Dec 52025 AI leaders flunk existential safety planning, new report finds Anthropic, OpenAI and Google DeepMind score low on planning for catastrophic risks.
