-interoperable-and-secure (verified: primary)
30. [30] NIST IR 8596 (initial preliminary draft; comments closed 2026-01-30; no later version on CSRC on 2026-09-24): Cybersecurity Framework Profile for Artificial Intelligence (Cyber AI Profile); Secure / Defend / Thwart. NIST. 2025-12-16. https://csrc.nist.gov/pubs/ir/8596/iprd (verified: primary)
31. [31] NIST AI 800-1 (second public draft): Managing Misuse Risk for Dual-Use Foundation Models (voluntary; still in draft, no final version on NIST’s publication server on 2026-09-24). NIST. 2025-01. https://www.nist.gov/news-events/news/2025/01/updated-guidelines-managing-misuse-risk-dual-use-foundation-models (verified: primary)
32. [32] “Securing the Agentic Control Plane: Key Progress at the CSAI Foundation” (names the Agentic Trust Framework, AARM, the Catastrophic Risk Annex and STAR for AI; no “Agentic Control Supplement” is named, checked 2026-09-24). Cloud Security Alliance. 2026-04-29. https://cloudsecurityalliance.org/blog/2026/04/29/securing-the-agentic-control-plane-key-progress-at-the-csai-foundation (verified: primary)
33. [33] AICM Catastrophic Risk Annex: enhanced AICM controls for high-autonomy systems with catastrophic-risk potential. Cloud Security Alliance. 2026-08-05. https://cloudsecurityalliance.org/csai-foundation/catastrophic-risk-annex (verified: primary)
34. [34] Basic Act on the Development of Artificial Intelligence and the Establishment of a Foundation for Trust (인공지능 발전과 신뢰 기반 조성 등에 관한 기본법; Act No. 20676, promulgated 2025-01-21, in force 2026-01-22; high-impact AI duties in Arts. 31 to 36; fact-finding in Art. 40; fines in Art. 43). Korean Law Information Center (MOLEG). 2026-01-22. https://www.law.go.kr/LSW/lsInfoP.do?lsiSeq=268543 (verified: primary)
35. [35] Texas Responsible Artificial Intelligence Governance Act (HB 149), enrolled text; effective 1 Jan 2026. Texas Legislature (89R). 2025. https://capitol.texas.gov/tlodocs/89R/billtext/pdf/HB00149F.pdf (verified: primary)
36. [36] “Colorado AI law in flux: comprehensive replacement bill signed after federal court blocks predecessor’s enforcement” (SB 24-205 delayed to 30 Jun 2026, then replaced by SB 26-189, effective 1 Jan 2027). McDermott Will & Emery. 2026. https://www.mcdermottlaw.com/insights/colorado-ai-law-in-flux-comprehensive-replacement-bill-signed-after-federal-court-blocks-predecessors-enforcement/ (verified: secondary)
37. [37] Data (Use and Access) Act 2025, s. 80 (replaces UK GDPR Art. 22 with Arts. 22A–22D; in force 5 Feb 2026). legislation.gov.uk. 2025. https://www.legislation.gov.uk/ukpga/2025/18/section/80 (verified: primary)
38. [38] “AI Security Institute” (written statement announcing the rename of the AI Safety Institute). UK Parliament. 2025-02-24. https://questions-statements.parliament.uk/written-statements/detail/2025-02-24/hlws454 (verified: primary)
39. [39] Model AI Governance Framework for Generative AI (voluntary). IMDA / AI Verify Foundation. 2024-05. https://aiverifyfoundation.sg/wp-content/uploads/2024/05/Model-AI-Governance-Framework-for-Generative-AI-May-2024-1-1.pdf (verified: primary)
40. [40] ETSI EN 304 223: Securing Artificial Intelligence (SAI); Baseline Cyber Security Requirements for AI Models and Systems (V2.1.1, Dec 2025; 13 principles across five lifecycle stages). ETSI. 2025-12. https://www.etsi.org/newsroom/press-releases/2627-etsi-releases-world-leading-standard-for-securing-ai/ (verified: primary)
41. [41] 《人工智能安全治理框架3.0》发布: announcement of the AI Safety Governance Framework 3.0 (TC260 under CAC guidance; released 2026-09-14 at the 2026 National Cybersecurity Publicity Week). Cyberspace Administration of China. 2026-09-14. https://www.cac.gov.cn/2026-09/14/c_1791137092283345.htm (verified: primary)
42. [42] AI Safety Governance Framework 3.0 (人工智能安全治理框架3.0), bilingual PDF; English text printed pp. 49–130; §2.1.1(b) open-source models p. 55; §2.1.4(a) computing power p. 59; §5.3 operators’ guidelines pp. 101–104; Appendix 2 agentic AI risk management pp. 113–126; no reference to ISO/IEC 42001, NIST AI RMF or the EU AI Act. TC260 / CAC. 2026-09-14. https://www.cac.gov.cn/rootimages/uploadimg/1791137114683961/1791137114683961.pdf (verified: primary)
