Working in safety at OpenAI or Anthropic is zeroth world problems.