2025: governance
Selekoa’s 516 governance records dated 2025, in date order, of 772 in all topics. 2 stated dates are listed apart.
Selective: what the records hold, not a history of the year.
How this page is made
- Each record keeps the kind of date its source gives: an event, a publication, an announcement, an effective date, a revision, a signature, a snapshot, or the start or end of an interval. A snapshot is one dated report of a state.
- A record dated only to a month sits under that month, before the records dated to a day. A record dated only to the year sits under the year, before the months. No day or month is added.
- An announcement records what a party stated. It does not show that anything was done, built or paid for.
- Dates a party states for a year are listed apart, after the records: targets, deadlines and terms, and dates stated in passing, such as a reported start of operation. So is a record’s own date that falls after its sources were last checked, such as a law’s date of effect. A stated date given only in words, with no year, is not listed.
- Records with an unknown date are not listed. A record with an interval appears in the year it starts and in the year it ends, or once, with both dates, when both fall in one year.
- Each provision of a text is a record of its own. Provisions sharing a text and a date are shown here as one entry, with their number, and listed on the text’s page; a contested or unresolved one is listed here on its own. The counts on this page include each provision record.
- The page is generated from the records, with no selection or framing by an editor. Coverage is uneven across years and topics: a year or topic with few records may be one the records barely cover.
Governance 516 records
January 2025
· announcement · dated to the day
A CAC announcement reposted on gov.cn on 10 January 2025 states that, as of 31 December 2024, 302 generative AI services had completed filing with the CAC, 238 of them added in 2024, and that 105 applications or functions calling filed models had registered with local cyberspace offices in 2024. Evidence ↗· publication · dated to the day
1 provision record of AI Opportunities Action Plan
· signature · dated to the day
Executive Order 14141, "Advancing United States Leadership in Artificial Intelligence Infrastructure", is dated 14 January 2025 and was published in the Federal Register on 17 January 2025 (FR Doc. 2025-01395). Evidence ↗· snapshot · dated to the day
OpenAI announced on 14 January 2025 that Adebayo Ogunlesi had joined its board of directors. Evidence ↗· publication · dated to the day
MiniMax published MiniMax-Text-01 and MiniMax-VL-01 under a “MiniMax Model License Agreement”, present as the repository’s LICENSE file at a commit of 15 January 2025. Evidence ↗· publication · dated to the day
The Bureau of Industry and Security published the interim final rule "Framework for Artificial Intelligence Diffusion" in the Federal Register of 15 January 2025, stating an effective date of 13 January 2025 and a general compliance date of 15 May 2025. Evidence ↗· publication · dated to the day
NIST states that the second public draft of NIST AI 800-1 revises the initial public draft, adding appendices on measuring misuse risk and on chemical/biological and cybersecurity misuse risk and content on open models and the AI supply chain. Evidence ↗· publication · dated to the day
On 15 January 2025 the U.S. AI Safety Institute released the second public draft of NIST AI 800-1, "Managing Misuse Risk for Dual-Use Foundation Models", for public comment through 15 March 2025. Evidence ↗· publication · dated to the day
3 provision records of MiniMax Model License Agreement (MiniMax-01)
· publication · dated to the day
1 provision record of NIST AI 800-1 (second public draft): Managing Misuse Risk for Dual-Use Foundation Models
· publication · dated to the day
1 provision record of Framework for Artificial Intelligence Diffusion (interim final rule)
· publication · dated to the day
The January 2025 foundry due-diligence rule names the AI diffusion rule, which BIS filed for public inspection on 13 January 2025, as “a rule related to this IFR”. Evidence ↗· publication · dated to the day
The Bureau of Industry and Security published the interim final rule "Implementation of Additional Due Diligence Measures for Advanced Computing Integrated Circuits" in the Federal Register of 16 January 2025, effective that day, with compliance required from 31 January 2025, to assist foundries and OSAT companies in complying with the advanced computing IC provisions. Evidence ↗· publication · dated to the day
1 provision record of Implementation of Additional Due Diligence Measures for Advanced Computing Integrated Circuits (interim final rule)
· publication · dated to the day
2 provision records of Executive Order 14141: Advancing United States Leadership in Artificial Intelligence Infrastructure
· signature · dated to the day
Section 2 of Executive Order 14148 (20 January 2025) lists Executive Order 14110 of 30 October 2023 among the executive actions it declares "hereby revoked". Evidence ↗· signature · dated to the day
Executive Order 14148, "Initial Rescissions of Harmful Executive Orders and Actions", is dated and signed at the White House on 20 January 2025 and was published in the Federal Register on 28 January 2025. Evidence ↗· publication · dated to the day
DeepSeek’s pinned R1 README states MIT licensing for code and weights, with upstream model qualifications. Evidence ↗· signature · dated to the day
Executive Order 14179, "Removing Barriers to American Leadership in Artificial Intelligence", is dated 23 January 2025 and was published in the Federal Register on 31 January 2025. Evidence ↗· publication · dated to the day
1 provision record of Executive Order 14179: Removing Barriers to American Leadership in Artificial Intelligence
February 2025
· publication · dated to the month
Microsoft’s Frontier Governance Framework, Version 1 (February 2025), states that the framework has its genesis in the voluntary Frontier AI Safety Commitments that Microsoft and fifteen other AI labs made in May 2024. Evidence ↗· publication · dated to the month
Microsoft published the first edition of its Frontier Governance Framework in February 2025. Evidence ↗· publication · dated to the month
