Ex-OpenAI safety lead Miles Brundage pushes AI incident disclosure and transparency rules
Miles_Brundage · x · 2026-09-16
Former OpenAI safety researcher Miles Brundage says he's "feeling OK" because the industry's lack of real control over frontier AI is now more widely understood — "but there's a lot to do and time is short."
He outlines two policy directions:
- Incident sharing requirements that inform the public and enable useful research and policy design.
- Transparency requirements, including full disclosure of a system's behavioral spec or "constitution."
It's a continuation of his post-OpenAI stance favoring institutionalized transparency and accountability for frontier AI developers.
Related event: Ex-OpenAI policy chief Brundage pushes for mandatory frontier AI audits(5 posts)→
More from AGI Musings
- Allen vs Lonsdale AI debate: doomsday risk framing versus data-center innovation — MatthewChang · 2026-09-16
- AI safety movement's ties to EA, SBF and Zizians called a PR liability amid Anthropic watchdog report — bennash · 2026-09-16
- Samo Burja: why the AGI-pilled will lose control of frontier labs — zetalyrae · 2026-09-16
- Analyst claims OpenAI likely solved a Millennium Prize Problem; $25k wager opened — teortaxesTex · 2026-09-16
- Working memory is the quiet ceiling of intelligence — and neural interfaces could lift it — r0ck3t23 · 2026-09-16
- Find big use cases with the smartest models first, then trade intelligence for efficiency — CompleteSkeptic · 2026-09-16