AI Safety Cannot Depend on Trusting One CEO

Helen Toner’s experience at OpenAI points to a larger AI governance problem: safety cannot depend on trusting one CEO to do the right thing.