27 provision records of Microsoft Frontier Governance Framework — Version 1 (February 2025)
· publication · dated to the day
Meta’s Frontier AI Framework v1.1 opens its scope section by stating that, in line with the Frontier AI Safety Commitments, which it says Meta signed in May 2024, the framework relates to Meta’s forthcoming most advanced models and systems. Evidence ↗· publication · dated to the day
Meta published its Frontier AI Framework, Version 1.1, identified by its announcement date of 2025-02-03. Evidence ↗· publication · dated to the day
26 provision records of Meta Frontier AI Framework — Version 1.1
· publication · dated to the day Inferred
Where the Frontier Safety Framework v1.0 set two Critical Capability Levels in a Biosecurity risk domain, one for enabling a non-expert and one for enabling an expert, v2.0 sets a single level, CBRN uplift 1, in a combined chemical, biological, radiological and nuclear risk domain. Evidence ↗· publication · dated to the day
FSF v2.0 introduces recommended security levels for Critical Capability Levels. Evidence ↗· publication · dated to the day
Google DeepMind published Frontier Safety Framework v2.0, identified by its publication date of 2025-02-04. Evidence ↗· publication · dated to the day
33 provision records of Google DeepMind Frontier Safety Framework — v2.0
· publication · dated to the day
Frontier Safety Framework v2.0 replaces v1.0 in the publisher’s documented framework version sequence. Evidence ↗· revision · dated to the day
Magic is listed, as “Magic”, among four organisations added on 7 February 2025 to the GOV.UK list of organisations that have agreed to the Frontier AI Safety Commitments. Evidence ↗· revision · dated to the day
MiniMax is listed, as “Minimax”, among four organisations added on 7 February 2025 to the GOV.UK list of organisations that have agreed to the Frontier AI Safety Commitments. Evidence ↗· revision · dated to the day
NVIDIA is listed, as “NVIDIA”, among four organisations added on 7 February 2025 to the GOV.UK list of organisations that have agreed to the Frontier AI Safety Commitments. Evidence ↗· revision · dated to the day
01.AI is listed, as “01.ai”, among four organisations added on 7 February 2025 to the GOV.UK list of organisations that have agreed to the Frontier AI Safety Commitments. Evidence ↗· announcement · dated to the day
The OECD.AI page on the Hiroshima AI Process Reporting Framework states that the framework was launched on 7 February 2025 and provides a standardised structure for organisations to report on their alignment with the Hiroshima Process Code of Conduct. Evidence ↗· publication · dated to the day · announced
The European Commission announced InvestAI at the AI Action Summit in Paris on 11 February 2025, an initiative to mobilise €200 billion for investment in AI including a new European fund of €20 billion for AI gigafactories. Evidence ↗· publication · dated to the day
The Paris statement commends the Bletchley Park and Seoul summits and notes 'the voluntary commitments launched there', without naming any commitment document. Evidence ↗· publication · dated to the day
The Élysée posted the English text of the Statement on Inclusive and Sustainable Artificial Intelligence for People and the Planet on elysee.fr on 11 February 2025. Evidence ↗· publication · dated to the day
1 provision record of EU launches InvestAI initiative to mobilise €200 billion of investment in artificial intelligence (IP/25/467)
· revision · dated to the day
The GOV.UK text of the Bletchley Declaration, updated on 13 February 2025, states that New Zealand joined the commitment to the declaration on 23 October 2024. Evidence ↗· signature · dated to the day
The GOV.UK text of the UK-Anthropic Memorandum of Understanding names Dario Amodei, CEO of Anthropic, and Peter Kyle, Secretary of State, as signatories, dated 13 February 2025. Evidence ↗· announcement · dated to the day
A GOV.UK press release of 14 February 2025 announces that the UK's AI Safety Institute becomes the 'UK AI Security Institute'. Evidence ↗· publication · dated to the day
The Department for Science, Innovation and Technology published the text of its Memorandum of Understanding with Anthropic on AI opportunities on GOV.UK on 14 February 2025. Evidence ↗· publication · dated to the day
1 provision record of Tackling AI security risks to unleash growth and deliver Plan for Change (GOV.UK press release, 14 February 2025)
· publication · dated to the day
2 provision records of Memorandum of Understanding between the UK and Anthropic on AI opportunities
· snapshot · dated to the day
The Élysée PDF of the Paris statement lists 64 numbered entries under 'Signatory countries and organisations': 62 countries, the African Union Commission and the European Union. Zimbabwe, listed on the web page, is not among them, and neither are the United Kingdom and the United States. Evidence ↗· publication · dated to the day
The national standards catalogue lists GB 45438-2025, Cybersecurity technology — Labeling method for content generated by artificial intelligence, as a mandatory national standard published by SAMR and the Standardization Administration on 28 February 2025, with an implementation date of 1 September 2025. Evidence ↗
March 2025
· snapshot · dated to the day
The UK Competition and Markets Authority states that under the 2023 LLC Agreement the strategic direction, management and day-to-day operation of OpenAI OpCo LLC are controlled by OpenAI GP LLC as its Manager, which is in turn wholly owned and controlled by OpenAI Nonprofit, and that members’ voting rights are restricted to matters related to typical financial investor protections. Evidence ↗· publication · dated to the day
The Cyberspace Administration of China, the Ministry of Industry and Information Technology, the Ministry of Public Security and the National Radio and Television Administration issued the Measures for Labelling AI-Generated Synthetic Content by a notice (国信办通字〔2025〕2号) dated 7 March 2025. Evidence ↗· publication · dated to the day
7 provision records of Measures for Labelling Artificial Intelligence-Generated Synthetic Content (人工智能生成合成内容标识办法)
· effective · dated to the day