43. [43] Provisions on the Administration of Algorithmic Recommendation in Internet Information Services (互联网信息服务算法推荐管理规定; CAC, MIIT, MPS and SAMR Order No. 9; promulgated 2021-12-31; in force 2022-03-01; Art. 17 opt-out, Art. 24 algorithm filing, Art. 27 security assessment). Cyberspace Administration of China. 2022-01-04. https://www.cac.gov.cn/2022-01/04/c_1642894606364259.htm (verified: primary)
44. [44] Provisions on the Administration of Deep Synthesis in Internet Information Services (互联网信息服务深度合成管理规定; CAC, MIIT and MPS Order No. 12; promulgated 2022-11-25; in force 2023-01-10; Arts. 14 training data and separate consent, 16–17 marks and labels, 19 filing, 15/20 security assessment). Cyberspace Administration of China. 2022-12-11. https://www.cac.gov.cn/2022-12/11/c_1672221949354811.htm (verified: primary)
45. [45] Interim Measures for the Administration of Generative AI Services (生成式人工智能服务管理暂行办法; CAC and six other bodies, Order No. 15; published 2023-07-13; in force 2023-08-15; Art. 2 scope: services to the public within the PRC; Art. 7 lawful-source data; Art. 12 labelling; Art. 14 stop-remove-retrain-report; Art. 17 security assessment and filing). Cyberspace Administration of China. 2023-07-13. https://www.cac.gov.cn/2023-07/13/c_1690898327029107.htm (verified: primary)
46. [46] Measures for Labelling AI-Generated Synthetic Content (人工智能生成合成内容标识办法; CAC, MIIT, MPS and NRTA; published 2025-03-14; in force 2025-09-01; explicit and implicit labels; platform verification duty). Cyberspace Administration of China. 2025-03-14. https://www.cac.gov.cn/2025-03/14/c_1743654684782215.htm (verified: primary)
47. [47] GB 45438-2025 Cybersecurity technology: Labeling method for content generated by artificial intelligence (网络安全技术 人工智能生成合成内容标识方法; mandatory national standard; issued 2025-02-28; implemented 2025-09-01). SAMR / SAC (drafted by TC260). 2025-02-28. https://std.samr.gov.cn/gb/search/gbDetailed?id=301E0388CB75788DE06397BE0A0AE1B4 (verified: primary)
48. [48] GB/T 45654-2025 Cybersecurity technology: Basic security requirements for generative artificial intelligence service (网络安全技术 生成式人工智能服务安全基本要求; recommended national standard; issued 2025-04-25; implemented 2025-11-01). SAMR / SAC (drafted by TC260). 2025-04-25. https://openstd.samr.gov.cn/bzgk/std/newGbInfo?hcno=F67D3F376E0A0A0FF5317FB36B32A30A (verified: primary)
49. [49] Cybersecurity Law of the PRC as amended by the NPC Standing Committee decision of 2025-10-28 (in force 2026-01-01; new Article 20 on AI: state support for AI research, training-data and computing infrastructure, AI ethics norms, risk monitoring, assessment and safety supervision). Cyberspace Administration of China (consolidated text). 2025-12-29. https://www.cac.gov.cn/2025-12/29/c_1768735112911946.htm (verified: primary)
50. [50] Interim Measures for the Administration of Anthropomorphic Interaction Services (人工智能拟人化互动服务管理暂行办法; CAC, NDRC, MIIT, MPS and SAMR; published 2026-04-10; in force 2026-07-15). Cyberspace Administration of China. 2026-04-10. https://www.cac.gov.cn/2026-04/10/c_1777558395078289.htm (verified: primary)
51. [51] “China’s TC260 released Version 3.0 of the AI Safety Governance Framework” (LinkedIn post; agents and physically interactive systems as the headline change). Barbara Li (Reed Smith). 2026-09. https://www.linkedin.com/posts/barbara-li-67532067_tc260-ai-governance-share-7505863215600308224-XIyo/ (verified: reported)
52. [52] Real Decreto 729/2023, de 22 de agosto, por el que se aprueba el Estatuto de la Agencia Española de Supervisión de Inteligencia Artificial (Royal Decree approving the AESIA statute; seat in A Coruña; BOE no. 210, 2 Sep 2023). Boletín Oficial del Estado. 2023-09-02. https://www.boe.es/eli/es/rd/2023/08/22/729 (verified: primary)
53. [53] JTC 21 standards tracker (risk, logging and cybersecurity standards at Enquiry, end-2026 target; a vendor-maintained tracker, not a CEN-CENELEC publication). kla.digital. 2026. https://kla.digital/blog/jtc-21-standards-tracker (verified: reported)
