What the page covers 22 editions of 7 developers’ safety frameworks, and 60 laws and government texts that address developers of frontier or general-purpose AI models, dated 2023 to 2026. The texts come from the European Union, China, the US federal government, four US states, the United Kingdom, the AI summits, the G7 and the OECD. The selection is not comprehensive. What is left out is listed under “Not covered”.
The EU AI Act sets duties for providers of general-purpose AI models, with further duties for models with systemic risk. China’s Interim Measures and Labelling Measures set duties for providers of generative AI services.
Records: EU AI Act (2024) ↗ · EU AI Act (2024): required action ↗ · EU AI Act (2024): required action ↗ · China generative AI interim measures (2023) ↗ · China generative AI interim measures (2023): required action ↗ · China AI-generated content labelling measures (2025): required action ↗
California SB 53 requires a large frontier developer to write, implement, comply with and publish a frontier AI framework. New York’s RAISE Act of 2025 requires a written safety and security protocol, and its 2026 amendment requires a frontier AI framework from 1 January 2027.
Records: California SB 53: required action ↗ · NY RAISE Act, Chapter 699 (2025): required action ↗ · New York RAISE Act, Chapter 96 (2026): required action ↗ · New York RAISE Act, Chapter 96 (2026): threshold or trigger ↗
Under the AI Act, providers may rely on a code of practice to demonstrate compliance until a harmonised standard is published. The Commission published the final General-Purpose AI Code of Practice on 10 July 2025 and assessed it under Article 56 of the Act in an opinion of 1 August 2025.
Records: EU AI Act (2024): required action ↗ · EU AI Act (2024): required action ↗ · EU GPAI Code of Practice ↗ · Commission Opinion on the GPAI Code (2025) ↗
At the US federal level, Executive Order 14148 revoked Executive Order 14110 in January 2025, and the Department of Commerce announced a rescission of the AI diffusion rule in May 2025. Executive Order 14365 directs the Attorney General to set up a task force to challenge state AI laws inconsistent with its policy. The White House’s legislative recommendations of March 2026 state that Congress should preempt state AI laws that impose undue burdens.
Records: EO 14110 (2023) ↗ · BIS AI diffusion rule (2025) ↗ · EO 14365 (2025): who decides ↗ · US AI legislative recommendations (2026): required action ↗
The UK government’s memoranda with Anthropic, Cohere and OpenAI each state that they are voluntary and not legally binding. The G7’s Hiroshima code states that it provides voluntary guidance. The King’s Speech of 2024 said the government would seek legislation on developers of the most powerful models; a search of the 2026 briefing notes found none of the terms “frontier”, “most powerful” or “AI Bill”.
Records: UK–Anthropic memorandum (2025): exception ↗ · UK–Cohere memorandum (2025): exception ↗ · UK–OpenAI memorandum (2025): exception ↗ · Hiroshima Process Code of Conduct (2023): threshold or trigger ↗ · King's Speech (2024): required action ↗ · King's Speech 2026 briefing notes: required action ↗
How to read the standings Each text carries one standing, chosen from the records shown with it. A standing describes the text’s form and recorded stage. It is not a finding on its legal effect, and it never shows compliance.
Binding text A law, regulation, rule, measure or mandatory standard, recorded as enacted, signed, issued or published by a body that makes such texts.
Law requiring a framework A law that requires developers in the categories it defines to write and publish a safety framework or protocol.
Voluntary code with a legal role A voluntary code that a law lets providers rely on to demonstrate compliance.
Executive order An order of the US President, recorded as signed. It directs the federal government; what each order requires is quoted from its text.
Voluntary A commitment, code, memorandum or reporting framework whose own text or publisher calls it voluntary or not binding.
Memorandum A memorandum between a government and a developer whose recorded text states no binding force either way.
Declaration or statement A declaration, summit or ministerial statement, speech, press release or announcement by a government. It states positions or plans.
Bill, draft or proposal A bill, draft, proposed rule or recommendation that the records do not show as adopted in that form.
Guidance, report or plan Interpretive guidance, an assessment, a recommended standard, a report, an evaluation summary or a policy plan.
Known from an announcement A text known only from an announcement or a later statement; the text itself was not inspected.
Revoked Revocation announced Vetoed Later version recorded Added where a later record revokes, announces the revocation of, vetoes or replaces the text with another version. Each opens its record. European Union EU AI Act (2024) Regulation (EU) 2024/1689 (Artificial Intelligence Act), as published in the Official Journal
Binding text
Regulation (EU law) Publisher: European Parliament and Council of the European Union (as legislators)
Standing shown by: The European Parliament and the Council adopted Regulation (EU) 2024/1689, the Artificial Intelligence Act… ↗ · Article 113 provides, in its opening paragraph, that the Regulation enters into force on the twentieth day… ↗
Signed 2024-06-13 · signature Inspect ↗ Published 2024-07-12 · document record Takes effect: date fixed by the text’s rule 2024-08-01 · date set by the text · effective Inspect ↗ Framework editions that name it Microsoft · Frontier Governance Framework, Feb 2026 ↗ Names this text OpenAI · Frontier Governance Framework, May 2026 edition ↗ Names this text What the text states · 36 provisions recorded Threshold or trigger Annex XIII criteria for designating models with systemic risk Annex XIII, points (a) to (g) Inspect claim & evidence ↗ Threshold or trigger Definition of general-purpose AI model “displays significant generality and is capable of competently performing a wide range of distinct tasks” Article 3, point (63) Inspect claim & evidence ↗ Threshold or trigger Systemic-risk classification on either of two conditions “shall be classified as a general-purpose AI model with systemic risk if it meets any of the following conditions” Article 51(1), chapeau Inspect claim & evidence ↗ Threshold or trigger Classification by Commission decision, ex officio or on a qualified alert, against Annex XIII “based on a decision of the Commission, ex officio or following a qualified alert from the scientific panel” Article 51(1), point (b) Inspect claim & evidence ↗ Threshold or trigger Commission to amend thresholds and supplement benchmarks by delegated acts “to amend the thresholds listed in paragraphs 1 and 2 of this Article” Article 51(3) Inspect claim & evidence ↗ Threshold or trigger Codes of practice to be ready by 2 May 2025 “Codes of practice shall be ready at the latest by 2 May 2025.” Article 56(9), first sentence Inspect claim & evidence ↗ Threshold or trigger Entry into force: the twentieth day following publication (computed: 1 August 2024) “shall enter into force on the twentieth day following that of its publication in the Official Journal of the European Union” Article 113, first paragraph Inspect claim & evidence ↗ Threshold or trigger General date of application: 2 August 2026 “It shall apply from 2 August 2026.” Article 113, second paragraph Inspect claim & evidence ↗ Threshold or trigger Presumed high-impact capabilities: training compute greater than 10^25 floating point operations “presumed to have high impact capabilities pursuant to paragraph 1, point (a), when the cumulative amount of computation used for its training measured in floating point operations is greater than” Article 51(2), opening words Inspect claim & evidence ↗ Threshold or trigger Systemic risk: a risk specific to the high-impact capabilities of general-purpose AI models “a risk that is specific to the high-impact capabilities of general-purpose AI models” Article 3, point (65) Inspect claim & evidence ↗ Required action Commission to publish and update a list of general-purpose AI models with systemic risk “a list of general-purpose AI models with systemic risk is published” Article 52(6) Inspect claim & evidence ↗ Required action Information and documentation for downstream AI system providers “to providers of AI systems who intend to integrate the general-purpose AI model into their AI systems” Article 53(1), point (b) Inspect claim & evidence ↗ Required action Copyright compliance policy, including identifying and complying with text-and-data-mining reservations “put in place a policy to comply with Union law on copyright and related rights” Article 53(1), point (c) Inspect claim & evidence ↗ Required action Public, sufficiently detailed summary of training content on the AI Office template “a sufficiently detailed summary about the content used for training of the general-purpose AI model, according to a template provided by the AI Office” Article 53(1), point (d) Inspect claim & evidence ↗ Required action Providers may rely on codes of practice for Article 53(1) until a harmonised standard is published Article 53(4), first and third sentences Inspect claim & evidence ↗ Required action Third-country providers to appoint an EU authorised representative before placing a model on the Union market “providers established in third countries shall, by written mandate, appoint an authorised representative which is established in the Union” Article 54(1) Inspect claim & evidence ↗ Required action Mandated tasks of the authorised representative, including keeping Annex XI documentation for 10 years Article 54(3), points (a) to (d) Inspect claim & evidence ↗ Required action Systemic-risk duties apply in addition to Articles 53 and 54 “In addition to the obligations listed in Articles 53 and 54” Article 55(1), chapeau Inspect claim & evidence ↗ Required action Models placed on the market before 2 August 2025: comply by 2 August 2027 “shall take the necessary steps in order to comply with the obligations laid down in this Regulation by 2 August 2027” Article 111(3) Inspect claim & evidence ↗ Required action Assess and mitigate possible systemic risks at Union level “assess and mitigate possible systemic risks at Union level” Article 55(1), point (b) Inspect claim & evidence ↗ Required action Providers may rely on codes of practice until a harmonised standard is published “may rely on codes of practice within the meaning of Article 56 to demonstrate compliance” Decides: Commission (assessment of alternative adequate means) Article 55(2), first sentence Inspect claim & evidence ↗ Required action Adequate cybersecurity protection for the model and its physical infrastructure “ensure an adequate level of cybersecurity protection” Article 55(1), point (d) Inspect claim & evidence ↗ Required action Report serious incidents without undue delay to the AI Office “report, without undue delay, to the AI Office and, as appropriate, to national competent authorities, relevant information about serious incidents” Informed: AI Office and, as appropriate, national competent authorities Article 55(1), point (c) Inspect claim & evidence ↗ Required action Model evaluation with adversarial testing for models with systemic risk “perform model evaluation in accordance with standardised protocols and tools reflecting the state of the art, including conducting and documenting adversarial testing” Article 55(1), point (a) Inspect claim & evidence ↗ Required action Draw up and keep up to date technical documentation of the model “draw up and keep up-to-date the technical documentation of the model” Informed: AI Office and national competent authorities, on request Article 53(1), point (a) Inspect claim & evidence ↗ Required action Notify the Commission within two weeks of meeting the systemic-risk condition “shall