“OpenAI's official reason for firing Leopold Aschenbrenner was that he leaked a brainstorming document on AGI preparedness to three external researchers.”