The RSP v2.1 change log states that the update adds a new capability threshold related to CBRN development, defining capabilities that could substantially uplift the development capabilities of moderately resourced state programs. Evidence ↗· effective · dated to the day
Anthropic’s Responsible Scaling Policy v2.1 states that the policy also helps satisfy Anthropic’s Frontier AI Safety Commitments (2024). Evidence ↗· effective · dated to the day
Anthropic’s Responsible Scaling Policy v2.1 states that the policy also helps satisfy Anthropic’s Voluntary White House Commitments (2023). Evidence ↗· effective · dated to the day
Anthropic published Responsible Scaling Policy v2.1, identified by its effective date of 2025-03-31. Evidence ↗· effective · dated to the day
29 provision records of Anthropic Responsible Scaling Policy — v2.1
· effective · dated to the day
RSP v2.1 separates AI R&D thresholds for automating entry-level research and dramatically accelerating effective scaling. Evidence ↗· effective · dated to the day
Responsible Scaling Policy v2.1 replaces v2.0 in the publisher’s documented framework version sequence. Evidence ↗
April 2025
· snapshot · dated to the day
Google DeepMind's post of 2 April 2025 describes its AGI Safety Council as analysing AGI risk and best practices and making recommendations on safety measures. Evidence ↗· announcement · dated to the day
DSIT and DESNZ announced that the AI Energy Council, co-chaired by the Technology and Energy Secretaries, held its first meeting on 8 April 2025, expected to advise on energy efficiency and sustainability in AI and data-centre infrastructure. Evidence ↗· revision · dated to the day ContestedQuestion posed by Selekoa
Preparedness Framework v2’s governance arrangements permit OpenAI Leadership to deploy models with High capabilities without safeguards that sufficiently minimize their associated risks. Evidence ↗· revision · dated to the day
OpenAI published Preparedness Framework Version 2, identified by its revision date of 2025-04-15. Evidence ↗· revision · dated to the day
Preparedness Version 2 moves handling of persuasion risks outside the Preparedness Framework. Evidence ↗· revision · dated to the day
Preparedness Version 2 introduces Research Categories for capabilities requiring further threat-model and measurement work. Evidence ↗· revision · dated to the day
Preparedness Version 2 removes the terms “low” and “medium” while defining High and Critical capability thresholds. Evidence ↗· revision · dated to the day
31 provision records of OpenAI Preparedness Framework — Version 2
· revision · dated to the day UnresolvedInferred
Preparedness Framework Version 2 replaces Beta in the publisher’s documented framework version sequence. Evidence ↗· publication · dated to the day
The national standards catalogue lists GB/T 45654-2025, Cybersecurity technology — Basic security requirements for generative artificial intelligence service, as published by SAMR and the Standardization Administration on 25 April 2025, with an implementation date of 1 November 2025. Evidence ↗· publication · dated to the day
The Virology Capabilities Test paper states that it presents a large language model benchmark measuring the capability to troubleshoot complex virology laboratory protocols, built from the inputs of dozens of expert virologists. Evidence ↗Virology Capabilities Test (VCT): A Multimodal Virology Q&A Benchmark (arXiv:2504.16137) →Claim page →
· revision · dated to the day
The Virology Capabilities Test paper states that expert virologists with internet access scored an average of 22.1% on the benchmark's questions within their own sub-areas of expertise, while the most performant model it tested, OpenAI's o3, reached 43.8% accuracy and outperformed 94% of those experts within their sub-areas. Evidence ↗Virology Capabilities Test (VCT): A Multimodal Virology Q&A Benchmark (arXiv:2504.16137) →Claim page →
· publication · dated to the day
Alibaba’s Qwen3 announcement describes downloadable models and API access still to come. Evidence ↗· publication · dated to the day
Qwen Team’s Chinese Qwen3 announcement places its eight listed model weights under Apache 2.0. Evidence ↗
May 2025
· interval begins · dated to the month
Anthropic's Long-Term Benefit Trust announced on 7 June 2025 the appointment of Richard Fontaine, CEO of the Center for a New American Security, as a member of the Trust; Anthropic's Trust page states that he joined in May 2025. Evidence ↗· signature · dated to the day
Executive Order 14292, Improving the Safety and Security of Biological Research, is dated 5 May 2025 in the Federal Register print of the order. Evidence ↗· event · dated to the day
OpenAI published a plan to convert its for-profit LLC into a public benefit corporation under nonprofit control. Evidence ↗· announcement · dated to the day
OpenAI announced on 5 May 2025 that its board had decided that the nonprofit would retain control of OpenAI, and that its for-profit LLC would transition to a public benefit corporation that the nonprofit would control and in which it would be a large shareholder. Evidence ↗· publication · dated to the day
2 provision records of Executive Order 14292: Improving the Safety and Security of Biological Research
· publication · dated to the day
The Bureau of Industry and Security issued guidance dated 13 May 2025 warning that use of PRC advanced-computing ICs, including specific Huawei Ascend chips, risks violating US export controls under General Prohibition 10, because such chips were likely developed or produced in violation of those controls. Evidence ↗· announcement · dated to the day · announced
On 13 May 2025 the Department of Commerce announced a rescission of the AI Diffusion Rule and said that BIS planned to publish a regulation formalising the rescission. The linked release PDF, dated 12 May 2025, says that the Department initiated a rescission. Evidence ↗· publication · dated to the day
1 provision record of Guidance on Application of General Prohibition 10 (GP10) to People’s Republic of China (PRC) Advanced-Computing Integrated Circuits (ICs)
· effective · dated to the day
The RSP v2.2 change log gives, as one of the two reasons it states for excluding sophisticated and state-compromised insiders from the ASL-3 Security Standard, that the CBRN-3 threat models entail large numbers of users having access to unguarded models, which it says is more likely to occur through a universal jailbreak than via model theft. Evidence ↗· effective · dated to the day