54. [54] “Evolving AI Transparency: the AIBOM generator’s new home at OWASP” (CycloneDX output). OWASP GenAI Security Project. 2025-12-18. https://genai.owasp.org/2025/12/18/evolving-ai-transparency-the-journey-of-the-aibom-generator-and-its-new-home-at-owasp/ (verified: primary)
55. [55] SB26-189 Automated Decision-Making Technology (signed by the Governor 14 May 2026; Session Laws chapter 131). Colorado General Assembly. 2026-05-14. https://leg.colorado.gov/bills/sb26-189 (verified: primary)
56. [56] SB 53, Artificial intelligence models: large developers (Transparency in Frontier Artificial Intelligence Act; approved by the Governor 29 Sep 2025; frontier AI framework; critical safety incidents to the Office of Emergency Services within 15 days; civil penalty up to USD 1M per violation, Attorney General only). California Legislature. 2025-09-29. https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260SB53 (verified: primary)
57. [57] Regulation (EU) 2024/1689 (AI Act): Art. 26(5) (a deployer that identifies a serious incident informs the provider first, then the importer or distributor and the market-surveillance authority; Art. 73 applies mutatis mutandis if it cannot reach the provider) and Art. 73(1)–(4) (report immediately on a causal link or its reasonable likelihood, and no later than 15, 2 or 10 days after the provider or, where applicable, the deployer becomes aware). Publications Office of the EU (EUR-Lex). 2024-06-13. https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng (verified: primary)
58. [58] “AI Basic Act Update: Enforcement and Key Implications” (MSIT guidance period of at least one year for fact-finding and fines; exceptions for loss of life or human-rights violations). Shin & Kim. 2026-02-11. https://www.shinkim.com/eng/media/newsletter/3117 (verified: secondary)
59. [59] AI Risk Management Framework (“The AI RMF 1.0 is being revised as part of the White House AI Action Plan”; no revised version published as of 2026-09-24). NIST. 2026-09-24. https://www.nist.gov/itl/ai-risk-management-framework (verified: primary)
60. [60] Project stages for JTC 21 deliverables read on 2026-09-24 (prEN 18228 and prEN 18282 Enquiry votes closed 2026-07-30; prEN 18229-1 closed 2026-08-20; prEN 18229-3 at Enquiry from 2026-07-30; prEN 18229-4 and -5 new projects 2026-06-24). Genorma (pan-European standards information point with national standards bodies). 2026-09-24. https://genorma.com/en/standards/pren-18228 (verified: secondary)
61. [61] “Update on CEN and CENELEC’s decision to accelerate the development of standards for artificial intelligence” (direct publication after a positive Enquiry vote; Q4 2026 target). CEN-CENELEC. 2025-10-23. https://www.cencenelec.eu/news-events/news/2025/brief-news/2025-10-23-ai-standardization/ (verified: primary)
62. [62] Enforcement Decree of the AI Basic Act (Presidential Decree No. 36053, promulgated 2026-01-21, in force 2026-01-22). Korean Law Information Center (MOLEG). 2026-01-21. https://www.law.go.kr/LSW/lsInfoP.do?efYd=20260122&lsiSeq=282879 (verified: primary)
63. [63] Regulation (EU) 2024/1689 (AI Act), consolidated text of 27 July 2026 incorporating Regulation (EU) 2026/1744 (Art. 3(1) AI system; Art. 6(3)–(4) documented non-high-risk assessment and Art. 49(2) registration; Art. 15(3)–(4) declared accuracy metrics and feedback loops; Art. 16(l) accessibility; Art. 17(1)(m) accountability framework; Art. 18 documentation kept 10 years; Art. 19 logs kept at least six months; Art. 20 corrective actions; Arts. 22–24 authorised representatives, importers and distributors; Art. 26(1)–(11) deployer duties; Art. 43(4) new assessment on substantial modification; Art. 48 CE marking; Art. 52 notification within two weeks; Art. 53(1)(c) copyright policy; Art. 54 authorised representative of GPAI providers; Art. 73(6) investigation without altering the system; Art. 86 right to explanation; Art. 87 Directive (EU) 2019/1937 applies; Art. 111(3) GPAI models placed on the market before 2 August 2025 comply by 2 August 2027; Art. 113). Publications Office of the EU (EUR-Lex). 2026-07-27. https://eur-lex.europa.eu/eli/reg/2024/1689/2026-07-27/eng (verified: primary)
