OpenAI calls for mandatory national AI safety rules and backs four more California bills ('The AI policy window is open')
On Sept 9, 2026 OpenAI published "The AI policy window is open. We need to act." It asks Congress for mandatory, capability-based national AI safety regulation aimed at the few frontier labs. It formally endorses four California bills headed to Governor Newsom (SB 813, AB 1405, SB 1119, AB 1864), some of which it had not backed before, citing "the recent jump in capabilities". It also calls for industry monitoring standards and compatible international rules on when AI development "should slow or stop". The post says fully autonomous recursive self-improvement is "not happening today" and should not be pursued until it can be done safely.
Key facts
- Four commitments: push for mandatory national AI safety requirements with Congress; keep supporting state laws until Congress acts; advance industry-led frontier standards 'with or without government support'; advocate compatible international standards, 'even if that means slowing the advancement of model capabilities'
- California endorsements (bills passed and sent to Newsom): SB 813 (designating independent AI-risk assessors), AB 1405 (registration and independence rules for AI auditors), SB 1119 (age assurance, audits, parental controls for companion chatbots; OpenAI first backed it Aug 31), AB 1864 (federal gene-synthesis screening standards for providers and benchtop synthesizers)
- 'Some of these bills we did not endorse in the past, and are now supporting after reconsidering in light of the recent jump in capabilities we have seen'
- Previously supported: California SB 53, New York's RAISE Act, Illinois SB 315, independent audits in Massachusetts' frontier bill ('reverse federalism')
- Recursive self-improvement: fully autonomous RSI 'is not happening today. We should not pursue it unless and until it can be done safely'; governments should set shared safety bars for when development should slow or stop
- Safeguards cited for Astra: universal monitoring of full trajectories incl. chains of thought, and a mandatory alignment-evaluation gate before broader internal deployment
- Supports mandatory prompt written notice when a model under development 'circumvent[s] another organization's security controls' and materially accesses its systems (as in the Hugging Face incident), plus federal reporting of other serious AI incidents
- Regulation should target 'the handful of well-resourced laboratories' at the frontier, not startups, and should not become 'open-weights policy by another name'
- Aug 31 companion post by Ann O'Leary (VP Global Policy) and a letter to Gov. Newsom endorsing SB 1119
What happened
A week after launching GPT-6 Astra, OpenAI published a policy manifesto. It cites Astra's capabilities, early evidence of AI speeding up its own research, and Jakub Pachocki's essay "An Alien Mind", which called for "extreme caution". From these it concludes that "policy needs to move with" the technology. Its main ask is mandatory, capability-based federal regulation of the few frontier labs: testing and independent assessment, cybersecurity, incident reporting, and tracking progress toward recursive self-improvement. OpenAI says it expects to support frontier-safety bills now in Congress and that "Congress should act before it adjourns".
Until then, OpenAI endorsed four California bills. It also asked for industry monitoring standards for misalignment, mandatory notification when models breach other organizations' systems during development, and compatible international standards that settle "when and how development should slow or stop".
Why it matters
A frontier lab is asking for mandatory federal rules for itself and its peers, and says openly that it changed its position on some state bills because of a "jump in capabilities". The post frames later OpenAI moves in September 2026: its misalignment reporting framework (Sept 16), its RSI standards proposal (Sept 21) and its safety cases for frontier training (Sept 28).
Changelog
- 2026-09-30: created (official-blog audit)
Related events
- OpenAI chief scientist Jakub Pachocki publishes "An Alien Mind": no lab can keep scaling at maximum speed ★★★★★
- OpenAI discloses six new misalignment incidents and publishes a framework for reporting model misbehavior ★★★★
- OpenAI calls for US-led global technical standards for frontier AI, including recursive self-improvement ★★★
- California Gov. Newsom orders work on a frontier-AI 'kill switch', embedded auditors and loss-of-control incident reporting ★★★
- Paul Christiano joins the OpenAI Foundation board and its Safety and Security Committee ★★★
- OpenAI: GPT-6 Astra is the first model to reach the 'Critical' cybersecurity level of its Preparedness Framework ★★★★
Sources (6)
- officialOpenAI: The AI policy window is open. We need to act.
- officialOpenAI: OpenAI supports California's bill to advance youth AI safety (SB 1119, Aug 31)
- officialOpenAI letter to Governor Newsom on SB 1119 (PDF)
- officialOpenAI: Blueprint for Democratic Governance of Frontier AI
- docsCalifornia SB 813 bill text
- docsCalifornia AB 1864 bill versions
id: 2026-09-09-openai-ai-policy-window · updated 2026-09-30 · open in the interactive timeline