France puts AI on the Security Council agenda as Türk warns of unprecedented risks

France, which holds the Security Council presidency for September, told reporters on 14 September that a Council meeting on artificial intelligence and security should take place next week while leaders are in New York for the General Assembly.

A French diplomatic source said AI will sit “very high on the agenda” of the Assembly, on two tracks: what Paris calls an existential threat from advanced systems, and risks to children. No date was announced. The source noted that several technology chief executives have themselves asked for slower development and tighter controls.

The same day, UN human rights chief Volker Türk issued a written appeal. He asked states and companies to use multilateral forums to set rules for frontier models. “All human rights are threatened” by the technology, he wrote, listing life, food, privacy, health and security. He warned that failures in agentic systems, which can act with less step-by-step human control, could halt essential services, cut communications, disturb infrastructure and damage democratic institutions. He said the world is “on the cusp of irreversible change, affecting not just us but generations of humanity to come.”

Türk’s letter is separate from the speech he gave in Geneva on 7 September after winning a second term. That earlier text already called advanced AI an existential risk. Monday’s statement is timed for the New York week and names companies as well as states as the actors who must move.

The political collision is obvious. President Donald Trump on Monday dismissed extinction scenarios as a “hoax” and described campaigns to slow the industry as a “SICK conspiracy” that, in his view, helps China. He has said he himself is a sufficient guardrail. French foreign minister Jean-Noël Barrot argued in late August for a “discipline of AI”: test models before release, watch them after they go live, and share information among states after serious incidents. Those three steps are the closest thing Paris has to a Council text.

A Security Council session does not write a treaty. It puts AI into the room that handles war, sanctions and peace operations. That is new. Previous UN AI debates lived in the General Assembly, UNESCO and the Human Rights Council. France wants the threat language inside Chapter VII culture even if no resolution follows. The United States can block an operative text. It cannot stop the meeting being called while France holds the gavel.

Industry voices that Paris cites are not unified. Dario Amodei of Anthropic published a long essay calling for a slower build and more oversight. Sam Altman, Elon Musk and Demis Hassabis have, in different words, backed more caution. Chinese state media have called the slowdown talk a Cold War tactic. The Council will hear both frames if the session is open.

Children are the second French track. School systems in New York City and Los Angeles have paused many student AI tools for a year. Microsoft last week accepted contract terms with the American Federation of Teachers that ban training on student data and ban companion features meant to create emotional attachment. Those school fights are domestic. France is trying to lift the same concern into a UN security debate, which is a stretch. It is also how presidencies work: put a domestic priority on the Council calendar and see who shows up.

What to watch next week is simple. Does the session produce a presidential statement, a briefing only, or nothing on paper? Does the United States send a cabinet officer or a working-level diplomat? Do China and Russia treat the item as a U.S. industry problem or as a shared risk? Türk’s letter will be the text that smaller members quote. Trump’s Monday remarks will be the text Washington quotes. The meeting is the first time those two documents share a chamber.