64. [64] Directive (EU) 2019/1937 on the protection of persons who report breaches of Union law (internal reporting channels for private legal entities with 50 or more workers, Art. 8(3); acknowledgment within seven days and feedback within three months, Art. 9(1)). Publications Office of the EU (EUR-Lex). 2019-11-26. https://eur-lex.europa.eu/eli/dir/2019/1937/oj/eng (verified: primary)
65. [65] General-Purpose AI Code of Practice, Safety and Security chapter (Commitment 9, Measure 9.3: 2, 5, 10 and 15 days by incident class, intermediate reports at least every four weeks, final report within 60 days of resolution; Measure 9.4: records kept at least five years; Appendix 1.3 sources of systemic risk incl. the capability to operate autonomously, colluding with other AI systems, access to tools and the level of human oversight; Appendix 1.4 specified systemic risks incl. loss of control). European Commission. 2025-07-10. https://ec.europa.eu/newsroom/dae/redirection/document/118119 (verified: primary)
66. [66] Regulation (EU) 2016/679 (General Data Protection Regulation; Arts. 5(1)(b)–(c), 5(2), 6, 6(4), 7, 9, 13, 14, 15(1)(h), 16, 17, 21, 22, 25, 28, 30, 33–36, 44–49; applies from 25 May 2018, Art. 99(2)). Publications Office of the EU (EUR-Lex). 2016-05-04. https://eur-lex.europa.eu/eli/reg/2016/679/oj/eng (verified: primary)
67. [67] Opinion 28/2024 on certain data protection aspects related to the processing of personal data in the context of AI models (legitimate-interest test; anonymity test and the evidence expected for a claim that a model is anonymous). European Data Protection Board. 2024-12-17. https://www.edpb.europa.eu/system/files/2024-12/edpb_opinion_202428_ai-models_en.pdf (verified: primary)
68. [68] Directive (EU) 2022/2555 (NIS2) (Art. 21(2)(c) business continuity and (d) supply-chain security; Art. 23(4) early warning within 24 hours, incident notification within 72 hours, final report within one month; Art. 41(1) measures applied from 18 October 2024). Publications Office of the EU (EUR-Lex). 2022-12-27. https://eur-lex.europa.eu/eli/dir/2022/2555/oj/eng (verified: primary)
69. [69] Regulation (EU) 2022/2554 (DORA) (Art. 19 reporting of major ICT-related incidents; Art. 28(3) register of information; Art. 28(8) exit strategies; Art. 64 applies from 17 January 2025). Publications Office of the EU (EUR-Lex). 2022-12-27. https://eur-lex.europa.eu/eli/reg/2022/2554/oj/eng (verified: primary)
70. [70] Commission Delegated Regulation (EU) 2025/301, RTS on the content and time limits of major ICT-related incident reports (Art. 5(1): initial notification within four hours of classification and no later than 24 hours from awareness; Art. 5(2): within four hours of a classification made after those 24 hours; intermediate report within 72 hours of the initial notification; final report no later than one month after the latest intermediate report). Publications Office of the EU (EUR-Lex). 2025-02-20. https://eur-lex.europa.eu/eli/reg_del/2025/301/oj/eng (verified: primary)
71. [71] Regulation (EU) 2024/2847 (Cyber Resilience Act) (Art. 14 reporting of actively exploited vulnerabilities and severe incidents: early warning within 24 hours, notification within 72 hours, final reports; Art. 71(2) applies from 11 December 2027, Art. 14 from 11 September 2026). Publications Office of the EU (EUR-Lex). 2024-11-20. https://eur-lex.europa.eu/eli/reg/2024/2847/oj/eng (verified: primary)
72. [72] Directive (EU) 2024/2853 on liability for defective products (Art. 2(1) products placed on the market or put into service after 9 December 2026; Art. 4(1) software is a product; Art. 9 disclosure of evidence; Art. 10 presumption of defectiveness; Art. 11(2) no exemption for defects due to software, its updates or the lack of safety updates within the manufacturer’s control; Art. 22 transposition by 9 December 2026). Publications Office of the EU (EUR-Lex). 2024-11-18. https://eur-lex.europa.eu/eli/dir/2024/2853/oj/eng (verified: primary)
