For the first time, the UN Security Council has invited American and Chinese frontier AI developers to the same session. On Wednesday, September 23, during the UN General Assembly's high-level week in New York, OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei and Hugging Face CEO Clement Delangue addressed the 15-member council. China's DeepSeek and Moonshot AI were invited to speak too — DeepSeek founder Liang Wenfeng skipped the trip, but having two leading Chinese labs at the table at all marks a genuine first.
France, which holds the Council's rotating presidency this month, convened the meeting; Foreign Minister Jean-Noël Barrot was expected to chair. The formal agenda item was "AI and international security" — only the second time the Council has tackled AI as a standalone topic since July 2023, when China warned against AI becoming a "runaway horse" and the US pushed back on using the technology for censorship.
Behind the diplomatic language sits an open fight over how fast AI should move. Since early September, Amodei has been pitching a coordinated slowdown for frontier models — essentially an arms-control-style pact, but for neural networks. Elon Musk backs the idea; Meta's Mark Zuckerberg and Nvidia's Jensen Huang want the opposite, arguing development shouldn't pause for anyone. China's Global Times has already dismissed slowdown talk as a "Cold War playbook" meant to lock in America's lead.
US Treasury Secretary Scott Bessent said Washington and Beijing agreed to keep talking privately about AI safety. Tellingly, Chinese AI firms are staying away from Xi Jinping's parallel summit in the US on September 24 — a sign, analysts say, that Beijing wants to keep AI off the trade-negotiation table entirely. Everyone agrees the session matters symbolically; nobody expects an actual agreement, given how far apart Washington and Beijing remain on who gets to set the rules.
