OpenAI shares embedded monitoring assessment by METR and Redwood Research

What happened

OpenAI said it invited METR and Redwood Research to evaluate its internal monitoring system this summer, providing deep infrastructure access over three weeks. OpenAI said the findings helped strengthen safeguards as it expanded the system. In the quoted post, METR said two employees from the organizations worked on premises to assess monitoring for misalignment in internal agents, and a brief write-up is now public.

Read next

• More changes in today’s briefingMicrosoft Research and Collaborator Resolve the KLS Conjecture With Human-Directed AI

The story so far

OpenAI shared findings from an assessment of its internal monitoring system by METR and Redwood Research, saying the findings helped strengthen safeguards as it expanded the system. OpenAI said it invited the organizations to conduct the evaluation over the summer and gave them deep access to its infrastructure for three weeks. METR said two employees from the organizations worked on premises to assess monitoring for misalignment in OpenAI’s internal agents, and that a brief write-up is now public.