In November 2025, Andrea Vallone, the head of a safety research team known as model policy, said she was leaving OpenAI at ...