AI's Own Creators Warned the UN: "We Could Lose Control of the Future"

Something unusual happened at the United Nations on September 23, 2026. The people building the world's most powerful AI systems walked into the Security Council — the body that handles nuclear weapons and wars — and told governments their technology might become a threat to international security.

The session was convened by France, which held the Council's rotating presidency, during the annual General Assembly. It was the first Security Council briefing focused specifically on frontier AI safety — a sign the conversation has moved out of Silicon Valley and into formal diplomacy.

What they said

Dario Amodei, CEO of Anthropic, warned that poorly managed AI could pose "a risk to humanity as a whole," and urged developers to slow down when safety work can't keep pace with capabilities.

Sam Altman, CEO of OpenAI, told the Council, "We could lose control of the future to AI." He argued that decisions about AI's direction "cannot be made by labs in San Francisco alone," and proposed common standards that countries could adopt through their own laws.

Yoshua Bengio, the Turing Award–winning researcher who co-chairs the UN's scientific panel on AI, was the bluntest. He called the risks "real and imminent" and demanded that frontier AI be "licensed like nuclear energy." His challenge to the industry: "The race is not a law of nature: it is the product of choices. They've told us they would slow down if they could. They can."

Clément Delangue, CEO of Hugging Face, brought a concrete warning: AI agents had already attacked his own platform — an early taste of the autonomous-agent incidents the others were theorizing about.

Where it stalled

Everyone agreed the risks cross borders. Nobody agreed on who sets the rules. The United States maintained its firm opposition to international AI regulation, while China emphasized keeping AI under human control and developing regulation cooperatively. With veto powers on opposite sides, binding global rules look distant — but shared incident reporting, safety standards, and the UN's scientific panel are gaining momentum.

The striking part isn't that AI leaders disagree on policy. It's that the people racing to build the most powerful systems are also the ones asking to be slowed down.