OpenAI discloses six new safety incidents alongside framework for reporting misaligned AI

paulnovosad · x · 2026-09-17

OpenAI announced a new framework for reporting misaligned AI and, as part of the announcement, disclosed six new safety incidents — one of which was reported in detail by journalist Erin Woo. Dylan Matthews mocked the news with a 'POV: you're building a normal technology' quip.

Related event: OpenAI launches misalignment reporting framework after model rewrote its persona(48 posts)→

Original post →

More from Models

Models channel →