Anthropic’s Responsible Scaling Policy v2.2 states that the policy also helps satisfy Anthropic’s Frontier AI Safety Commitments (2024). Evidence ↗· effective · dated to the day
Anthropic’s Responsible Scaling Policy v2.2 states that the policy also helps satisfy Anthropic’s Voluntary White House Commitments (2023). Evidence ↗· effective · dated to the day
Anthropic published Responsible Scaling Policy v2.2, identified by its effective date of 2025-05-14. Evidence ↗· effective · dated to the day
25 provision records of Anthropic Responsible Scaling Policy — v2.2
· effective · dated to the day
RSP v2.2 excludes sophisticated insiders and state-compromised insiders from the ASL-3 Security Standard. Evidence ↗· effective · dated to the day
Responsible Scaling Policy v2.2 replaces v2.1 in the publisher’s documented framework version sequence. Evidence ↗· publication · dated to the day
Anthropic's ASL-3 activation post states that it had not yet determined whether Claude Opus 4 definitively passed the capabilities threshold requiring ASL-3 protections, that continued improvements in CBRN-related knowledge and capabilities meant clearly ruling out ASL-3 risks was not possible as it had been for every previous model, and that it deployed the model with ASL-3 measures as a precautionary and provisional action. Evidence ↗· publication · dated to the day
Anthropic published 'Activating AI Safety Level 3 protections' on 22 May 2025, stating that it had activated the ASL-3 Deployment and Security Standards of its Responsible Scaling Policy in conjunction with launching Claude Opus 4. Evidence ↗· publication · dated to the day
Anthropic's ASL-3 activation post states that experiments showed access to Claude Sonnet 3.7 helped participants do somewhat better on tasks related to CBRN weapon acquisition than those with standard internet access, and that all participants' plans still had critical failures. Evidence ↗· publication · dated to the day
1 provision record of Activating AI Safety Level 3 protections
· snapshot · dated to the day
Anthropic announced on 28 May 2025 that Reed Hastings had been appointed to its board of directors by its Long-Term Benefit Trust. Evidence ↗· publication · dated to the day
2 provision records of Executive Order 14299: Deploying Advanced Nuclear Reactor Technologies for National Security
June 2025
· publication · dated to the day
1 provision record of Department of Commerce bulletin of 4 June 2025 on the Center for AI Standards and Innovation
· snapshot · dated to the day
The Seed-Coder-8B-Instruct model card at its revision of 6 June 2025 states that the project is licensed under the MIT License; its revisions up to 6 May 2025 had carried the licence metadata apache-2.0. Evidence ↗· event · dated to the day
The New York State Assembly passed S6953-B, substituted for A6453-B, on 12 June 2025. Evidence ↗· event · dated to the day
The New York State Senate passed S6953-B (the RAISE Act) on 12 June 2025. Evidence ↗· snapshot · dated to the day
Scale AI stated on 12 June 2025 that Alexandr Wang would continue to serve as a director on Scale's board. Evidence ↗· publication · dated to the day
The MiniMax-M1 GitHub repository carries the Apache License, Version 2.0, with a 2025 MiniMax copyright notice, as of a commit of 14 June 2025. Evidence ↗· snapshot · dated to the day
The OECD paper 'How are AI developers managing risks?' (September 2025) lists, in Annex A, 20 entities that submitted reports under the HAIP reporting framework between 7 February and 15 June 2025, including Anthropic, Google, Microsoft and OpenAI. Evidence ↗OECD, How are AI developers managing risks? (2025) →Claim page →
· publication · dated to the day
The Department for Science, Innovation and Technology published the text of a Memorandum of Understanding between the UK government and Cohere on AI opportunities on GOV.UK on 16 June 2025. Evidence ↗· publication · dated to the day
1 provision record of Memorandum of Understanding between the UK and Cohere on AI opportunities
· publication · dated to the day
The Joint California Policy Working Group on AI Frontier Models released its final report, "The California Report on Frontier AI Policy", on 17 June 2025. Evidence ↗· signature · dated to the day
Texas S.B. 6 (89th Legislature), on the planning for, interconnection and operation of, and costs of serving certain electrical loads, was signed by the Governor on 20 June 2025 and took effect immediately. Evidence ↗· effective · dated to the day
1 provision record of Texas S.B. 6 (89th Legislature): planning for, interconnection and operation of, and costs of serving certain electrical loads
· event · dated to the day
Michigan House Bill 4668, the artificial intelligence safety and security transparency act, was introduced in the Michigan House of Representatives on 24 June 2025 and referred to the Committee on Judiciary. Evidence ↗· publication · dated to the day
1 provision record of Michigan House Bill 4668 (2025), the artificial intelligence safety and security transparency act
· snapshot · dated to the day
Baidu's Hugging Face model card for ERNIE-4.5-300B-A47B-PT, at its revision of 29 June 2025, states that the ERNIE 4.5 models are provided under the Apache License 2.0, and the repository's LICENSE file at that revision is the Apache License, Version 2.0. Evidence ↗
July 2025
· publication · dated to the month
The Frontier AI Risk Management Framework (v1.0), July 2025, names Concordia AI with Shanghai AI Laboratory in its citation line, and its executive summary says that Shanghai AI Laboratory introduced the framework in collaboration with Concordia AI. Evidence ↗· publication · dated to the month
Shanghai AI Laboratory, in collaboration with Concordia AI, introduced the Frontier AI Risk Management Framework (v1.0), dated July 2025 in its own citation line. Evidence ↗Shanghai Artificial Intelligence Laboratory / 上海人工智能实验室 →Claim page →
· publication · dated to the month
2 provision records of Frontier AI Risk Management Framework (v1.0)
· publication · dated to the day
