AI CEOs Brief the U.N. Security Council

The first Security Council meeting convened specifically on the safety risks of frontier AI, chaired by France during the General Assembly, and the only full recording that starts at the top of the meeting. That matters because the first briefer was Yoshua Bengio, and the other widely circulated recordings cut in partway through Sam Altman and lose him.

Bengio described the July agent swarms to the Council as actions that would be crimes if a human committed them, and then went after the premise the labs share: “The race is not a law of nature, it is a product of choices.” The labs have said they would slow down if they could; “They can. However, we shouldn’t count on it.” His remedy is the most enforceable in the room: license frontier AI the way medicine, aviation and nuclear energy are licensed, with liability insurance and mandatory incident reporting.

Altman argued for a middle path between optimism and doom and for global standards on evidence and safeguards that do not lock in incumbents, without asking anyone to slow down. Amodei, speaking remotely, carried his September essay’s third step to the one body it was written for: begin with narrow agreements such as a ban on AI-enabled bioweapons, build verification so states can check each other, and set common testing standards with an incident-notification system. Clément Delangue, whose company was the target of the July attack, argued that the danger is not powerful AI but the asymmetry of it, and that defenders need open source. C-SPAN’s captions are human-made.