OpenAI starts external evaluator pilot with Apollo Research
OpenAI has begun a pilot with Apollo Research in which six external evaluators examine whether its models could act covertly against their developers.
According to The Neuron, the pilot lasts four months, and OpenAI also announced agreements with METR, Transluce and Goodfire. Apollo says its access and disclosure rights are limited, which could affect how independent the findings are. The briefing adds that bounded access is not full independent oversight. The pilot targets covert behavior and attempts to undermine safeguards. First results from the Apollo assessments are worth watching in the coming months.
Source: The Neuron, published on November 17, 2026.