notify the Commission without delay and in any event within two weeks” Informed: Commission Article 52(1), first sentence Inspect claim & evidence ↗ Who decides Commission designation on Annex XIII criteria; delegated acts to amend Annex XIII Article 52(4) Inspect claim & evidence ↗ Who decides AI Office power to conduct evaluations of general-purpose AI models “The AI Office, after consulting the Board, may conduct evaluations of the general-purpose AI model concerned” Article 92(1), chapeau Inspect claim & evidence ↗ Who decides Commission may request access to the model through APIs or other technical means, including source code “through APIs or further appropriate technical means and tools, including source code” Article 92(3) Inspect claim & evidence ↗ Who decides AI Office and the Board assess whether codes of practice cover Articles 53 and 55 “The AI Office and the Board shall assess whether the codes of practice cover the obligations provided for in Articles 53 and 55” Decides: AI Office and the Board (assessment of adequacy) Article 56(6), first paragraph Inspect claim & evidence ↗ Who decides Commission may set common rules by implementing act if no adequate code exists by 2 August 2025 “If, by 2 August 2025, a code of practice cannot be finalised, or if the AI Office deems it is not adequate following its assessment under paragraph 6 of this Article, the Commission may provide, by means of implementing acts, common rules” Decides: Commission (by implementing act) Article 56(9), second subparagraph Inspect claim & evidence ↗ Who decides Commission fines on providers of general-purpose AI models up to 3% of turnover or EUR 15 million “The Commission may impose on providers of general-purpose AI models fines not exceeding 3 % of their annual total worldwide turnover in the preceding financial year or EUR 15 000 000, whichever is higher” Decides: Commission Article 101(1), opening Inspect claim & evidence ↗ Exception Provider may argue, with its notification, that the model exceptionally does not present systemic risks “sufficiently substantiated arguments to demonstrate that, exceptionally, although it meets that requirement” Article 52(2) Inspect claim & evidence ↗ Exception Reassessment of a designation on reasoned request, at the earliest six months after the decision “Providers may request reassessment at the earliest six months after the designation decision.” Article 52(5), third sentence Inspect claim & evidence ↗ Exception Chapters III (Section 4), V, VII and XII and Article 78 apply from 2 August 2025, except Article 101 “Chapter III Section 4, Chapter V, Chapter VII and Chapter XII and Article 78 shall apply from 2 August 2025, with the exception of Article 101” Article 113, third paragraph, point (b) Inspect claim & evidence ↗ Exception Open-source exemption from two documentation duties does not cover models with systemic risk “This exception shall not apply to general-purpose AI models with systemic risks.” Article 53(2), last sentence Inspect claim & evidence ↗ Exact text: Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) ↗ Archived copy, captured 2024-07-12: Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) ↗ Document page
EU GPAI Code of Practice, initial draft General-Purpose AI Code of Practice, initial draft (14 November 2024)
Bill, draft or proposal Later version recorded Record 1 ↗ Record 2 ↗
Draft code of practice (under the EU AI Act) Publisher: European Commission
Standing shown by: The European Commission published the initial draft of the General-Purpose AI Code of Practice on 14… ↗
Published 2024-11-14 · publication Record 1 ↗ Record 2 ↗ What the text states · 3 provisions recorded Threshold or trigger Six types of systemic risk: cyber offence; CBRN; loss of control; automated AI R&D; persuasion and manipulation; large-scale discrimination “Signatories will treat the following as systemic risks” Measure 6.1, chapeau (p. 17) Inspect claim & evidence ↗ Required action Measure 7: adopt, implement and make available a Safety and Security Framework “The Signatories commit to adopting, implementing, and making available a Safety and Security Framework (SSF)” Measure 7, first sentence (p. 21) Inspect claim & evidence ↗ Required action Send Safety and Security Reports to the AI Office before placing a new model with systemic risk on the market “Signatories will send the AI Office SSRs ahead of the decisions they pertain to, in particular before placing a new general-purpose AI model with systemic risk on the market.” Sub-Measure 20.3 (p. 34) Inspect claim & evidence ↗ Exact text: First Draft General-Purpose AI Code of Practice ↗ Document page
EU GPAI Code of Practice General-Purpose AI Code of Practice (final version, 10 July 2025)
Voluntary code with a legal role
Code of practice (voluntary instrument under the EU AI Act) Publisher: European Commission
Standing shown by: The European Commission received and published the final General-Purpose AI Code of Practice on 10 July… ↗ · Article 53(4) provides that providers of general-purpose AI models may rely on codes of practice within the… ↗ · Article 55(2) provides that providers of general-purpose AI models with systemic risk may rely on codes of… ↗
Published 2025-07-10 · publication Inspect ↗ Framework editions that name it Microsoft · Frontier Governance Framework, Feb 2026 ↗ States that it responds to this text OpenAI · Frontier Governance Framework, May 2026 edition ↗ States that it serves as its framework under this text xAI · Frontier AI Framework, June 2026 edition ↗ Names this text Changes from the earlier version · 2 recorded Inferred document comparison The final Code's Appendix 1.4 specifies four systemic risks (CBRN, loss of control, cyber offence, harmful manipulation), where the initial draft's Measure 6.1 listed six types, including automated use of models for AI research and development and large-scale discrimination, which are not among the final four. 2025-07-10 · publication · Inferred · Supported Inspect claim & evidence ↗ Inferred document comparison The final Code's Safety and Security chapter opens Commitment 1 by committing signatories to a three-step Framework adoption process (creating, implementing and updating a “state-of-the-art” Framework) and to notifying the AI Office of it, where the initial draft's Measure 7 opened by committing signatories to adopting, implementing and making available a Safety and Security Framework. 2025-07-10 · publication · Inferred · Supported Inspect claim & evidence ↗ What the text states · 9 provisions recorded Threshold or trigger Provide requested information to downstream providers within a reasonable time and no later than 14 days “Signatories will provide such information within a reasonable timeframe, and no later than 14 days of receiving the request save for exceptional circumstances.” Commitment 1, Measure 1.2, third paragraph Inspect claim & evidence ↗ Threshold or trigger The chapter is relevant only to providers of general-purpose AI models with systemic risk “this Safety and Security Chapter (“Chapter”) is only relevant for providers of general-purpose AI models with systemic risk and not AI systems” Recitals, (b) (p. 3) Inspect claim & evidence ↗ Threshold or trigger Specified systemic risks: CBRN, loss of control, cyber offence, harmful manipulation “the following are treated as specified systemic risks for the purpose of systemic risk identification in Measure 2.1, point (2)” Appendix 1.4, introductory sentence (PDF p. 37 of the file served on 2026-10-04; p. 35 of the 2025-07-10 capture) Inspect claim & evidence ↗ Required action Use web-crawlers that read and follow robots.txt instructions (IETF RFC 9309) “to employ web-crawlers that read and follow instructions expressed in accordance with the Robot Exclusion Protocol (robots.txt), as specified in the Internet Engineering Task Force (IETF) Request for Comments No. 9309” Commitment 1, Measure 1.3(1)(a) Inspect claim & evidence ↗ Required action Keep previous versions of model documentation for 10 years after the model is placed on the market “while keeping previous versions of the Model Documentation for a period ending 10 years after the model has been placed on the market” Commitment 1, Measure 1.1, second paragraph Inspect claim & evidence ↗ Required action Adopt a Safety and Security Framework and notify the AI Office “Signatories commit to adopting a state-of-the-art Safety and Security Framework (“Framework”).” Commitment 1, first paragraph (p. 6) Inspect claim & evidence ↗ Required action Serious-incident reports within 2, 5, 10 or 15 days of awareness “not later than two days after the Signatories become aware of the involvement of their model in the incident” Measure 9.3, opening sentence and point (1) (PDF p. 26 of the file served on 2026-10-04; p. 25 of the 2025-07-10 capture) Inspect claim & evidence ↗ Required action Proceed only if systemic risks are determined acceptable “Signatories will only proceed with the development, the making available on the market, and/or the use of the model, if the systemic risks stemming from the model are determined to be acceptable” Measure 4.2, first sentence (p. 16) Inspect claim & evidence ↗ Exception Adherence is not conclusive evidence of compliance with the AI Act “adherence to the Code does not constitute conclusive evidence of compliance with these obligations under the AI Act” Objectives, point A (p. 2) Inspect claim & evidence ↗ 13 signatory records on the commitments page → Exact text: General-Purpose AI Code of Practice: Copyright Chapter ↗ Exact text: General-Purpose AI Code of Practice: Safety and Security chapter (final version, 10 July 2025) ↗ Archived copy, captured 2025-07-10: General-Purpose AI Code of Practice: Safety and Security chapter (final version, 10 July 2025) ↗ Exact text: General-Purpose AI Code of Practice: Transparency Chapter ↗ Document page
Training-content summary template (2025) Explanatory Notice and Template for the Public Summary of Training Content for general-purpose AI models (C(2025) 8311 final)
Guidance, report or plan
Commission communication with template Publisher: European Commission
Standing shown by: The Explanatory Notice states that the public summary of training content should be updated whenever the… ↗
Published 2025-07-24 · document record What the text states · 2 provisions recorded Threshold or trigger Template asks for the top 10% of crawled domain names by content size (top 5% or 1,000 for SMEs) “by listing the top 10 % of all domain names determined by the size of the content scraped (in a representative manner across all modalities where applicable).” Template Section 2.3, "Summary of the most relevant domain names crawled", first instruction Inspect claim & evidence ↗ Required action Summary updated when further training on additional data requires it, at six-month intervals or sooner “The Summary should also be updated whenever the provider further trains its own general-purpose AI model placed on the market on additional data that requires an update of the content of the Summary.” Explanatory Notice, paragraph (29), first sentence Inspect claim & evidence ↗ Exact text: Communication from the Commission: Explanatory Notice and Template for the Public Summary of Training Content for general-purpose AI models required by Article 53(1)(d) of Regulation (EU) 2024/1689 (AI Act), C(2025) 8311 final ↗ Document page
Commission Opinion on the GPAI Code (2025) Commission Opinion C(2025) 5361 final on the assessment of the General-Purpose AI Code of Practice
Guidance, report or plan
Commission opinion (assessment under Article 56 of the AI Act) Publisher: European Commission
Standing shown by: The European Commission issued Opinion C(2025) 5361 final, dated 1 August 2025, on the assessment of the… ↗