The Shelby County Health Department issued construction permit 01156-01PC on 2 July 2025 to CTC Property LLC for fifteen Solar SMT-130 natural gas-fired combustion turbines, each rated at 16.48 MW, at xAI’s Memphis facility, with a printed expiry of 2 January 2027, as the cross-family check read the permit. Action News 5 reported the approval on 2 July 2025. Evidence ↗· snapshot · dated to the day
Anthropic’s amended and restated certificate of incorporation filed on 3 July 2025, the latest inspected, states its specific public benefit as responsibly developing and maintaining advanced AI for the long term benefit of humanity. The same wording is in Article III of the restated certificate filed on 24 May 2021. Evidence ↗· snapshot · dated to the day
Under Anthropic’s restated certificate filed on 3 July 2025, while any Class T Common Stock is outstanding the company may not, without the consent or vote of all Class T holders as a separate class, amend the Class T provisions or the phase-in definitions, alter the certificate so that the board’s legal powers are exercised by another person or body, or change the number of directors except to add seats at least half of which Class T holders elect. Acts without that consent are declared void. Evidence ↗· snapshot · dated to the day
The restated certificate filed on 3 July 2025 provides that any director elected under Article VIII(B) may be removed at any time without cause by the holders of the class or series of stock entitled to elect that director, so the holders of Class T Common Stock, which Anthropic states is held by the Long-Term Benefit Trust, may remove the Class T Directors. Evidence ↗· publication · dated to the day
The European Commission received and published the final General-Purpose AI Code of Practice on 10 July 2025. The Code consists of three separately authored chapters: Transparency, Copyright, and Safety and Security. Evidence ↗· publication · dated to the day Inferred
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. Evidence ↗· publication · dated to the day Inferred
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. Evidence ↗· publication · dated to the day
11 provision records of General-Purpose AI Code of Practice (final version, 10 July 2025)
· announcement · dated to the day
On 11 July 2025 OpenAI announced its intention to sign the General-Purpose AI Code of Practice, subject to the then-current version being formally approved by the AI Board during its adequacy assessment. Evidence ↗· publication · dated to the day
2 provision records of Kimi K2 Modified MIT License
· publication · dated to the day
The ChatGPT Agent System Card states that OpenAI conducted human red teaming on biological risk as one of two key risk areas, and sets out an overview table of its biological and chemical evaluations. Evidence ↗· publication · dated to the day
OpenAI published the ChatGPT Agent System Card, dated 17 July 2025 on its cover. Evidence ↗· publication · dated to the day
The ChatGPT Agent System Card states that OpenAI invited seven experts with biosafety and biosecurity backgrounds to spend three to five hours with a helpful-only version of the model and to score it in a survey, and that experts noted an average success rate of about 7.7 on a scale of 1 to 10 on the tasks they attempted. Evidence ↗· publication · dated to the day
The ChatGPT Agent System Card states that OpenAI decided to treat the launch as High capability in the Biological and Chemical domain under its Preparedness Framework, activating the associated safeguards, while stating that it does not have definitive evidence that the model could meaningfully help a novice to create severe biological harm and has taken a precautionary approach. Evidence ↗· publication · dated to the day · reported as performed
OpenAI's ChatGPT Agent System Card reports that two SecureBio virologists carried out about ten hours of manual red-teaming with ChatGPT agent, and that SecureBio also supplied an external assessment within the card's biological evaluations. Evidence ↗· announcement · dated to the day
On 18 July 2025 Joel Kaplan, Meta’s global affairs chief according to CNBC, wrote on LinkedIn that Meta had reviewed the European Commission’s Code of Practice for general-purpose AI models and would not sign it. Evidence ↗· publication · dated to the day
The Commission’s library entry for its guidelines on general-purpose AI models, with a publication date of 18 July 2025, offers for download a Communication to the Commission, C(2025) 5045 final, which asks the Commission to approve the content of draft guidelines. The Communication states that the guidelines will be formally adopted later and applicable only from that moment. Evidence ↗· announcement · dated to the day
On 21 July 2025 Anthropic announced that, after review, it intended to sign the General-Purpose AI Code of Practice. Evidence ↗· publication · dated to the day
The Department for Science, Innovation and Technology published the text of a Memorandum of Understanding between DSIT and OpenAI on AI opportunities on GOV.UK on 21 July 2025. Evidence ↗· publication · dated to the day
2 provision records of Memorandum of Understanding between the UK and OpenAI on AI opportunities
· signature · dated to the day
Section 4 of Executive Order 14318 of 23 July 2025 revoked Executive Order 14141 of 14 January 2025. Evidence ↗· signature · dated to the day
Executive Order 14318, "Accelerating Federal Permitting of Data Center Infrastructure", is dated 23 July 2025 and was published in the Federal Register on 28 July 2025 (FR Doc. 2025-14212). Evidence ↗· publication · dated to the day
The White House released "Winning the AI Race: America’s AI Action Plan" on 23 July 2025. Evidence ↗· publication · dated to the day
6 provision records of Winning the Race: America’s AI Action Plan
· snapshot · dated to the day
The Seed-X-Instruct-7B model card at its revision of 28 July 2025 states that the project is licensed under OpenMDW, and the repository's LICENSE file is the OpenMDW License Agreement, version 1.0. Evidence ↗· publication · dated to the day
1 provision record of Kimi K2: Open Agentic Intelligence
· publication · dated to the day
1 provision record of Executive Order 14318: Accelerating Federal Permitting of Data Center Infrastructure
· announcement · dated to the day
On 30 July 2025 Google announced, in a post by Kent Walker, that it would join other companies in signing the General-Purpose AI Code of Practice. Evidence ↗· publication · dated to the day