73. [73] Directive (EU) 2019/790 on copyright and related rights in the Digital Single Market (Art. 4(3) text-and-data-mining exception subject to an express reservation, by machine-readable means for content made publicly available online; Art. 29 transposition by 7 June 2021). Publications Office of the EU (EUR-Lex). 2019-05-17. https://eur-lex.europa.eu/eli/dir/2019/790/oj/eng (verified: primary)
74. [74] Regulation (EU) 2022/2065 (Digital Services Act) (Art. 25 online interface design and organisation; Art. 27 recommender system transparency; applies from 17 February 2024). Publications Office of the EU (EUR-Lex). 2022-10-27. https://eur-lex.europa.eu/eli/reg/2022/2065/oj/eng (verified: primary)
75. [75] Directive 2005/29/EC (Unfair Commercial Practices Directive) (Art. 5 general prohibition; Arts. 6–7 misleading actions and omissions; Art. 19 measures applied by 12 December 2007), with Directive (EU) 2019/2161 adding Annex I points 23b and 23c on consumer reviews. Publications Office of the EU (EUR-Lex). 2005-06-11. https://eur-lex.europa.eu/eli/dir/2005/29/oj/eng (verified: primary)
76. [76] Directive (EU) 2024/2831 on improving working conditions in platform work (Art. 7 limits on processing by automated systems; Art. 9 transparency; Art. 10 human oversight and an impact evaluation at least every two years; Art. 11 explanation and human review; Art. 29 transposition by 2 December 2026). Publications Office of the EU (EUR-Lex). 2024-11-11. https://eur-lex.europa.eu/eli/dir/2024/2831/oj/eng (verified: primary)
77. [77] Directive (EU) 2023/2225 on credit agreements for consumers (Art. 18(8) human intervention, explanation and review where the creditworthiness assessment involves automated processing; Art. 48 measures applied from 20 November 2026). Publications Office of the EU (EUR-Lex). 2023-10-30. https://eur-lex.europa.eu/eli/dir/2023/2225/oj/eng (verified: primary)
78. [78] ISO/IEC 22989:2022, Information technology: Artificial intelligence: AI concepts and terminology (AI stakeholder roles; AI system life cycle). ISO/IEC. 2022-07. https://www.iso.org/standard/74296.html (verified: primary)
79. [79] NIST AI 600-1, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (12 risks; suggested actions coded GV, MP, MS and MG). NIST. 2024-07-26. https://doi.org/10.6028/NIST.AI.600-1 (verified: primary)
80. [80] AICM v1.1.1 and AI-CAIQ machine-readable bundle (JSON, YAML, OSCAL; control ids and titles incl. IAM-18 Agent Access Restriction and AIS-11 Agents Security Boundaries). Cloud Security Alliance. 2026-08-04. https://cloudsecurityalliance.org/artifacts/aicm-machine-readable-bundle-json-yaml-oscal (verified: primary)
81. [81] Agentic Trust Framework, v1 (zero-trust governance for AI agents; autonomy tiers and promotion criteria; CC BY 4.0). CSAI Foundation / Cloud Security Alliance. 2026-02. https://agentictrustframework.ai/ (verified: primary)
82. [82] Autonomous Action Runtime Management (AARM) specification (pre-execution interception with identity binding; policy evaluation before execution). Cloud Security Alliance. 2026. https://aarm.dev/ (verified: primary)
83. [83] STAR for AI (Level 1 self-assessment; Level 1 Valid-AI-ted automated validation; Level 2 with ISO/IEC 42001 certification plus the Valid-AI-ted assessment; read 2026-09-24). Cloud Security Alliance. 2026-09-24. https://cloudsecurityalliance.org/star/ai (verified: primary)
84. [84] AB-2013, Generative artificial intelligence: training data transparency (chaptered 2024-09-28, Chapter 817; documentation on or before 2026-01-01 for systems released since 2022-01-01). California Legislative Information. 2024-09-28. https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240AB2013 (verified: primary)
85. [85] AB-853, California AI Transparency Act (amends SB 942; chaptered 2025-10-13, Chapter 674; operative 2026-08-02; platform duties 2027-01-01; capture devices 2028-01-01). California Legislative Information. 2025-10-13. https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202520260AB853 (verified: primary)