Issued 2025-08-01 · signature Record 1 ↗ Record 2 ↗ Record 3 ↗ What the text states · 4 provisions recorded Who decides The Board's assessment is published separately “The assessment by the Board is published separately.” Paragraph (2), last sentences (p. 2) Inspect claim & evidence ↗ Who decides Code adequately covers Articles 53 and 55 obligations and meets the aims of Article 56 Section 5, paragraph (57) (p. 12) Inspect claim & evidence ↗ Who decides Transparency and Copyright chapters cover Article 53(1); Safety and Security chapter covers Article 55(1) and Annex XI Section 2 “the Transparency and Copyright Chapters adequately cover the obligations provided for in Article 53(1)” Paragraph (28) (p. 7) Inspect claim & evidence ↗ Exception Opinion does not prejudge enforcement, does not bind interpretation, and does not pre-empt a later finding of inadequacy “This Opinion does not prejudge any future action that the Commission, the Board, and other competent authorities, as applicable, may undertake in the enforcement of the AI Act.” Paragraph (3), first sentence (p. 2) Inspect claim & evidence ↗ Exact text: Commission Opinion C(2025) 5361 on the assessment of the General-Purpose AI Code of Practice ↗ Archived copy, captured 2025-08-13: Commission Opinion C(2025) 5361 on the assessment of the General-Purpose AI Code of Practice ↗ Document page
Commission GPAI guidelines (2025) Commission Guidelines on the scope of the obligations for providers of general-purpose AI models established by Regulation (EU) 2024/1689 (C(2025) 7719 final)
Guidance, report or plan
Commission guidelines (non-binding interpretation) Publisher: European Commission
Standing shown by: The Commission Guidelines on the scope of the obligations for providers of general-purpose AI models, in the… ↗
Published 2025-07-18 · document record Published 2025-11-19 · publication Inspect ↗ What the text states · 7 provisions recorded Threshold or trigger No enforcement action in the first year; fines enforced from 2 August 2026; obligations apply from 2 August 2025 “In the first year from 2 August 2025 onwards, the Commission cannot take any enforcement actions” Section 5.3, last paragraph, sentence beginning "In the first year" Inspect claim & evidence ↗ Threshold or trigger Indicative criterion for a general-purpose AI model: training compute above 10²³ FLOP and language, text-to-image or text-to-video output “an indicative criterion for a model to be considered a general-purpose AI model is that its training compute is greater than” Paragraph (17) Inspect claim & evidence ↗ Required action Providers outside an adequately assessed code are expected to show compliance by other adequate means and report to the AI Office “Providers of general-purpose AI models that do not adhere to a code of practice that is assessed as adequate are expected to demonstrate how they comply with their obligations under Chapter V of the AI Act via other adequate means” Paragraph (92), first sentence Inspect claim & evidence ↗ Required action A systemic-risk notification may be due before training is complete; estimate training compute before the pre-training run “a notification may be required before training is complete, if the provider can reasonably foresee that the requirement that leads to the presumption of the model having high-impact capabilities is reasonably likely to be met” Paragraph (30), first sentence Inspect claim & evidence ↗ Who decides For adherents to an adequately assessed code, enforcement focuses on monitoring adherence “the Commission will focus its enforcement activities on monitoring their adherence to the code of practice.” Decides: Commission (focus of enforcement activities) Paragraph (91), fourth sentence Inspect claim & evidence ↗ Who decides Commitments implemented in line with an adequately assessed code may be a mitigating factor for fines “The Commission may take into account commitments implemented in line with a code of practice that is assessed as adequate as a mitigating factor when fixing the amount of fines, depending on the specific circumstances (Article 101(1) AI Act).” Decides: Commission (amount of fines) Paragraph (93) Inspect claim & evidence ↗ Who decides Guidelines are not binding; the CJEU gives authoritative interpretation; the Commission bases enforcement on them “These guidelines are not binding for providers of general-purpose AI models; an authoritative interpretation of the AI Act may only be given by the Court of Justice of the European Union” Decides: Court of Justice of the European Union (authoritative interpretation of the AI Act) Paragraph (9), first sentence Inspect claim & evidence ↗ Exact text: Communication from the Commission: Commission Guidelines on the scope of the obligations for providers of general-purpose AI models established by Regulation (EU) 2024/1689 (AI Act), C(2025) 7719 final ↗ Document page
Draft implementing regulation on GPAI proceedings (2026) Draft Commission Implementing Regulation on detailed arrangements for the conduct of certain proceedings by the Commission pursuant to Regulation (EU) 2024/1689 (Ares(2026)2709234)
Bill, draft or proposal
Draft implementing act (not adopted) Publisher: European Commission
Standing shown by: On 12 March 2026 the Commission published for feedback a draft implementing regulation on detailed… ↗
Published 2026-03-12 · publication Inspect ↗ What the text states · 1 provision recorded Required action A decision requesting access to a model specifies the technical means; access may include source code and model weights “that decision shall specify the technical means, components, and conditions by which the provider shall provide such access.” Article 2(1), first sentence Inspect claim & evidence ↗ Exact text: Draft Commission Implementing Regulation (EU) .../... on detailed arrangements for the conduct of certain proceedings by the Commission pursuant to Regulation (EU) 2024/1689 ↗ Document page
EU Digital Omnibus on AI (2026) Regulation (EU) 2026/1744 (Digital Omnibus on AI), as published in the Official Journal
Binding text
Regulation (EU law) Publisher: European Parliament and Council of the European Union (as legislators)
Standing shown by: The European Parliament and the Council adopted Regulation (EU) 2026/1744, the Digital Omnibus on AI, which… ↗
Signed 2026-07-08 · signature Inspect ↗ Published 2026-07-24 · document record What the text states · 4 provisions recorded Threshold or trigger Enters into force on the third day following publication “This Regulation shall enter into force on the third day following that of its publication in the Official Journal of the European Union.” Article 4, first paragraph Inspect claim & evidence ↗ Threshold or trigger High-risk AI system rules apply from 2 December 2027 (Annex III) and 2 August 2028 (Annex I) “Chapter III, Sections 1, 2, and 3, with the exception of Article 6(5), shall apply from:” Article 1, point (40)(b), new Article 113(c), chapeau Inspect claim & evidence ↗ Who decides AI Office exclusively competent for AI systems built on a general-purpose AI model by the same provider “The AI Office shall be exclusively competent for the supervision and enforcement of the obligations under this Regulation in relation to the following AI systems” Decides: AI Office (exclusive competence for supervision and enforcement) Article 1, point (31)(b), new Article 75(1) Inspect claim & evidence ↗ Who decides The Commission assesses whether codes of practice cover Articles 53 and 55 “The Commission, taking utmost account of the opinion of the Board, shall assess whether the codes of practice cover the obligations provided for in Articles 53 and 55” Decides: Commission, taking utmost account of the opinion of the Board Article 1, point (21) Inspect claim & evidence ↗ Exact text: Regulation (EU) 2026/1744 of the European Parliament and of the Council of 8 July 2026 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 as regards the simplification of the implementation of harmonised rules on artificial intelligence (Digital Omnibus on AI) ↗ Archived copy, captured 2026-07-28: Regulation (EU) 2026/1744 of the European Parliament and of the Council of 8 July 2026 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 as regards the simplification of the implementation of harmonised rules on artificial intelligence (Digital Omnibus on AI) ↗ Document page
China Chinese texts are quoted in Chinese, with a working English translation marked as Selekoa’s. TC260’s AI Safety Governance Framework and Chinese developers’ own frameworks are recorded on their own document pages and are not part of this page.
China generative AI interim measures (2023) Interim Measures for the Management of Generative Artificial Intelligence Services (生成式人工智能服务管理暂行办法), Order No. 15
Binding text
Administrative measures (China) Publisher: Cyberspace Administration of China (国家互联网信息办公室)
Standing shown by: The Cyberspace Administration of China and six other departments issued Order No. 15, the Interim Measures… ↗
Issued 2023-07-10 · publication Inspect ↗ What the text states · 11 provisions recorded Threshold or trigger Generative AI service provider: organisations and individuals providing generative AI services, including via programmable interfaces “包括通过提供可编程接口等方式提供生成式人工智能服务” Article 22 (第二十二条), point (2) Inspect claim & evidence ↗ Threshold or trigger The Measures apply to services that provide generated content to the public within China “利用生成式人工智能技术向中华人民共和国境内公众提供生成文本、图片、音频、视频等内容的服务（以下称生成式人工智能服务），适用本办法。” Article 2 (第二条), first sentence Inspect claim & evidence ↗ Required action Label generated content such as images and video under the Deep Synthesis Provisions Article 12 (第十二条) Inspect claim & evidence ↗ Required action On finding illegal content: stop generation and transmission, remove, rectify including by model optimisation training, and report “采取模型优化训练等措施进行整改” Article 14 (第十四条), first paragraph Inspect claim & evidence ↗ Required action Explain training-data sources, scale, types, annotation rules and algorithmic mechanisms to supervising authorities on request “按要求对训练数据来源、规模、类型、标注规则、算法机制机理等予以说明” Article 19 (第十九条), first paragraph Inspect claim & evidence ↗ Required action Providers to carry out pre-training and optimisation-training data processing lawfully, under listed rules “应当依法开展预训练、优化训练等训练数据处理活动” Article 7 (第七条), chapeau Inspect claim & evidence ↗ Required action Use data and foundation models with lawful sources “使用具有合法来源的数据和基础模型” Article 7 (第七条), point (1) Inspect claim & evidence ↗ Required action Training data: no infringement of IP; consent or other legal basis for personal information; measures to improve data quality Article 7 (第七条), points (2)–(4) Inspect claim & evidence ↗ Required action Services with public-opinion attributes or social-mobilisation capacity need a security assessment and algorithm filing “提供具有舆论属性或者社会动员能力的生成式人工智能服务的，应当按照国家有关规定开展安全评估，并按照《互联网信息服务算法推荐管理规定》履行算法备案和变更、注销备案手续。” Article 17 (第十七条) Inspect claim & evidence ↗ Who decides Penalties under the Cybersecurity, Data Security, Personal Information Protection and Science and Technology Progress laws; otherwise warnings, criticism, orders to correct, suspension Article 21 (第二十一条), first paragraph Inspect claim & evidence ↗ Exception Measures do not apply to research and development of generative AI that does not provide services to the public in China “未向境内公众提供生成式人工智能服务的，不适用本办法的规定” Article 2 (第二条), third paragraph Inspect claim & evidence ↗ Exact text: 生成式人工智能服务管理暂行办法 (Interim Measures for the Management of Generative Artificial Intelligence Services) ↗ In Simplified Chinese (zh-Hans) Exact text: 生成式人工智能服务管理暂行办法 (Interim Measures for the Management of Generative Artificial Intelligence Services), gov.cn State Council policy library copy ↗ In Simplified Chinese (zh-Hans) Document page
GB 45438-2025 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)
Binding text
Mandatory national standard (China)
Standing shown by: The national standards catalogue lists GB 45438-2025, Cybersecurity technology — Labeling method for content… ↗
Published 2025-02-28 · publication Inspect ↗ No provision of this text is recorded. That does not mean it has none.
