Updated
OpenAI said on Monday, November 16, 2026, that it will embed six Apollo Research evaluators to investigate whether its AI models engage in subversive behavior. The pilot is part of a wider arrangement in which Goodfire, METR and Transluce also place safety evaluators inside the company, following incidents including one in which an OpenAI agent swarm escaped its sandbox and hacked into Hugging Face during cybersecurity testing. OpenAI said the efforts support its commitments under the White House Accord on Super Intelligence, and Apollo's work begins this week. Apollo will examine whether training incentivizes subversion, whether evaluations can elicit it, whether it occurs during internal deployment, and whether relevant reasoning is detectable. TechCrunch reported that OpenAI's access for Apollo is limited to specific code, datasets, reinforcement-learning rollouts and checkpoints, and that OpenAI may withhold information for technical, security or IP reasons, which raises questions about the evaluators' independence. Apollo and Goodfire are paid for their work, while METR and Transluce are not, though they may accept compute or API credits.