OpenAI safety reviewer walks out

The Signal · 2026-10-04

The person who helped write OpenAI's safety reports for major model releases has quit. He's now raising concerns publicly in an Atlantic editorial. The Verge frames it as another sign of internal tension over OpenAI's safety practices and release process.

Also today: The White House gathered Zuckerberg, Bezos, Musk and Anthropic's Dario Amodei to sign an AI safety pledge, reframing the technology as super intelligence.

Apple is tightening macOS Full Disk Access, saying some developers abused privacy permissions to reach users' message histories. It's the first platform-level limit on what agents can touch.

That's today's Signal. Thanks for watching.