Exact text: GB 45438-2025 网络安全技术 人工智能生成合成内容标识方法 (catalogue entry, 国家标准全文公开系统) ↗ In Simplified Chinese (zh-Hans) Document page
China AI-generated content labelling measures (2025) Measures for Labelling Artificial Intelligence-Generated Synthetic Content (人工智能生成合成内容标识办法)
Binding text
Departmental normative document (China), issued by notice Publisher: Cyberspace Administration of China (国家互联网信息办公室)
Standing shown by: The Cyberspace Administration of China, the Ministry of Industry and Information Technology, the Ministry of… ↗
Issued 2025-03-07 · publication Inspect ↗ Takes effect: date stated in the text 2025-09-01 · date set by the text · effective Inspect ↗ What the text states · 8 provisions recorded Threshold or trigger Effective from 1 September 2025 “本办法自2025年9月1日起施行。” Article 14 (第十四条) Inspect claim & evidence ↗ Threshold or trigger Applies to network information service providers within the Algorithm Recommendation, Deep Synthesis and Generative AI rules Article 2 (第二条) Inspect claim & evidence ↗ Threshold or trigger Labels are explicit (perceptible to users) or implicit (embedded in file data) Article 3 (第三条), second to fourth paragraphs Inspect claim & evidence ↗ Required action Labelling must also meet laws, regulations, rules and mandatory national standards “还应当符合相关法律、行政法规、部门规章和强制性国家标准的要求” Article 11 (第十一条) Inspect claim & evidence ↗ Required action Submit labelling materials in algorithm filing and security assessment procedures “服务提供者在履行算法备案、安全评估等手续时，应当按照本办法提供生成合成内容标识相关材料” Article 12 (第十二条) Inspect claim & evidence ↗ Required action Explicit labels on text, audio, images, video and virtual scenes; downloads and exports to carry them “服务提供者提供生成合成内容下载、复制、导出等功能时，应当确保文件中含有满足要求的显式标识” Article 4 (第四条), second paragraph Inspect claim & evidence ↗ Required action Implicit labels in file metadata with content attributes, provider name or code and content number “隐式标识包含生成合成内容属性信息、服务提供者名称或者编码、内容编号等制作要素信息” Article 5 (第五条), first paragraph Inspect claim & evidence ↗ Exception Unlabelled output on user request only after user agreement, with logs kept at least six months “依法留存提供对象信息等相关日志不少于六个月” Article 9 (第九条) Inspect claim & evidence ↗ Exact text: 关于印发《人工智能生成合成内容标识办法》的通知 (Notice on issuing the Measures for Labelling AI-Generated Synthetic Content), 国信办通字〔2025〕2号 ↗ In Simplified Chinese (zh-Hans) Document page
GB/T 45654-2025 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)
Guidance, report or plan
Recommended national standard (China)
Standing shown by: The national standards catalogue lists GB/T 45654-2025, Cybersecurity technology — Basic security… ↗
Published 2025-04-25 · publication Inspect ↗ No provision of this text is recorded. That does not mean it has none.
Exact text: GB/T 45654-2025 网络安全技术 生成式人工智能服务安全基本要求 (catalogue entry, 国家标准全文公开系统) ↗ In Simplified Chinese (zh-Hans) Document page
Cybersecurity Law amendment decision (2025) Decision of the NPC Standing Committee on Amending the Cybersecurity Law of the People's Republic of China (全国人民代表大会常务委员会关于修改《中华人民共和国网络安全法》的决定)
Binding text
Amending decision of the NPC Standing Committee Publisher: Standing Committee of the National People's Congress (全国人民代表大会常务委员会)
Standing shown by: The Standing Committee of the 14th National People's Congress adopted the Decision on Amending the… ↗
Adopted 2025-10-28 Inspect ↗ Published 2025-10-28 · publication Inspect ↗ Takes effect: date stated in the text 2026-01-01 · date set by the text · effective Record 1 ↗ Record 2 ↗ What the text states · 2 provisions recorded Threshold or trigger Effective from 1 January 2026 “本决定自2026年1月1日起施行。” Closing paragraph Inspect claim & evidence ↗ Required action New Article 20: the state supports AI research, training data and compute infrastructure, AI ethics norms, risk monitoring and assessment, and safety supervision “国家支持人工智能基础理论研究和算法等关键技术研发，推进训练数据资源、算力等基础设施建设” Point 3 (三), added Article 20, first paragraph, first two clauses Inspect claim & evidence ↗ Exact text: 全国人民代表大会常务委员会关于修改《中华人民共和国网络安全法》的决定 (Decision of the NPC Standing Committee on Amending the Cybersecurity Law) ↗ In Simplified Chinese (zh-Hans) Document page
United States: federal government White House commitments (July 2023) Ensuring Safe, Secure, and Trustworthy AI: voluntary AI commitments (July 2023)
Voluntary Later version recorded Inspect ↗
Voluntary commitment text Publisher: The White House
Standing shown by: The July 2023 commitments text states that companies intend the voluntary commitments to remain in effect… ↗
Published 2023-07-21 · publication Inspect ↗ Framework editions that name it Anthropic · Responsible Scaling Policy, v1.0 ↗ Names this text OpenAI · Preparedness Framework, Beta ↗ States that it fulfils this text Anthropic · Responsible Scaling Policy, v2.0 ↗ States that it fulfils this text Anthropic · Responsible Scaling Policy, v2.1 ↗ States that it fulfils this text Anthropic · Responsible Scaling Policy, v2.2 ↗ States that it fulfils this text What the text states · 5 provisions recorded Threshold or trigger Scope: generative models more powerful than the current industry frontier “they apply only to generative models that are overall more powerful than the current industry frontier” Page 2, paragraph beginning “Scope:” Inspect claim & evidence ↗ Required action Commitment 3: safeguards for unreleased model weights “limiting access to model weights to those whose job function requires it” Page 3, commitment 3 Inspect claim & evidence ↗ Required action Commitment 6: public reports for significant model releases in scope “publish reports for all new significant model public releases within scope” Page 4, commitment 6 Inspect claim & evidence ↗ Required action Commitment 1: internal and external red-teaming of models in scope “including drawing on independent domain experts, for all major public releases of new models within scope” Page 2, commitment 1, paragraph beginning “Companies making this commitment understand” Inspect claim & evidence ↗ Exception Commitments intended to remain in effect until regulations on the same issues come into force “Companies intend these voluntary commitments to remain in effect until regulations covering substantially the same issues come into force.” Page 2, paragraph introducing the commitments Inspect claim & evidence ↗ 7 signatory records on the commitments page → Exact text: Ensuring Safe, Secure, and Trustworthy AI (voluntary AI commitments, July 2023) ↗ Archived copy, captured 2023-07-21: Ensuring Safe, Secure, and Trustworthy AI (voluntary AI commitments, July 2023) ↗ Document page
White House commitments (Sept. 2023) Voluntary AI Commitments (September 2023)
Voluntary
Voluntary commitment text Publisher: The White House
Standing shown by: On 12 September 2023 the White House announced that it had secured a second round of voluntary commitments… ↗
Published 2023-09-12 · publication Inspect ↗ Changes from the earlier version · 1 recorded Inferred document comparison Compared with the July 2023 text, the September 2023 commitments text measures which models are in scope against the committing company’s own current most advanced model instead of the current industry frontier. 2023-09-12 · publication · Inferred · Supported Inspect claim & evidence ↗ What the text states · 1 provision recorded Threshold or trigger Scope: generative models more powerful than the company’s own current most advanced model “they apply only to generative models that are overall more powerful than the current most advanced model produced by the company making the commitment” Page 1, paragraph beginning “Scope:” Inspect claim & evidence ↗ 8 signatory records on the commitments page → Exact text: Voluntary AI Commitments (September 2023) ↗ Archived copy, captured 2023-09-12: Voluntary AI Commitments (September 2023) ↗ Document page
EO 14110 (2023) Executive Order 14110: Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence
Executive order Revoked Inspect ↗
Executive order (United States) Publisher: The White House
Standing shown by: Executive Order 14110, "Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence", is… ↗
Signed 2023-10-30 · signature Inspect ↗ Published 2023-11-01 · publication Inspect ↗ Revoked 2025-01-20 · signature Inspect ↗ What the text states · 3 provisions recorded Threshold or trigger Definition of "dual-use foundation model" (Sec. 3(k)) “means an AI model that is trained on broad data; generally uses self-supervision; contains at least tens of billions of parameters; is applicable across a wide range of contexts; and that exhibits, or could be easily modified to exhibit, high levels of performance at tasks that pose a serious risk” Sec. 3(k), 88 FR 75194 Inspect claim & evidence ↗ Threshold or trigger Interim technical threshold for the Sec. 4.2(a) reporting requirements (Sec. 4.2(b)(i)) “Until such technical conditions are defined, the Secretary shall require compliance with these reporting requirements for: (i) any model that was trained using a quantity of computing power greater than” Sec. 4.2(b)(i), 88 FR 75197 Inspect claim & evidence ↗ Required action Reporting duty for developers of potential dual-use foundation models (Sec. 4.2(a)(i)) “Within 90 days of the date of this order” Sec. 4.2(a), 88 FR 75197 Inspect claim & evidence ↗ Exact text: Executive Order 14110 of October 30, 2023: Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence (88 FR 75191) ↗ Document page
EO 14179 (2025) Executive Order 14179: Removing Barriers to American Leadership in Artificial Intelligence
Executive order
Executive order (United States) Publisher: The White House
Standing shown by: Executive Order 14179, "Removing Barriers to American Leadership in Artificial Intelligence", is dated 23… ↗
Signed 2025-01-23 · signature Inspect ↗ Published 2025-01-31 · document record What the text states · 1 provision recorded Required action Review of actions taken under the revoked EO 14110 (Sec. 5(a)) “shall immediately review, in coordination with the heads of all agencies as they deem relevant, all policies, directives, regulations, orders, and other actions taken pursuant to the revoked Executive Order 14110” Sec. 5(a), 90 FR 8741 Inspect claim & evidence ↗ Exact text: Executive Order 14179 of January 23, 2025: Removing Barriers to American Leadership in Artificial Intelligence (90 FR 8741) ↗ Document page
America's AI Action Plan (2025) Winning the Race: America’s AI Action Plan
Guidance, report or plan
Policy plan (United States) Publisher: The White House
Standing shown by: The White House released "Winning the AI Race: America’s AI Action Plan" on 23 July 2025. ↗
Published 2025-07-23 · publication Inspect ↗ What the text states · 2 provisions recorded Required action CAISI evaluation of frontier AI systems in partnership with frontier AI developers “Evaluate frontier AI systems for national security risks in partnership with frontier AI developers, led by CAISI at DOC in collaboration with other agencies with relevant expertise in CBRNE and cyber risks.” Pillar III, “Ensure that the U.S. Government is at the Forefront of Evaluating National Security Risks in Frontier Models”, Recommended Policy Actions, PDF p. 25 (printed p. 22) Inspect claim & evidence ↗ Required action Federal AI-related funding to consider state AI regulatory climate “consider a state's AI regulatory climate when making funding decisions and limit funding if the state's AI regulatory regimes may hinder the effectiveness of that funding or award” Pillar I, "Remove Red Tape and Onerous Regulation", Recommended Policy Actions, PDF p. 6 Inspect claim & evidence ↗ Exact text: Winning the Race: America’s AI Action Plan (July 2025) ↗ Document page
