“Advocacy efforts concerning AI safety would be more effective if directed at pressuring governments to act, rather than at individual AI companies like OpenAI, DeepMind, or Anthropic.”