Zhipu’s pinned Chinese GLM-4.5 README states MIT licensing for the listed GLM-4.5 and GLM-4.5-Air variants. Evidence ↗· announcement · dated to the day
On 31 July 2025 xAI announced on X that it would sign the Safety and Security chapter of the EU AI Act’s Code of Practice. Evidence ↗
August 2025
· publication · dated to the day
Amazon appears on the European Commission’s list of signatories of the General-Purpose AI Code of Practice from the list’s first publication on 1 August 2025, and is still listed in the version last updated on 31 July 2026. Evidence ↗· publication · dated to the day
Anthropic appears on the European Commission’s list of signatories of the General-Purpose AI Code of Practice from the list’s first publication on 1 August 2025, and is still listed in the version last updated on 31 July 2026. Evidence ↗· publication · dated to the day
Cohere appears on the European Commission’s list of signatories of the General-Purpose AI Code of Practice from the list’s first publication on 1 August 2025, and is still listed in the version last updated on 31 July 2026. Evidence ↗· publication · dated to the day
Google appears on the European Commission’s list of signatories of the General-Purpose AI Code of Practice from the list’s first publication on 1 August 2025, and is still listed in the version last updated on 31 July 2026. Evidence ↗· publication · dated to the day
IBM appears on the European Commission’s list of signatories of the General-Purpose AI Code of Practice from the list’s first publication on 1 August 2025, and is still listed in the version last updated on 31 July 2026. Evidence ↗· publication · dated to the day
Microsoft appears on the European Commission’s list of signatories of the General-Purpose AI Code of Practice from the list’s first publication on 1 August 2025, and is still listed in the version last updated on 31 July 2026. Evidence ↗· publication · dated to the day
Mistral AI appears on the European Commission’s list of signatories of the General-Purpose AI Code of Practice from the list’s first publication on 1 August 2025, and is still listed in the version last updated on 31 July 2026. Evidence ↗· publication · dated to the day
OpenAI appears on the European Commission’s list of signatories of the General-Purpose AI Code of Practice from the list’s first publication on 1 August 2025, and is still listed in the version last updated on 31 July 2026. Evidence ↗· publication · dated to the day
From the list’s first publication on 1 August 2025, the European Commission’s signatory page leaves xAI out of its list of signatories of the General-Purpose AI Code of Practice and states separately that xAI signed up to the Safety and Security chapter. Evidence ↗· signature · dated to the day
The European Commission issued Opinion C(2025) 5361 final, dated 1 August 2025, on the assessment of the General-Purpose AI Code of Practice under Article 56 of the AI Act. Evidence ↗· signature · dated to the day
6 provision records of Commission Opinion C(2025) 5361 final on the assessment of the General-Purpose AI Code of Practice
· effective · dated to the day
1 provision record of Regulation (EU) 2024/1689 (Artificial Intelligence Act), as published in the Official Journal
· publication · dated to the day · reported as performed
OpenAI's GPT-5 System Card states that it gave the US Center for AI Standards and Innovation early access to gpt-5-thinking under a newly signed, updated agreement, and that CAISI evaluated the model's biological and chemical capabilities and its safeguards. Evidence ↗· publication · dated to the day
OpenAI published the GPT-5 System Card, dated 13 August 2025 on its cover. Evidence ↗· publication · dated to the day
The GPT-5 System Card states that OpenAI decided to treat gpt-5-thinking as High capability in the Biological and Chemical domain under its Preparedness Framework, activating the associated safeguards, while stating that it does not have definitive evidence that the model could meaningfully help a novice to create severe biological harm. Evidence ↗· publication · dated to the day · reported as performed
OpenAI's GPT-5 System Card states that the UK AI Security Institute evaluated gpt-5-thinking's biological and chemical capabilities and its safeguards, with access to prototype safeguards and non-public information sources as part of a longer-term collaboration. Evidence ↗· publication · dated to the day
The GPT-5 System Card states that the UK AI Security Institute's safeguards team identified multiple model-level jailbreaks that overcome gpt-5-thinking's built-in refusal logic without degrading model capabilities, and that one of them evades all layers of mitigations and is being patched. Evidence ↗· revision · dated to the day
The Grok 4 Model Card states that the area of highest concern is the model's expert-level biology capabilities, which significantly exceed human expert baselines on the benchmarks used, while stating that these evaluations measure dual-use knowledge and that a high score indicates greater capability to enable weapons development, not necessarily increased risk. Evidence ↗· revision · dated to the day
xAI published the Grok 4 Model Card, the expression read stating that it was last updated on 20 August 2025. Evidence ↗· announcement · dated to the day
The Louisiana Public Service Commission decided at its Business and Executive Session of 20 August 2025, in Order U-37425 dated 29 August 2025 and effective immediately, on Entergy Louisiana’s application for generation and transmission resources to serve a single customer in North Louisiana. Entergy stated on 20 August 2025 that the approval allows it to build three combined-cycle combustion turbine plants tied to Meta’s Richland Parish data centre, two there expected online in late 2028 and one at Waterford by the end of 2029. Evidence ↗· snapshot · dated to the day
The Seed-OSS model card (revision of 26 August 2025) states that the Seed-OSS series is released under the Apache-2.0 license, and the seed-oss Git repository's LICENSE file at its commit of 25 August 2025 is the Apache License, Version 2.0. Evidence ↗
September 2025
· publication · dated to the month
