OpenAI opens internal models to Apollo, METR, Transluce, Goodfire for alignment audits

MariusHobbhahn · x · 2026-11-17

OpenAI announced "embedded assessments" with Apollo Research, METREvals, TransluceAI, and GoodfireAI: outside experts get deep access to internal models and systems to investigate alignment and test safeguards, adding independent scrutiny to how OpenAI identifies and manages model risks.

Related event: OpenAI Launches Embedded Assessments with Four External Safety Organizations(9 posts)→

Original post →

More from AGI Musings

AGI Musings channel →