
OpenAI Apollo Research Pilot to Investigate AI Model Subversion
OpenAI will embed six Apollo Research evaluators for four months to investigate whether AI models attempt to undermine safety controls. The researchers will examine model training, evaluations, and…
















































































