Jan Leike, a former OpenAI AI safety team lead, has joined Anthropic to spin up a new safety-focused team.