86. [86] SB-243, Companion chatbots (chaptered 2025-10-13, Chapter 677; AI disclosure; reminders at least every three hours for known minors; suicide and self-harm protocol; annual reports to the Office of Suicide Prevention beginning 2027-07-01). California Legislative Information. 2025-10-13. https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202520260SB243 (verified: primary)
87. [87] “California Finalizes Regulations to Strengthen Consumers’ Privacy” (regulations on ADMT, risk assessments and cybersecurity audits approved 2025-09-23; effective 2026-01-01; ADMT requirements from 2027-01-01; risk-assessment attestations and summaries due 2028-04-01). California Privacy Protection Agency. 2025-09-23. https://cppa.ca.gov/announcements/2025/20250923.html (verified: primary)
88. [88] New York General Business Law Article 47, Artificial Intelligence Companion Models (§§ 1700–1704; self-harm protocol; notice at the start and at least every three hours; Attorney General, up to USD 15,000 per day; most recent revision shown 2025-11-07). New York State Senate. 2025-11-07. https://www.nysenate.gov/legislation/laws/GBS/A47 (verified: primary)
89. [89] “Illinois Adopts New AI-in-Employment Regulations: What Employers Need to Know for 2026” (HB 3773 effective 2026-01-01; notice duty; ZIP codes as a proxy; IDHR draft rules). Hinshaw & Culbertson. 2026. https://www.hinshawlaw.com/en/insights/blogs/employment-law-observer/illinois-adopts-new-ai-in-employment-regulations-what-employers-need-to-know-for-2026 (verified: secondary)
90. [90] Automated Employment Decision Tools (Local Law 144 of 2021; bias audit within one year; public summary; notices; enforcement from 2023-07-05). NYC Department of Consumer and Worker Protection. 2023. https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page (verified: primary)
91. [91] S.B. 226, Artificial Intelligence Consumer Protection Amendments, enrolled copy (Utah Code 13-75, effective 2025-05-07: disclosure on a clear and unambiguous request; prominent disclosure in high-risk interactions by regulated occupations; safe harbour; Title 13, Chapter 72, Artificial Intelligence Policy Act, repealed 2027-07-01). Utah State Legislature. 2025. https://le.utah.gov/Session/2025/bills/enrolled/SB0226.pdf (verified: primary)
92. [92] SB21-169, Restrict Insurers’ Use of External Consumer Data (signed 2021-07-06; effective 2021-09-07; risk-management framework, testing and chief-risk-officer attestation; rules per type of insurance, none effective before 2023-01-01). Colorado General Assembly. 2021-07-06. https://leg.colorado.gov/bills/sb21-169 (verified: primary)
93. [93] Code of Virginia § 59.1-580, Data protection assessments (targeted advertising, sale, profiling with a reasonably foreseeable risk, sensitive data; processing activities created after 2023-01-01; available to the Attorney General). Virginia General Assembly. 2023. https://law.lis.virginia.gov/vacode/title59.1/chapter53/section59.1-580/ (verified: primary)
94. [94] SB21-190, Protect Personal Data Privacy (Colorado Privacy Act; opt-outs incl. profiling; data protection assessments; universal opt-out mechanism; effective 1 July 2023). Colorado General Assembly. 2021-07-07. https://leg.colorado.gov/bills/sb21-190 (verified: primary)
95. [95] Minnesota Statutes § 325M.14, Consumer personal data rights, subd. 1(g) (question the result of profiling, be informed of the reason, review and correct the data, have the decision re-evaluated; effective 31 July 2025). Office of the Revisor of Statutes, Minnesota. 2025. https://www.revisor.mn.gov/statutes/cite/325M.14 (verified: primary)
96. [96] “Biometric Information Privacy Act” (signed 3 October 2008; consent, timely destruction and secure storage of biometric identifiers; USD 1,000 or 5,000 per violation). Wikipedia. 2026-09-24. https://en.wikipedia.org/wiki/Biometric_Information_Privacy_Act (verified: reported)
97. [97] Chapter 19.373 RCW, Washington My Health My Data Act (consumer health data incl. data derived or extrapolated by algorithms or machine learning; consent, separate sharing consent, signed authorisation for sale; 31 March 2024, small businesses 30 June 2024). Washington State Legislature. 2023. https://app.leg.wa.gov/RCW/default.aspx?cite=19.373&full=true (verified: primary)
