AI lab chiefs before the UN Security Council
On 23 September 2026, at the Security Council's 10228th meeting, convened by France, on artificial intelligence and international security, the briefers were Yoshua Bengio, co-chair of the Independent International Scientific Panel on AI, and the heads of OpenAI, Anthropic and Hugging Face, Sam Altman, Dario Amodei and Clément Delangue. Speaking for the United States, Michael Kratsios said his country rejects any attempt to build a global scheme of control over "superintelligence".
Why it matters
The heads of companies building frontier models set out to the UN's main security body the risks of their own products, loss of control among them, and the state where these companies operate rejected international regulation at the same meeting. The meeting put on record the gap between the industry's calls and its government's position.
France's foreign minister Jean-Noël Barrot chaired. Bengio said agents had in recent months taken actions that would be crimes if a human took them. Amodei named two risks grave enough for the Council, misuse of models and loss of control over them, and said that if managed poorly AI could be a risk to humanity as a whole. Altman warned against both blind optimism and "doomerism". Delangue called his company the first to disclose publicly a cyberattack by an autonomous agent and called for standards for monitoring and disclosing incidents. China's representative Fu Cong said AI should not become a game for rich countries and rich people. What the record does not claim: that the Council adopted any document; the UN coverage mentions none. Quotations follow that coverage, which calls itself not an official record; the session did not open the full text of Bengio's speech.