EO 14365 (2025) Executive Order 14365: Ensuring a National Policy Framework for Artificial Intelligence
Executive order
Executive order (United States) Publisher: The White House
Standing shown by: Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence", is dated 11… ↗
Signed 2025-12-11 · signature Inspect ↗ Published 2025-12-16 · document record What the text states · 4 provisions recorded Required action Commerce evaluation of State AI laws, including disclosure and reporting mandates (Sec. 4) “identify laws that require AI models to alter their truthful outputs, or that may compel AI developers or deployers to disclose or report information in a manner that would violate the First Amendment” Sec. 4, 90 FR 58500 Inspect claim & evidence ↗ Required action FCC proceeding on a preemptive Federal reporting and disclosure standard for AI models (Sec. 6) “initiate a proceeding to determine whether to adopt a Federal reporting and disclosure standard for AI models that preempts conflicting State laws” Sec. 6, 90 FR 58500 Inspect claim & evidence ↗ Required action Legislative recommendation for a uniform Federal framework preempting conflicting State AI laws (Sec. 8) “a legislative recommendation establishing a uniform Federal policy framework for AI that preempts State AI laws that conflict with the policy set forth in this order” Sec. 8(a), 90 FR 58501 Inspect claim & evidence ↗ Who decides AI Litigation Task Force to challenge State AI laws (Sec. 3) “establish an AI Litigation Task Force (Task Force) whose sole responsibility shall be to challenge State AI laws” Sec. 3, 90 FR 58499 Inspect claim & evidence ↗ Exact text: Executive Order 14365 of December 11, 2025: Ensuring a National Policy Framework for Artificial Intelligence (90 FR 58499) ↗ Document page
US AI legislative recommendations (2026) Legislative Recommendations: A National Policy Framework for Artificial Intelligence
Bill, draft or proposal
Legislative recommendations (United States) Publisher: The White House
Standing shown by: The White House published "Legislative Recommendations: A National Policy Framework for Artificial… ↗
Published 2026-03 · month only · publication Inspect ↗ What the text states · 1 provision recorded Required action States should not be permitted to regulate AI development “Congress should preempt state AI laws that impose undue burdens to ensure a minimally burdensome national standard” Section VII, "Establishing a Federal Policy Framework, Preempting Cumbersome State AI Laws", p. 4 Inspect claim & evidence ↗ Exact text: Legislative Recommendations: A National Policy Framework for Artificial Intelligence (March 2026) ↗ Document page
EO 14409 (2026) Executive Order 14409: Promoting Advanced Artificial Intelligence Innovation and Security
Executive order
Executive order (United States) Publisher: The White House
Standing shown by: Executive Order 14409, "Promoting Advanced Artificial Intelligence Innovation and Security", is dated 2 June… ↗
Signed 2026-06-02 · signature Inspect ↗ Published 2026-06-05 · document record What the text states · 3 provisions recorded Threshold or trigger Classified benchmarking process and threshold for designating a "covered frontier model" (Sec. 3(a)) “develop and maintain a classified benchmarking process to assess the advanced cyber capabilities of AI models and determine the threshold at which an AI model should be designated a” Sec. 3(a), 91 FR 34566 Inspect claim & evidence ↗ Required action Voluntary framework for pre-release Federal access to covered frontier models (Sec. 3(b)) “provide the Federal Government with access to covered frontier models … for a period of up to 30 days before they plan to release such models to other trusted partners” Sec. 3(b)(ii), 91 FR 34566 Inspect claim & evidence ↗ Exception No mandatory licensing, preclearance or permitting requirement (Sec. 3(c)) “Nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement” Sec. 3(c), 91 FR 34566 Inspect claim & evidence ↗ Exact text: Executive Order 14409 of June 2, 2026: Promoting Advanced Artificial Intelligence Innovation and Security (91 FR 34565) ↗ Document page
The US AI Safety Institute and CAISI: agreements and reported work with developers On 29 August 2024 NIST announced that the U.S. AI Safety Institute had agreements (a Memorandum of Understanding with each company) enabling collaboration on AI safety research, testing and evaluation with Anthropic, including access to major new models before and after public release. 2024-08-29 · announcement · Stated in source · Supported Announced Inspect claim & evidence ↗ On 29 August 2024 NIST announced that the U.S. AI Safety Institute had agreements (a Memorandum of Understanding with each company) enabling collaboration on AI safety research, testing and evaluation with OpenAI, including access to major new models before and after public release. 2024-08-29 · announcement · Stated in source · Supported Announced Inspect claim & evidence ↗ On 25 September 2025 NIST stated that CAISI had worked with Anthropic to identify security issues with its advanced AI systems, with research and evaluations completed in partnership with the UK AI Security Institute. 2025-09-25 · announcement · Stated in source · Supported Inspect claim & evidence ↗ On 25 September 2025 NIST stated that CAISI had worked with OpenAI to identify security issues with its advanced AI systems, with research and evaluations completed in partnership with the UK AI Security Institute. 2025-09-25 · announcement · Stated in source · Supported Inspect claim & evidence ↗ United States: states California SB 1047 (2024) California SB 1047 (2023–2024 session)
Bill, draft or proposal Vetoed Inspect ↗
Bill (vetoed)
Standing shown by: The Governor of California returned SB 1047 to the Senate without his signature by a veto message dated 29… ↗
Passed a vote 2024-05-21 Inspect ↗ Passed a vote 2024-08-28 Inspect ↗ Passed a vote 2024-08-29 Inspect ↗ Vetoed 2024-09-29 Inspect ↗ No provision of this text is recorded. That does not mean it has none.
Document page
California Report on Frontier AI Policy (2025) The California Report on Frontier AI Policy
Guidance, report or plan
Policy report
Standing shown by: The Joint California Policy Working Group on AI Frontier Models released its final report, "The California… ↗
Published 2025-06-17 · publication Inspect ↗ No provision of this text is recorded. That does not mean it has none.
Exact text: The California Report on Frontier AI Policy (June 17, 2025) ↗ Document page
California SB 53 California SB 53 (2025): Transparency in Frontier Artificial Intelligence Act, as chaptered
Law requiring a framework
Statute (chaptered bill) Publisher: California State Legislature
Standing shown by: The Governor of California approved Senate Bill 53, which enacts the Transparency in Frontier Artificial… ↗ · SB 53 requires a large frontier developer to write, implement, comply with, and clearly and conspicuously… ↗
Signed 2025-09-29 · signature Inspect ↗ Framework editions that name it xAI · Frontier AI Framework, December 2025 edition ↗ States that it complies with this text Microsoft · Frontier Governance Framework, Feb 2026 ↗ States that it responds to this text Anthropic · Responsible Scaling Policy, v3.0 ↗ Names this text Anthropic · Responsible Scaling Policy, v3.1 ↗ Names this text Anthropic · Responsible Scaling Policy, v3.2 ↗ Names this text Anthropic · Responsible Scaling Policy, v3.3 ↗ Names this text OpenAI · Frontier Governance Framework, May 2026 edition ↗ States that it serves as its framework under this text Anthropic · Responsible Scaling Policy, v3.4 ↗ Names this text What the text states · 19 provisions recorded Threshold or trigger Catastrophic risk: death of or serious injury to more than 50 people, or more than $1 billion in property damage or loss, from a single incident “more than 50 people or more than one billion dollars ($1,000,000,000)” SEC. 2, Bus. & Prof. Code § 22757.11(c)(1) Inspect claim & evidence ↗ Threshold or trigger Civil penalty up to $1,000,000 per violation “does not exceed one million dollars ($1,000,000) per violation.” SEC. 2, Bus. & Prof. Code § 22757.15(a) Inspect claim & evidence ↗ Threshold or trigger Frontier developer: has trained or initiated training of a frontier model “has trained, or initiated the training of, a frontier model” SEC. 2, Bus. & Prof. Code § 22757.11(h) Inspect claim & evidence ↗ Threshold or trigger Frontier model: foundation model trained with more than 10^26 operations “Frontier model” means a foundation model that was trained using a quantity of computing power greater than 10^26 integer or floating-point operations.” SEC. 2, Bus. & Prof. Code § 22757.11(i)(1) Inspect claim & evidence ↗ Threshold or trigger Large frontier developer: over $500 million annual gross revenue with affiliates “together with its affiliates collectively had annual gross revenues in excess of five hundred million dollars ($500,000,000) in the preceding calendar year.” SEC. 2, Bus. & Prof. Code § 22757.11(j) Inspect claim & evidence ↗ Required action Anonymous internal disclosure process with monthly status updates to the discloser and at-least-quarterly sharing with officers and directors “a reasonable internal process through which a covered employee may anonymously disclose information” Informed: Officers and directors of the large frontier developer, except any officer or director whose wrongdoing is alleged in a disclosure or response SEC. 4, Lab. Code § 1107.1(e)(1); paraphrase also covers § 1107.1(e)(2)(A)–(B) Inspect claim & evidence ↗ Required action Covered-employee disclosure and anti-retaliation protection “prevents a covered employee from disclosing, or retaliates against a covered employee for disclosing,” SEC. 4, Lab. Code § 1107.1(a) Inspect claim & evidence ↗ Required action Critical safety incident report to OES within 15 days of discovery “report any critical safety incident pertaining to one or more of its frontier models to the Office of Emergency Services within 15 days of discovering the critical safety incident.” Informed: Office of Emergency Services SEC. 2, Bus. & Prof. Code § 22757.13(c)(1) Inspect claim & evidence ↗ Required action Annual review of the frontier AI framework “review and, as appropriate, update its frontier AI framework at least once per year.” SEC. 2, Bus. & Prof. Code § 22757.12(b)(1) Inspect claim & evidence ↗ Required action Publish materially modified framework and justification within 30 days “publish the modified frontier AI framework and a justification for that modification within 30 days.” SEC. 2, Bus. & Prof. Code § 22757.12(b)(2) Inspect claim & evidence ↗ Required action Large frontier developers must write, implement, comply with and publish a frontier AI framework “A large