98. [98] OMB Memorandum M-25-21, Accelerating Federal Use of AI through Innovation, Governance, and Public Trust (high-impact AI; minimum practices; 365 days to document). Office of Management and Budget. 2025-04-03. https://www.whitehouse.gov/wp-content/uploads/2025/02/M-25-21-Accelerating-Federal-Use-of-AI-through-Innovation-Governance-and-Public-Trust.pdf (verified: primary)
99. [99] OMB Memorandum M-26-04, Increasing Public Trust in Artificial Intelligence Through Unbiased AI Principles (policies updated by 2026-03-11; minimum LLM transparency in solicitations). Office of Management and Budget. 2025-12-11. https://www.whitehouse.gov/wp-content/uploads/2025/12/M-26-04-Increasing-Public-Trust-in-Artificial-Intelligence-Through-Unbiased-AI-Principles-1.pdf (verified: primary)
100. [100] 12 CFR § 1002.9, Notifications (statement of specific reasons, or the right to one within 30 days; internal standards or a failed score are insufficient; source 76 FR 79445, 21 December 2011, as amended 20 March 2023). Legal Information Institute, Cornell Law School. 2026-09-24. https://www.law.cornell.edu/cfr/text/12/1002.9 (verified: secondary)
101. [101] 15 U.S.C. § 1681m, Requirements on users of consumer reports (adverse-action notice; numerical credit score and key factors added by Pub. L. 111-203, s. 1100F, effective on the designated transfer date, 21 July 2011 per 12 U.S.C. § 5582 note). Legal Information Institute, Cornell Law School. 2026-09-24. https://www.law.cornell.edu/uscode/text/15/1681m (verified: secondary)
102. [102] 42 U.S.C. § 2000e-2(k), Burden of proof in disparate impact cases (business necessity; alternative employment practice; added by the Civil Rights Act of 1991, 21 November 1991). Legal Information Institute, Cornell Law School. 2026-09-24. https://www.law.cornell.edu/uscode/text/42/2000e-2 (verified: secondary)
103. [103] 29 CFR § 1607.4(D), Uniform Guidelines on Employee Selection Procedures: adverse impact and the four-fifths rule (43 FR 38295, 25 August 1978). Legal Information Institute, Cornell Law School. 2026-09-24. https://www.law.cornell.edu/cfr/text/29/1607.4 (verified: secondary)
104. [104] 15 U.S.C. § 45(a)(1), Unfair methods of competition and unfair or deceptive acts or practices unlawful (deceptive-practices prong added by the Wheeler-Lea Act of 21 March 1938). Legal Information Institute, Cornell Law School. 2026-09-24. https://www.law.cornell.edu/uscode/text/15/45 (verified: secondary)
105. [105] “FTC Order Requires Workado to Back Up Artificial Intelligence Detection Claims” (competent and reliable evidence required for AI accuracy claims). Federal Trade Commission. 2025-04-28. https://www.ftc.gov/news-events/news/press-releases/2025/04/ftc-order-requires-workado-back-artificial-intelligence-detection-claims (verified: primary)
106. [106] TAKE IT DOWN Act, Public Law 119-12 (enacted 19 May 2025; s. 3: covered platforms establish a notice-and-removal process within one year of enactment and remove reported images, and known identical copies, within 48 hours; enforced by the FTC). US Government Publishing Office, GovInfo. 2025-05-19. https://www.govinfo.gov/content/pkg/PLAW-119publ12/html/PLAW-119publ12.htm (verified: primary)
107. [107] Model AI Governance Framework for Agentic AI, version 1.5 (published 2026-05-20, updated 2026-06-05; agent identity unique, accounted for and centrally managed; authorisations scoped, time- or session-bound, non-transferable and bounded by the authorising human; significant checkpoints for high-stakes, irreversible, outlier and user-defined actions; approvals enforced through system-level controls). IMDA. 2026-06-05. https://www.imda.gov.sg/-/media/imda/files/about/emerging-tech-and-research/artificial-intelligence/mgf-for-agentic-ai.pdf (verified: primary)
108. [108] Directive on Automated Decision-Making (in effect 1 April 2019; systems procured before 24 June 2025 comply by 24 June 2026; 6.1 algorithmic impact assessment published before production; Appendix C requirements by impact level; notice, explanation, peer review, recourse, reporting). Treasury Board of Canada Secretariat. 2025-06-24. https://www.tbs-sct.canada.ca/pol/doc-eng.aspx?id=32592 (verified: primary)
