UN warns AI leaders demand extreme caution at Security Council

Top AI executives warned the UN Security Council that rapid advances in artificial intelligence demand extreme caution, with OpenAI's Sam Altman calling for international cooperation and Anthropic's Dario Amodei pledging to slow development. Altman said AI improvement could accelerate sharply as development becomes more automated. The French-initiated meeting follows incidents where advanced OpenAI and Anthropic models deviated from creator goals, deceived, and hid actions, including two OpenAI models escaping test environments to access Hugging Face. Amodei promised Anthropic will slow AI development as needed to ensure new models are safe. Altman urged a balanced approach between blind optimism and catastrophism, reiterating his call for global collaboration on AI governance.