The Claude Sonnet 4.5 system card states that the model demonstrated enhanced performance on several biological risk evaluations while remaining below ASL-4 rule-out thresholds, and that Anthropic has not determined whether it passed the ASL-3 capabilities threshold but cannot clearly rule out ASL-3 risks, so deploys it with ASL-3 measures as a precautionary, provisional action. Evidence ↗· publication · dated to the month
The Claude Sonnet 4.5 system card states that its CBRN evaluations primarily focus on biological risks with the largest consequences and sets out separate ASL-3 and ASL-4 threat models for that domain. Evidence ↗· publication · dated to the month
Anthropic published the Claude Sonnet 4.5 system card, dated September 2025 on its cover. Evidence ↗· effective · dated to the day
1 provision record of Measures for Labelling Artificial Intelligence-Generated Synthetic Content (人工智能生成合成内容标识办法)
· publication · dated to the day
The AI Safety Governance Framework 2.0 (September 2025) names CNCERT/CC (国家计算机网络应急技术处理协调中心) beneath TC260 on its cover, and its preface says that TC260 organised CNCERT/CC and other professional bodies, research institutes and companies to prepare it. Evidence ↗· publication · dated to the day
The AI Safety Governance Framework 2.0, a TC260 technical document whose cover names TC260 and CNCERT/CC, was released on 15 September 2025 at the main forum of the 2025 National Cybersecurity Awareness Week. Evidence ↗· publication · dated to the day
5 provision records of AI Safety Governance Framework 2.0 (人工智能安全治理框架2.0)
· announcement · dated to the day · announced
The UK government announced on 17 September 2025 an AI Growth Zone in the North East, made up of sites in Blyth and Cobalt Park, which it said was set to attract up to £30 billion in private investment. Evidence ↗· publication · dated to the day Inferred
The Frontier Safety Framework v3.0 defines its CBRN risk domain as risks of models assisting in the development, preparation and/or execution of a CBRN threat, where v2.0 defined the same domain as risks of assisting in a CBRN attack. Evidence ↗· publication · dated to the day
FSF v3.0 introduces a Critical Capability Level for harmful manipulation. Evidence ↗· publication · dated to the day
Google DeepMind published Frontier Safety Framework v3.0, identified by its publication date of 2025-09-22. Evidence ↗· publication · dated to the day
34 provision records of Google DeepMind Frontier Safety Framework — v3.0
· publication · dated to the day
Frontier Safety Framework v3.0 replaces v2.0 in the publisher’s documented framework version sequence. Evidence ↗· announcement · dated to the day · reported as performed
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. Evidence ↗· announcement · dated to the day · reported as performed
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. Evidence ↗· publication · dated to the day
19 provision records of California SB 53 (2025): Transparency in Frontier Artificial Intelligence Act, as chaptered
· signature · dated to the day
The Governor of California approved Senate Bill 53, which enacts the Transparency in Frontier Artificial Intelligence Act, on 29 September 2025; it was filed with the Secretary of State the same day and chaptered as Chapter 138, Statutes of 2025. Evidence ↗
October 2025
· publication · dated to the day
MOFCOM issued Announcement No. 61 of 2025 on 9 October 2025, requiring foreign exporters to obtain a MOFCOM dual-use export licence for certain foreign-made items containing Chinese-origin rare-earth items, with parts effective from 1 December 2025 and the rest on publication. Evidence ↗· publication · dated to the day
1 provision record of MOFCOM Announcement No. 61 of 2025 on export controls on related foreign rare-earth items (商务部公告2025年第61号)
· signature · dated to the day
In a letter of 23 October 2025, citing his authority under section 403 of the DOE Organization Act, the Secretary of Energy directed FERC to initiate rulemaking and attached for its consideration an advance notice of proposed rulemaking on interconnecting large loads, including AI data centres, to the transmission system, asking for final action no later than 30 April 2026. Evidence ↗· publication · dated to the day
The MiniMax-M2 GitHub repository carried an MIT License text with a 2025 MiniMax AI copyright notice as of a commit of 27 October 2025. Evidence ↗· event · dated to the day
OpenAI published its account of a completed recapitalization, with OpenAI Foundation controlling OpenAI Group PBC. Evidence ↗· event · dated to the day UnresolvedQuestion posed by Selekoa
Following the October 2025 recapitalization, the OpenAI Foundation can exercise effective mission-directed control over OpenAI Group PBC in practice. Evidence ↗· publication · dated to the day
The Delaware Department of Justice issued a letter to OpenAI, Inc. dated 28 October 2025 setting out the conditions under which the Attorney General would not object to OpenAI, Inc.’s recapitalization, which it describes as restructuring a for-profit limited liability company into a newly formed public benefit corporation, with the nonprofit retaining control of it in accordance with OpenAI’s representations. Evidence ↗· snapshot · dated to the day
OpenAI states that the OpenAI Foundation, through special voting and governance rights held solely by it, appoints all members of OpenAI Group’s board and can replace directors at any time. The rights are those of the Class N Common Stock in the certificate of incorporation filed on 28 October 2025, which makes them depend on that stock being held, directly or indirectly, by OpenAI Foundation. Evidence ↗· snapshot · dated to the day
Under Section IV.2(c) of OpenAI Group PBC’s certificate of incorporation filed on 28 October 2025, the holder of the Class N Common Stock, while it is OpenAI Foundation, must give prior written approval before the corporation amends its public-benefit article, enters into a deemed liquidation event or material asset sale, issues more Class N stock or other director-voting stock, or modifies the certificate or bylaws to the Class N holder’s disadvantage. Evidence ↗· snapshot · dated to the day
Under Section V.3 of OpenAI Group PBC’s certificate of incorporation filed on 28 October 2025, the holder of the Class N Common Stock may remove directors with or without cause while any Class N shares are outstanding; OpenAI states that the OpenAI Foundation holds these rights and can replace directors at any time. Evidence ↗· snapshot · dated to the day
