“OpenAI's AI governance and policy research team is working on dangerous capabilities evaluations and developing agreements on when to halt AI development.”