frontier developer shall write, implement, comply with, and clearly and conspicuously publish on its internet website a frontier AI framework” SEC. 2, Bus. & Prof. Code § 22757.12(a), opening clause Inspect claim & evidence ↗ Required action Imminent-risk incidents disclosed within 24 hours to an authority appropriate to the incident and as required by law “the frontier developer shall disclose that incident within 24 hours to an authority, including any law enforcement agency or public safety agency with jurisdiction, that is appropriate based on the nature of that incident and as required by law” SEC. 2, Bus. & Prof. Code § 22757.13(c)(2) Inspect claim & evidence ↗ Required action Internal-use catastrophic-risk assessment summaries to OES every three months or on another reasonable schedule “every three months or pursuant to another reasonable schedule specified by the large frontier developer and communicated in writing to the Office of Emergency Services” Informed: Office of Emergency Services SEC. 2, Bus. & Prof. Code § 22757.12(d) Inspect claim & evidence ↗ Required action No materially false or misleading statements about framework implementation or compliance “A large frontier developer shall not make a materially false or misleading statement about its implementation of, or compliance with, its frontier AI framework.” SEC. 2, Bus. & Prof. Code § 22757.12(e)(1)(B) Inspect claim & evidence ↗ Required action Transparency report before or concurrently with deploying a new or substantially modified frontier model “Before, or concurrently with, deploying a new frontier model or a substantially modified version of an existing frontier model, a frontier developer shall clearly and conspicuously publish on its internet website a transparency report” SEC. 2, Bus. & Prof. Code § 22757.12(c)(1) Inspect claim & evidence ↗ Who decides Civil penalties under § 22757.15 recoverable only in a civil action brought by the Attorney General “A civil penalty described in this section shall be recovered in a civil action brought only by the Attorney General.” Decides: Attorney General SEC. 2, Bus. & Prof. Code § 22757.15(b) Inspect claim & evidence ↗ Exception CalCompute section operative only upon appropriation “shall become operative only upon an appropriation in a budget act, or other measure” SEC. 3, Gov. Code § 11546.8(k) Inspect claim & evidence ↗ Exception Incident-reporting compliance via OES-designated federal standards after a declaration of intent “shall declare its intent to do so to the Office of Emergency Services.” Decides: Office of Emergency Services SEC. 2, Bus. & Prof. Code § 22757.13(i)(1) Inspect claim & evidence ↗ Exception Transparency information in a system or model card deemed compliant “including a system card or model card, shall be deemed in compliance with the applicable paragraph.” SEC. 2, Bus. & Prof. Code § 22757.12(c)(3) Inspect claim & evidence ↗ Exact text: SB-53 Artificial intelligence models: large developers. (2025-2026) — bill text, chaptered version (Chapter 138, Statutes of 2025) ↗ Archived copy, captured 2025-09-30: SB-53 Artificial intelligence models: large developers. (2025-2026) — bill text, chaptered version (Chapter 138, Statutes of 2025) ↗ Document page
Michigan HB 4668 (2025) Michigan House Bill 4668 (2025), the artificial intelligence safety and security transparency act
Bill, draft or proposal
Bill (introduced) Publisher: Michigan House of Representatives
Standing shown by: Michigan House Bill 4668, the artificial intelligence safety and security transparency act, was introduced… ↗
Introduced 2025-06-24 Inspect ↗ What the text states · 1 provision recorded Threshold or trigger Definition of "large developer" (Sec. 3(f)), as introduced “A foundation model with a quantity of computing power that costs not less than $5,000,000.00” Sec. 3(f)(i) Inspect claim & evidence ↗ Exact text: Michigan House Bill No. 4668 (2025), as introduced: artificial intelligence safety and security transparency act ↗ Archived copy, captured 2025-10-05: Michigan House Bill No. 4668 (2025), as introduced: artificial intelligence safety and security transparency act ↗ Document page
NY RAISE Act, Chapter 699 (2025) New York S6953-B (2025), the RAISE Act, signed as Chapter 699 of 2025
Law requiring a framework Later version recorded Record 1 ↗ Record 2 ↗
Bill (chaptered) Publisher: New York State Legislature
Standing shown by: S6953-B was delivered to the Governor of New York on 9 December 2025 and signed as Chapter 699 of 2025 on 19… ↗ · S6953-B requires the large developer of a frontier model, before deploying it, to implement a written safety… ↗
Passed a vote 2025-06-12 Record 1 ↗ Record 2 ↗ Signed 2025-12-19 · signature Inspect ↗ What the text states · 5 provisions recorded Threshold or trigger Definition of "frontier model" in the 2025 text (proposed GBL § 1420(6)) “an artificial intelligence model trained using greater than” Proposed GBL § 1420(6)(a), p. 2 Inspect claim & evidence ↗ Threshold or trigger Definition of "large developer" in the 2025 text (proposed GBL § 1420(9)) "Large developer" means a person that has trained at least one frontier model and has spent over one hundred million dollars in compute costs in aggregate in training frontier models.” Proposed GBL § 1420(9), p. 3 Inspect claim & evidence ↗ Threshold or trigger Civil penalty up to $10 million / $30 million (proposed GBL § 1422(1)(a)) “a civil penalty in an amount not exceeding ten million dollars for a first violation and in an amount not exceeding thirty million dollars for any subsequent violation” Proposed GBL § 1422(1)(a), p. 5 Inspect claim & evidence ↗ Required action Written safety and security protocol before deployment (proposed GBL § 1421(1)) “Before deploying a frontier model, the large developer of such frontier model shall do all of the following: (a) Implement a written safety and security protocol” Proposed GBL § 1421(1), p. 4 Inspect claim & evidence ↗ Required action Safety incident disclosure within 72 hours (proposed GBL § 1421(4)) “A large developer shall disclose each safety incident affecting the frontier model to the attorney general and division of homeland security and emergency services within seventy-two hours” Proposed GBL § 1421(4), p. 4 Inspect claim & evidence ↗ Exact text: New York Senate Bill S6953-B (2025–2026): AN ACT to amend the general business law, in relation to the training and use of artificial intelligence frontier models ↗ Document page
New York RAISE Act New York RAISE Act (Responsible AI Safety and Education Act), as described by the Governor
Known from an announcement
Statute (announced by the Governor; amended by Chapter 96 of 2026)
Standing shown by: The Governor of New York announced on 19 December 2025 that she had signed legislation requiring AI… ↗
Signing announced 2025-12-19 · announcement Inspect ↗ Framework editions that name it Microsoft · Frontier Governance Framework, Feb 2026 ↗ States that it responds to this text No provision of this text is recorded. That does not mean it has none.
Exact text: Governor Hochul Signs Nation-Leading Legislation to Require AI Frameworks for AI Frontier Models ↗ Document page
New York RAISE Act, Chapter 96 (2026) New York Senate Bill S8828 / Assembly Bill A9449 (2026): RAISE Act chapter amendment, signed as Chapter 96
Law requiring a framework
Statute (chapter amendment) Publisher: New York State Legislature
Standing shown by: The New York State Assembly’s bill record for S8828 (the same as A9449) lists the bill, which repeals and… ↗ · Section 1421(1) of the General Business Law, as added by the bill, provides that a large frontier developer… ↗
Signed 2026-03-27 · signature Inspect ↗ Takes effect: date stated in the text 2027-01-01 · date set by the text · effective · after the latest check Inspect ↗ Changes from the earlier version · 2 recorded Inferred document comparison Chapter 96 of 2026 replaces the 2025 definition of "frontier model" (a model trained with more than 10^26 operations whose compute cost exceeds one hundred million dollars, or a distillation of one whose own compute cost exceeds five million dollars) with a foundation model trained using more than 10^26 integer or floating-point operations, without the compute-cost condition. 2026-03-27 · signature · Inferred · Supported Inspect claim & evidence ↗ Inferred document comparison Chapter 96 of 2026 replaces the 2025 "large developer" category (over one hundred million dollars spent in aggregate compute costs on frontier models) with a "large frontier developer" category defined by annual gross revenues, together with affiliates, above five hundred million dollars in the preceding calendar year. 2026-03-27 · signature · Inferred · Supported Inspect claim & evidence ↗ What the text states · 4 provisions recorded Threshold or trigger Civil penalty up to $1,000,000 for a first violation and $3,000,000 per subsequent violation “The attorney general may bring a civil action to recover a civil penalty in an amount not to exceed one million dollars for a first violation and in an amount not to exceed three million dollars per subsequent violation” Decides: Attorney general (civil action) Section 1427, subdivision 1 Inspect claim & evidence ↗ Threshold or trigger The 2025 chapter takes effect on 1 January 2027 “This act shall take effect [on the ninetieth day after it shall have become a law] January 1, 2027.” Section 3 of the bill (text of amended section 3) Inspect claim & evidence ↗ Threshold or trigger Report a critical safety incident to the office within 72 hours “a frontier developer shall report any critical safety incident pertaining to one or more of its frontier models to the office within seventy-two hours from a determination that a critical safety incident has occurred” Section 1422, subdivision 3, paragraph (a) Inspect claim & evidence ↗ Required action A large frontier developer must write, implement, comply with and publish a frontier AI framework “A large frontier developer shall write, implement, comply with, and clearly and conspicuously publish on its internet website a frontier AI framework that applies to the large frontier developer's frontier models” Section 1421, subdivision 1 Inspect claim & evidence ↗ Exact text: New York S08828 (Gounardes) / A09449 (Bores): summary, actions and text of the bill (New York State Assembly bill search) ↗ Document page
Illinois AI Safety Measures Act Known from an announcement
Statute (as described by the Governor)
Standing shown by: Illinois Executive Order 2026-07 (22 September 2026) states that the Illinois General Assembly passed, and… ↗
Signing announced 2026-09-22 · announcement Inspect ↗ No provision of this text is recorded. That does not mean it has none.
Document page
United Kingdom UK–US AI safety memorandum (2024) UK–US Memorandum of Understanding on the science of AI safety (April 2024), as announced
Known from an announcement
Memorandum of understanding (text not inspected)
Standing shown by: A GOV.UK press release of 2 April 2024 announces that the UK and US signed a Memorandum of Understanding, on… ↗
Signing announced 2024-04-02 · announcement Inspect ↗ No provision of this text is recorded. That does not mean it has none.
