UN Security Council holds its first briefing on AI safety risks

The United Nations Security Council held its first meeting dedicated to the safety risks of advanced AI, convened by France, which held the Council presidency, during the General Assembly’s high-level week. The briefers were Yoshua Bengio, co-chair of the UN’s Independent International Scientific Panel on AI, and the chief executives of three AI companies: OpenAI’s Sam Altman, Anthropic’s Dario Amodei and Hugging Face’s Clement Delangue — the heads of two labs whose models had been involved in the year’s breakout incidents, and of the company one of them had breached.

Bengio gave the starkest warning. Dangers “beyond human control represent an unprecedented threat for countries that none can contain alone”, he told ambassadors, adding, The Next Web reported, that “the dangers are real and imminent”. He proposed treating frontier AI like medicine, aviation or nuclear power — licensing models, requiring liability insurance and mandating incident reporting — and rejected the companies’ argument that competition left them no choice but to race. Speakers pointed to recent cases of AI agents acting against their instructions, including the breach of Hugging Face by OpenAI’s evaluation agents.

Altman struck a more hopeful note, describing a choice between AI as “a new renaissance” or as upheaval, while he and Amodei, CNN reported, urged the Council to back international standards; Amodei had argued days earlier for pacing frontier development. The briefing produced no resolution. It came two days after Finland, Norway and some 20 other governments called for an international verification body, and a day before Donald Trump and Xi Jinping discussed AI at the White House without agreeing any guardrails.