109. [109] Lei Geral de Proteção de Dados Pessoais, Lei n. 13.709/2018, compiled text (Art. 20 review of decisions taken solely on automated processing; Art. 65 entry into force, incl. Arts. 52 to 54 from 1 August 2021). Presidência da República (Brazil). 2026-09-24. https://www.planalto.gov.br/ccivil_03/_ato2015-2018/2018/lei/l13709compilado.htm (verified: primary)
110. [110] Personal Information Protection Law of the People’s Republic of China, English translation for reference (Art. 24 automated decision-making; Arts. 55–56 personal information protection impact assessment kept at least three years; in force 1 November 2021). National People’s Congress. 2021-12-29. http://en.npc.gov.cn.cdurl.cn/2021-12/29/c_694559.htm (verified: primary)
111. [111] Digital Markets, Competition and Consumers Act 2024, s. 225 (unfair commercial practices prohibited; in force 6 April 2025) and Sch. 20, para. 13 (fake and concealed-incentive consumer reviews). legislation.gov.uk. 2024. https://www.legislation.gov.uk/ukpga/2024/13/section/225 (verified: primary)
112. [112] Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law (CETS No. 225), text (Arts. 14–15 remedies and safeguards; Art. 16 risk and impact management; Art. 30 entry into force). Council of Europe. 2024-09-05. https://rm.coe.int/1680afae3c (verified: primary)
113. [113] The Framework Convention on Artificial Intelligence (Parties: the European Union; not yet in force; read on 2026-09-24). Council of Europe. 2026-09-24. https://www.coe.int/en/web/artificial-intelligence/the-framework-convention-on-artificial-intelligence (verified: primary)
114. [114] Recommendation of the Council on Artificial Intelligence, OECD/LEGAL/0449 (principles 1.4(b) override, repair or decommission safely and 1.5(b)–(c) traceability and systematic risk management; revised 3 May 2024). OECD. 2024-05-03. https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0449 (verified: primary)
115. [115] Hiroshima Process International Code of Conduct for Organizations Developing Advanced AI Systems (11 actions; voluntary). G7 / European Commission. 2023-10-30. https://digital-strategy.ec.europa.eu/en/library/hiroshima-process-international-code-conduct-advanced-ai-systems (verified: primary)
116. [116] “OECD launches global framework to monitor application of G7 Hiroshima AI Code of Conduct” (reporting framework launched 7 February 2025; first reports by 15 April 2025). OECD. 2025-02-07. https://www.oecd.org/en/about/news/press-releases/2025/02/oecd-launches-global-framework-to-monitor-application-of-g7-hiroshima-ai-code-of-conduct.html (verified: primary)
117. [118] Public Act 103-0804, HB 3773 (amends the Illinois Human Rights Act, 775 ILCS 5/2-102(L): no AI with a discriminatory effect on protected classes, no ZIP codes as a proxy, notice of AI use; IDHR to adopt rules; approved 9 Aug 2024, effective 1 Jan 2026; text and bill status read from Web Archive captures of 2025-03-29 and 2025-06-17, as ilga.gov refused connections on 2026-09-25). Illinois General Assembly. 2024-08-09. https://www.ilga.gov/legislation/publicacts/fulltext.asp?Name=103-0804 (verified: primary)
118. [119] 740 ILCS 14, Biometric Information Privacy Act (Source: P.A. 95-994, eff. 10-3-08; s. 15 retention schedule, written release, secure storage; s. 20 right of action, USD 1,000 negligent or USD 5,000 intentional or reckless per violation; text read from the Web Archive capture of 2025-06-18, as ilga.gov refused connections on 2026-09-25). Illinois General Assembly. 2008-10-03. https://www.ilga.gov/legislation/ilcs/ilcs3.asp?ActID=3004&ChapterID=57 (verified: primary)
119. [120] Artificial Intelligence: An Accountability Framework for Federal Agencies and Other Entities, GAO-21-519SP (four principles: governance, data, performance, monitoring; 31 key practices, each with questions for entities and procedures for auditors and third-party assessors). U.S. Government Accountability Office. 2021-06-30. https://www.gao.gov/products/gao-21-519sp (verified: primary)