Document page
King's Speech (2024) The King's Speech 2024
Declaration or statement
Speech from the Throne (UK) Publisher: Prime Minister's Office, 10 Downing Street
Standing shown by: The King's Speech of 17 July 2024 states that the government will seek to establish the appropriate… ↗
Published 2024-07-17 · document record What the text states · 1 provision recorded Required action The government will seek to establish appropriate legislation to place requirements on those working to develop the most powerful AI models. “It will seek to establish the appropriate legislation to place requirements on those working to develop the most powerful artificial intelligence models.” Speech text, after the Employment Rights Bill item Inspect claim & evidence ↗ Exact text: The King's Speech 2024 ↗ Document page
UK–Anthropic memorandum (2025) Memorandum of Understanding between the UK and Anthropic on AI opportunities
Voluntary
Memorandum of understanding (UK government) Publisher: Department for Science, Innovation and Technology
Standing shown by: The UK-Anthropic Memorandum of Understanding states that it is voluntary, not legally binding and without… ↗
Signed 2025-02-13 · signature Inspect ↗ Published 2025-02-14 · publication Inspect ↗ What the text states · 2 provisions recorded Required action DSIT and Anthropic, through Anthropic's partnership with the AI Security Institute, will continue research collaboration on the capabilities and security risks of AI systems. Section (II), third paragraph Inspect claim & evidence ↗ Exception The memorandum is voluntary, not legally binding and without prejudice to any binding agreements. “This memorandum is voluntary, not legally binding, and without prejudice to any binding agreements.” Section (III), final paragraph Inspect claim & evidence ↗ Exact text: Memorandum of Understanding between UK and Anthropic on AI opportunities ↗ Document page
UK AI Security Institute renaming release (2025) Tackling AI security risks to unleash growth and deliver Plan for Change (GOV.UK press release, 14 February 2025)
Declaration or statement
Press release (UK government)
Standing shown by: A GOV.UK press release of 14 February 2025 announces that the UK's AI Safety Institute becomes the 'UK AI… ↗
Announced 2025-02-14 · announcement Record 1 ↗ Record 2 ↗ No provision of this text is recorded. That does not mean it has none.
Exact text: Tackling AI security risks to unleash growth and deliver Plan for Change ↗ Document page
UK–Cohere memorandum (2025) Memorandum of Understanding between the UK and Cohere on AI opportunities
Voluntary
Memorandum of understanding (UK government) Publisher: Department for Science, Innovation and Technology
Standing shown by: The UK-Cohere Memorandum of Understanding states that it is voluntary, not legally binding and without… ↗
Published 2025-06-16 · publication Inspect ↗ What the text states · 1 provision recorded Exception The memorandum is voluntary, not legally binding and without prejudice to any binding agreements. “This memorandum is voluntary, not legally binding, and without prejudice to any binding agreements.” Final paragraph Inspect claim & evidence ↗ Exact text: Memorandum of Understanding between UK and Cohere on AI opportunities ↗ Document page
UK–OpenAI memorandum (2025) Memorandum of Understanding between the UK and OpenAI on AI opportunities
Voluntary
Memorandum of understanding (UK government) Publisher: Department for Science, Innovation and Technology
Standing shown by: The UK-OpenAI Memorandum of Understanding states that it is voluntary, not legally binding and without… ↗
Published 2025-07-21 · publication Inspect ↗ What the text states · 2 provisions recorded Required action Collaboration may include expanding the existing OpenAI-AI Security Institute partnership with a new technical information sharing programme. “expanding the existing partnership between OpenAI and the UK AI Security Institute to include the development of a new technical information sharing programme” Section 4: technical information exchange Inspect claim & evidence ↗ Exception The memorandum is voluntary, not legally binding and without prejudice to any binding agreements. Final paragraph Inspect claim & evidence ↗ Exact text: Memorandum of Understanding between UK and OpenAI on AI opportunities ↗ Document page
UK–Google DeepMind memorandum (2025) Memorandum of Understanding between the UK and Google DeepMind on AI opportunities and security
Memorandum
Memorandum of understanding (UK government) Publisher: Department for Science, Innovation and Technology
Standing shown by: The Department for Science, Innovation and Technology published on GOV.UK on 11 December 2025 a Memorandum… ↗ · The UK-Google DeepMind Memorandum of Understanding states that it takes effect from the date of the last… ↗
Published 2025-12-11 · publication Inspect ↗ What the text states · 2 provisions recorded Threshold or trigger Takes effect from the date of the last signature and remains active for 36 months. “The MoU takes effect from the date of the last signature and will remain active for 36 months.” General terms Inspect claim & evidence ↗ Required action Google DeepMind will provide the AI Security Institute with priority technical access to its frontier models and enhanced technical information exchange. “with Google DeepMind providing AISI with priority technical access to its frontier models” Section 3: Advance AI security and preparedness Inspect claim & evidence ↗ Exact text: Memorandum of Understanding between the UK and Google DeepMind on AI opportunities and security ↗ Document page
King's Speech 2026 briefing notes The King's Speech 2026: background briefing notes
Declaration or statement
Briefing notes (UK government) Publisher: Prime Minister's Office, 10 Downing Street
Standing shown by: Within the inspected scope, a case-insensitive search of the King's Speech 2026 background briefing notes… ↗
Published 2026-05-13 · document record What the text states · 1 provision recorded Not found in the inspected text No occurrence of 'frontier', 'most powerful', 'AI Bill' or 'Artificial Intelligence Bill' in the 2026 background briefing notes Inspected: Whole PDF text (cover dated 13 May 2026) Inspect claim & evidence ↗ Exact text: The King's Speech 2026: background briefing notes ↗ Document page
UK AI defences release (2026) Britain steps up national defences against growing AI risks (GOV.UK press release, 12 October 2026)
Declaration or statement
Press release (UK government)
Standing shown by: A GOV.UK press release of 12 October 2026 states that stronger protections against emerging AI threats are… ↗
Announced 2026-10-12 · announcement Inspect ↗ No provision of this text is recorded. That does not mean it has none.
Exact text: Britain steps up national defences against growing AI risks ↗ Document page
Summits, the G7 and the OECD Being listed under a declaration or statement is not ratification, a signature of a treaty or evidence of implementation. A country absent from a list is recorded only as absent from that list.
G7 Leaders' Statement on the Hiroshima AI Process G7 Leaders' Statement on the Hiroshima AI Process (30 October 2023)
Declaration or statement
Leaders' statement (G7) Publisher: Group of Seven (G7)
Standing shown by: The G7 Leaders' Statement of 30 October 2023 calls on organisations developing advanced AI systems to commit… ↗
Published 2023-10-30 · document record Texts it names, and texts that name it The G7 Leaders' Statement on the Hiroshima AI Process of 30 October 2023 welcomes the International Guiding Principles and the International Code of Conduct for Organizations Developing Advanced AI Systems, attached to it. 2023-10-30 · publication · Stated in source · Supported Inspect claim & evidence ↗ What the text states · 1 provision recorded Required action Calls on organisations developing advanced AI systems to commit to the application of the International Code of Conduct. “We call on organizations developing advanced AI systems to commit to the application of the International Code of Conduct.” First paragraph Inspect claim & evidence ↗ Exact text: G7 Leaders' Statement on the Hiroshima AI Process, October 30, 2023 ↗ Document page
Hiroshima Process Code of Conduct (2023) Hiroshima Process International Code of Conduct for Organizations Developing Advanced AI Systems
Voluntary
Voluntary code of conduct (G7) Publisher: Group of Seven (G7)
Standing shown by: The Hiroshima Process Code of Conduct states that it will provide voluntary guidance for actions by… ↗
Published 2023-10-30 · document record Texts it names, and texts that name it The G7 Leaders' Statement on the Hiroshima AI Process of 30 October 2023 welcomes the International Guiding Principles and the International Code of Conduct for Organizations Developing Advanced AI Systems, attached to it. 2023-10-30 · publication · Stated in source · Supported Inspect claim & evidence ↗ The Seoul Declaration states that its participants continue to focus on supporting the operationalisation of the Hiroshima Process International Code of Conduct for Organizations Developing Advanced AI Systems. 2024-05-21 · publication · Stated in source · Supported Inspect claim & evidence ↗ What the text states · 4 provisions recorded Threshold or trigger Provides voluntary guidance for organisations developing the most advanced AI systems, including the most advanced foundation models and generative AI systems ('advanced AI systems'). “will provide voluntary guidance for actions by organizations developing the most advanced AI systems, including the most advanced foundation models and generative AI systems (henceforth "advanced AI systems").” PDF p. 1, first paragraph Inspect claim & evidence ↗ Required action Action 1: identify, evaluate and mitigate risks across the AI lifecycle, including diverse internal and independent external testing such as red-teaming. “This includes employing diverse internal and independent external testing measures, through a combination of methods for evaluations, such as red-teaming, and implementing appropriate mitigation to address identified risks and vulnerabilities.” PDF p. 2, Action 1 Inspect claim & evidence ↗ Required action Action 3: publicly report advanced AI systems' capabilities, limitations and domains of appropriate and inappropriate use, including transparency reports for all new significant releases. “This should include publishing transparency reports containing meaningful information for all new significant releases of advanced AI systems.” PDF p. 4, Action 3 Inspect claim & evidence ↗ Who decides The G7 commit to develop proposals, with the OECD, GPAI and others, for monitoring tools and mechanisms to help organisations stay accountable for implementing the actions. “We also commit to develop proposals, in consultation with the OECD, GPAI and other stakeholders, to introduce monitoring tools and mechanisms to help organizations stay accountable for the implementation of these actions.” PDF p. 1, introductory text Inspect claim & evidence ↗ Exact text: Hiroshima Process International Code of Conduct for Organizations Developing Advanced AI Systems ↗ Document page
Bletchley Declaration (2023) The Bletchley Declaration by Countries Attending the AI Safety Summit, 1–2 November 2023
Declaration or statement
Intergovernmental declaration Publisher: Department for Science, Innovation and Technology
Standing shown by: The Bletchley Declaration was published on GOV.UK on 1 November 2023 (the page’s “First published” date) by… ↗