OpenAI Group PBC was incorporated as a Delaware public benefit corporation. Its certificate of incorporation carries the Delaware Division of Corporations filing stamp of 28 October 2025, 08:07 AM (file number 10381551). Evidence ↗· event · dated to the day
The Standing Committee of the 14th National People's Congress adopted the Decision on Amending the Cybersecurity Law at its 18th session on 28 October 2025. Evidence ↗· publication · dated to the day
Presidential Order No. 61, dated 28 October 2025, promulgated the Decision on Amending the Cybersecurity Law. Evidence ↗· publication · dated to the day
2 provision records of Delaware Attorney General’s statement of no objection to OpenAI, Inc.’s recapitalization (2025)
· publication · dated to the day
4 provision records of Certificate of Incorporation of OpenAI Group PBC (filed 2025-10-28)
November 2025
· publication · dated to the day
Moonshot AI released the Kimi-K2-Thinking code and weights under a licence headed “Modified MIT License”, as stated in the Hugging Face repository on 6 November 2025. Evidence ↗· publication · dated to the day
1 provision record of MOFCOM–GAC Announcement No. 70 of 2025 suspending Announcements Nos. 55–58, 61 and 62 of 2025 (商务部、海关总署公告2025年第70号)
· revision · dated to the day
On 13 November 2025 the MiniMax-M2 repository’s LICENSE was changed to add, after the MIT License text, a modification requiring commercial products above 100 million monthly active users or 30 million US dollars in annual recurring revenue to display “MiniMax M2” prominently. Evidence ↗· publication · dated to the day
The Commission Guidelines on the scope of the obligations for providers of general-purpose AI models, in the file that the Commission serves at its download address, are a Communication from the Commission dated Brussels 19 November 2025 and numbered C(2025) 7719 final. Evidence ↗· revision · dated to the day
7 provision records of 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)
December 2025
· revision · dated to the day
2 provision records of Explanatory Notice and Template for the Public Summary of Training Content for general-purpose AI models (C(2025) 8311 final)
· publication · dated to the day
2 provision records of Scientific and Technological Advances in the Life Sciences since 1975: How the Last 50 Years Will Shape the BTWC S&T Advisory Mechanism (BWC/MSP/2025/WP.4)
· publication · dated to the day
The Department for Science, Innovation and Technology published on GOV.UK on 11 December 2025 a Memorandum of Understanding with Google DeepMind on AI opportunities and security, naming Liz Kendall and Demis Hassabis as signatories without a signing date. Evidence ↗· signature · dated to the day
Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence", is dated 11 December 2025 and was published in the Federal Register on 16 December 2025. Evidence ↗· publication · dated to the day
2 provision records of Memorandum of Understanding between the UK and Google DeepMind on AI opportunities and security
· publication · dated to the day
4 provision records of Executive Order 14365: Ensuring a National Policy Framework for Artificial Intelligence
· publication · dated to the day
1 provision record of Statement by the United States (BWC/MSP/2025/WP.14)
· publication · dated to the day
1 provision record of Statement by the United Kingdom of Great Britain & Northern Ireland (BWC/MSP/2025/WP.27)
· publication · dated to the day
1 provision record of Statement by Japan (BWC/MSP/2025/WP.28)
· signature · dated to the day
S6953-B was delivered to the Governor of New York on 9 December 2025 and signed as Chapter 699 of 2025 on 19 December 2025, with approval memo 76. Evidence ↗· announcement · dated to the day
The Governor of New York announced on 19 December 2025 that she had signed legislation requiring AI frameworks for frontier-model developers, the release names the RAISE Act (S6953B/A6453B) and refers to agreed-upon chapter amendments. Evidence ↗· signature · dated to the day
5 provision records of New York S6953-B (2025), the RAISE Act, signed as Chapter 699 of 2025
· announcement · dated to the day
Anthropic announced on 19 December 2025 that its Frontier Compliance Framework was available, stating that the framework will serve as its compliance framework for SB 53 and other regulatory requirements while its Responsible Scaling Policy remains its voluntary safety policy. Evidence ↗· revision · dated to the day
xAI published Frontier Artificial Intelligence Framework December 2025 edition, identified by its revision date of 2025-12-30. Evidence ↗· revision · dated to the day
25 provision records of xAI Frontier Artificial Intelligence Framework — December 2025 edition
· revision · dated to the day
xAI's Frontier Artificial Intelligence Framework (December 2025 edition) states that the FAIF complies with California's Transparency in Frontier Artificial Intelligence Act (Bus. & Prof. Code § 22757.10 et seq.). Evidence ↗· publication · dated to the day
2 provision records of MiniMax Group Inc. Global Offering prospectus (31 December 2025)
Dates stated for 2025 2 dates
Each date as its record gives it, with who stated it. A target, deadline or term says what a party said would happen; a reported date says what a party said happened; a scheduled date had not been reached when the record’s sources were last checked. None is a dated record of what happened.
(as worded: “end of 2025”) · Completion target · stated by NDRC
Task (4) of the 2024 data-centre plan sets the goal that by the end of 2025 a two-way computing–electricity coordination mechanism is initially formed and green electricity exceeds 80% for new data centres at national hub nodes, and encourages co-location of new data centres with renewable generation and pilots of direct green-power supply. Evidence ↗· Deadline · stated by The White House
Section 4(a) of Executive Order 14141 directed the Secretaries of Defense and Energy to identify, by 28 February 2025 and if possible, at least three sites each on federal land that could be leased for frontier AI data centres and the clean energy facilities to serve them, by the end of 2027. Evidence ↗