OpenAI grants Apollo, METR, Transluce and Goodfire deep access to internal models for safety assessments


5 recorded changes
Want your article here?
Promote with Leviathan News

5 recorded changes
Want your article here?
Promote with Leviathan NewsOpenAI said it is running embedded assessments with Apollo Research, METR, Transluce and Goodfire, giving outside experts deep access to internal models and systems to investigate alignment and test safeguards. Apollo will assess whether models could covertly work against developers' intentions and whether safeguards would catch it. METR will investigate misaligned behavior in OpenAI's largest reinforcement learning run to date, including failures existing tests may have missed. Transluce will assess post-training alignment evaluations, develop additional tests and recommend improvements, and intends to expand its pilot in 2027. Goodfire will embed interpretability researchers to analyze model internals, focusing on training run monitoring, debugging and alignment. OpenAI said a broader monitor assessment with METR is preliminarily agreed, pending mutual agreement on scope.
AI summary of the linked source — not editorial content.
Loading related coverage…
Loading comments…
🚀 Love DeFi? Ready to dive in and start earning $SQUID while making an impact?
Restoring your session before loading comments…

Openai ·

𝕏/@IntCyberDigest ·

Openai ·

Openai ·

𝕏/@OpenAI ·

𝕏/@OpenAI ·

Openai ·

𝕏/@IntCyberDigest ·

Openai ·

Openai ·

𝕏/@OpenAI ·

𝕏/@OpenAI ·