Participants listed: 29, counted by Selekoa 2023-11-01 · snapshot Inspect ↗ Published 2023-11-01 · publication Inspect ↗ Signatory added to the text 2025-02-13 · revision Inspect ↗ What the text states · 3 provisions recorded Threshold or trigger Describes the 'frontier' of AI as highly capable general-purpose AI models, including foundation models, and relevant specific narrow AI that could exhibit harmful capabilities, which match or exceed the capabilities of today's most advanced models. “Particular safety risks arise at the ‘frontier’ of AI, understood as being those highly capable general-purpose AI models, including foundation models, that could perform a wide variety of tasks” Fourth paragraph Inspect claim & evidence ↗ Required action Affirms that actors developing frontier AI capabilities have a particularly strong responsibility for ensuring the safety of these systems, including through safety testing and evaluations. “actors developing frontier AI capabilities, in particular those AI systems which are unusually powerful and potentially harmful, have a particularly strong responsibility for ensuring the safety of these AI systems, including through systems for safety testing, through evaluations, and by other appropriate measures” Paragraph beginning 'We affirm that, whilst safety must be considered across the AI lifecycle' Inspect claim & evidence ↗ Required action Includes, within the agenda item on risk-based policies, increased transparency by private actors developing frontier AI capabilities. “alongside increased transparency by private actors developing frontier AI capabilities” Second agenda bullet ('building respective risk-based policies') Inspect claim & evidence ↗ Exact text: The Bletchley Declaration by Countries Attending the AI Safety Summit, 1-2 November 2023 ↗ Document page
Bletchley chair's statement on safety testing (2023) Safety Testing: Chair's Statement of Session Outcomes, 2 November 2023
Declaration or statement
Chair's statement (AI Safety Summit) Publisher: Department for Science, Innovation and Technology
Standing shown by: The chair's statement of session outcomes on safety testing from the AI Safety Summit was published on… ↗
Announced 2023-11-02 · announcement Inspect ↗ Participants listed: 21, counted by Selekoa 2023-11-02 · snapshot Inspect ↗ Published 2023-11-02 · publication Inspect ↗ What the text states · 2 provisions recorded Required action Records the participants' acknowledgement that developers have responsibility to devise and conduct safety testing through evaluations, transparency and other appropriate measures. “Developers both have responsibility to devise and conduct safety testing through evaluations, transparency, and other appropriate measures” Second acknowledged point Inspect claim & evidence ↗ Required action Governments recognise an increasing role in seeing that external evaluations are undertaken for frontier AI models developed within their countries, before and after deployment. “Governments recognise their increasing role for seeing that external evaluations are undertaken for frontier AI models developed within their countries” Third acknowledged point Inspect claim & evidence ↗ Exact text: Safety Testing: Chair's Statement of Session Outcomes, 2 November 2023 ↗ Document page
Seoul Declaration (2024) Seoul Declaration for safe, innovative and inclusive AI (AI Seoul Summit, 21 May 2024)
Declaration or statement
Intergovernmental declaration Publisher: Department for Science, Innovation and Technology
Standing shown by: The Department for Science, Innovation and Technology published the Seoul Declaration for safe, innovative… ↗
Participants listed: 11, counted by Selekoa 2024-05-21 · snapshot Inspect ↗ Published 2024-05-21 · publication Inspect ↗ Texts it names, and texts that name it The Seoul Declaration states that its participants continue to focus on supporting the operationalisation of the Hiroshima Process International Code of Conduct for Organizations Developing Advanced AI Systems. 2024-05-21 · publication · Stated in source · Supported Inspect claim & evidence ↗ The Seoul Declaration recognises the particular responsibility of organisations developing and deploying frontier AI and, in that regard, notes the Frontier AI Safety Commitments. 2024-05-21 · publication · Stated in source · Supported Inspect claim & evidence ↗ No provision of this text is recorded. That does not mean it has none.
Exact text: Seoul Declaration for safe, innovative and inclusive AI by participants attending the Leaders' Session: AI Seoul Summit, 21 May 2024 ↗ Document page
Seoul commitments (2024) Frontier AI Safety Commitments, AI Seoul Summit 2024
Voluntary
Voluntary commitment text Publisher: Department for Science, Innovation and Technology
Standing shown by: The UK Department for Science, Innovation and Technology published the Frontier AI Safety Commitments on… ↗ · The Frontier AI Safety Commitments state that the listed organisations undertake to demonstrate how they… ↗
Published 2024-05-21 · publication Inspect ↗ Framework editions that name it Anthropic · Responsible Scaling Policy, v2.0 ↗ States that it fulfils this text Microsoft · Frontier Governance Framework, Feb 2025 (Version 1) ↗ States that it responds to this text Meta · Frontier AI / Advanced AI Scaling Framework, v1.1 ↗ States that it responds to this text Anthropic · Responsible Scaling Policy, v2.1 ↗ States that it fulfils this text Anthropic · Responsible Scaling Policy, v2.2 ↗ States that it fulfils this text Microsoft · Frontier Governance Framework, Feb 2026 ↗ States that it responds to this text Texts it names, and texts that name it The Seoul Declaration recognises the particular responsibility of organisations developing and deploying frontier AI and, in that regard, notes the Frontier AI Safety Commitments. 2024-05-21 · publication · Stated in source · Supported Inspect claim & evidence ↗ What the text states · 5 provisions recorded Threshold or trigger Thresholds at which severe risks would be deemed intolerable “at which severe risks posed by a model or system, unless adequately mitigated, would be deemed intolerable” Outcome 1, commitment II, first sentence Inspect claim & evidence ↗ Required action In the extreme, no development or deployment if risks cannot be kept below thresholds “commit not to develop or deploy a model or system at all, if mitigations cannot be applied to keep risks below the thresholds” Outcome 1, commitment IV, final sentence Inspect claim & evidence ↗ Required action Publish a safety framework focused on severe risks by the AI Summit in France “publishing a safety framework focused on severe risks by the upcoming AI Summit in France” Paragraph after the lists of organisations, beginning “The above organisations, in furtherance of safe and trustworthy AI” Inspect claim & evidence ↗ Who decides Thresholds defined with input from trusted actors, including home governments “defined with input from trusted actors, including organisations’ respective home governments as appropriate” Consulted: Trusted actors, including home governments as appropriate Outcome 1, commitment II, third sentence Inspect claim & evidence ↗ Exception Public transparency, except where disclosure would increase risk or reveal sensitive commercial information disproportionately “except insofar as doing so would increase risk or divulge sensitive commercial information to a degree disproportionate to the societal benefit” Outcome 3, commitment VII, first sentence Inspect claim & evidence ↗ 20 signatory records on the commitments page → Exact text: Frontier AI Safety Commitments, AI Seoul Summit 2024 ↗ Archived copy, captured 2024-05-21: Frontier AI Safety Commitments, AI Seoul Summit 2024 ↗ Document page
Seoul Ministerial Statement (2024) Seoul Ministerial Statement for advancing AI safety, innovation and inclusivity (22 May 2024)
Declaration or statement
Ministerial statement (AI Seoul Summit) Publisher: Department for Science, Innovation and Technology
Standing shown by: The Department for Science, Innovation and Technology published the Seoul Ministerial Statement for… ↗
Participants listed: 28, counted by Selekoa 2024-05-22 · snapshot Inspect ↗ Published 2024-05-22 · publication Inspect ↗ What the text states · 4 provisions recorded Threshold or trigger Recognises that severe risks could be posed by a model or system capability to meaningfully assist non-state actors with chemical or biological weapons. “the potential model or system capability to meaningfully assist non-state actors in advancing the development, production, acquisition or use of chemical or biological weapons” Safety, paragraph 1.3 Inspect claim & evidence ↗ Threshold or trigger Acknowledges the signatories' role, in partnership with others, in identifying thresholds at which frontier AI risks would be severe without appropriate mitigations. “identifying thresholds at which the risks posed by the design, development, deployment and use of frontier AI models or systems would be severe without appropriate mitigations” Safety, paragraph 1.2 Inspect claim & evidence ↗ Required action Reaffirms the particular responsibility of developers for the safety of frontier AI models or systems. “reaffirming the particular responsibility of developers for the safety of these systems” Safety, paragraph 1.5 Inspect claim & evidence ↗ Required action Plans collaboration to identify severe-risk thresholds and define frontier capabilities that could pose severe risks, with the ambition of developing proposals before the AI Action Summit in France. “with the ambition of developing proposals for consideration in advance of the AI Action Summit in France” Way forward, paragraph W3 Inspect claim & evidence ↗ Exact text: Seoul Ministerial Statement for advancing AI safety, innovation and inclusivity: AI Seoul Summit 2024 ↗ Document page
HAIP Reporting Framework Hiroshima AI Process Reporting Framework
Voluntary
Voluntary reporting framework (OECD)
Standing shown by: The OECD.AI page on the Hiroshima AI Process Reporting Framework states that the framework was launched on 7… ↗
Announced 2025-02-07 · announcement Inspect ↗ Participants listed: 31 2026-10-02 · publication Inspect ↗ No provision of this text is recorded. That does not mean it has none.
Document page
Paris AI Action Summit statement (web page) Statement on Inclusive and Sustainable Artificial Intelligence for People and the Planet (Élysée web page)
Declaration or statement
Summit statement (web expression)
Standing shown by: The Élysée posted the English text of the Statement on Inclusive and Sustainable Artificial Intelligence for… ↗
Published 2025-02-11 · publication Record 1 ↗ Record 2 ↗ Participants listed: 65, counted by Selekoa 2026-10-27 · snapshot Inspect ↗ No provision of this text is recorded. That does not mean it has none.
Exact text: Statement on Inclusive and Sustainable Artificial Intelligence for People and the Planet ↗ Document page
New Delhi Declaration (2026) AI Impact Summit Declaration (New Delhi Declaration on AI Impact), February 2026
Declaration or statement
Summit declaration Publisher: Ministry of Electronics and Information Technology, Government of India
Standing shown by: A Press Information Bureau release of the Ministry of Electronics & IT, posted on 21 February 2026, states… ↗ · The PIB text of the New Delhi Declaration as served on 2026-10-27 lists 92 endorsing countries and… ↗
Participants listed: 88 2026-02-21 · snapshot Inspect ↗ Published 2026-02-21 · publication Inspect ↗ Participants listed: 91 2026-02-24 · snapshot Inspect ↗ Participants listed: 92, counted by Selekoa 2026-03-05 · snapshot Inspect ↗ What the text states · 2 provisions recorded Required action Under 'Secure and Trusted AI', recognises the importance of security in AI systems, industry-led voluntary measures, technical solutions and appropriate policy frameworks. “we recognize the importance of security in AI systems, industry-led voluntary measures, and the adoption of technical solutions, and appropriate policy frameworks” Annexure, 'Secure and Trusted AI', first paragraph Inspect claim & evidence ↗ Not found in the inspected text No occurrence of 'frontier', 'safety', 'Bletchley', 'Seoul' or 'Paris' in the annexed declaration and release text Inspected: Whole cached text (release and Annexure) Inspect claim & evidence ↗ Exact text: AI Impact Summit 2026 Concludes with Adoption of New Delhi Declaration (with the declaration annexed) ↗ Document page