OpenAI calls for US-led global technical standards for frontier AI, including recursive self-improvement
In "Building standards for the next phase of AI" (Sept 21, 2026), OpenAI proposed US-led international technical standards for frontier AI, coordinated through CAISI and a network of AI safety institutes. Topics include measuring progress toward recursive self-improvement (RSI), triggers for human oversight, and incident classification. OpenAI wrote that "fully autonomous RSI is not happening today, and we should not pursue it unless and until it can be done safely".
Key facts
- Published Sept 21, 2026 on openai.com
- Quote: 'Fully autonomous RSI is not happening today, and we should not pursue it unless and until it can be done safely'
- Network of AI safety institutes (Australia, Canada, Germany, France, Kenya, Japan, Korea, Singapore, India, UK) coordinated through the US CAISI
- Standards would 'not be licenses, mandatory prerelease review, or approval requirements'
- Proposed standards: measuring RSI progress, triggers for human oversight, incident classification and reporting; supports US–China dialogue
What happened
OpenAI set out its preferred governance model: voluntary but shared technical standards run through government safety institutes, not licensing. It was published two days before Altman's Security Council remarks.
Why it matters
It is the first time a frontier lab has publicly named RSI as something to standardize and not pursue until safe. It also stakes out a middle position between the US administration's opposition to global oversight and calls for an international agency.
Changelog
- 2026-09-29: created
- 2026-09-29: sweep 2026-09-29: added Axios link; related to the industry standards-body report
Related posts (1)
- Building standards for the next phase of AI OpenAI @OpenAI · blog · 2026-09-21
OpenAI says fully autonomous recursive self-improvement should not be pursued until safe and proposes international technical standards via safety institutes.
Related events
- Hinton, Bengio, Pachocki, Jack Clark and others: automating AI R&D could trigger an 'intelligence explosion' ★★★★
- Altman and Amodei ask the UN Security Council for international AI standards and incident reporting ★★★★
- OpenAI publishes early guidelines for 'safety cases' before frontier training runs ★★★
- Google, OpenAI and Anthropic plan an industry safety standards body, the 'Standards Authority for Frontier AI', without government oversight ★★★
- Mirendil, an ex-Anthropic startup building self-improving AI, in talks to raise up to $1B at a $5B valuation ★★★
- Rep. Ro Khanna announces the Human Control Over AI Act: ban on recursive self-improving AI, strict liability, FDA-style agency ★★★
Sources (3)
- officialOpenAI: Building standards for the next phase of AI
- pressTechNode Global: OpenAI urges US to lead global technical standards for frontier AI
- pressAxios: OpenAI urges US to lead global AI safety standards effort ahead of Altman's UN address
id: 2026-09-21-openai-building-standards-rsi · updated 2026-09-29 · open in the interactive timeline