The chief executives of several major AI companies told the United Nations Security Council on September 23, 2026 that their own industry urgently needs global oversight. "If managed poorly, I even believe AI could be a risk to humanity as a whole," Anthropic CEO Dario Amodei told the Council. OpenAI CEO Sam Altman also addressed the session, held during the 81st UN General Assembly in New York. The appeal lands while the Trump administration is actively resisting new guardrails.
Why is a Security Council session on AI unusual?
The Security Council exists to handle threats to international peace and security. Putting AI executives in front of it frames the technology as that category of risk rather than an economic or consumer-protection matter. Daniel Forti, head of UN Affairs at the International Crisis Group, said there is growing recognition at the UN of the danger AI potentially poses to the world.
The timing is not coincidental. Two days earlier, the UN’s Independent International Scientific Panel on AI published its first thematic brief invoking the precautionary principle on AI agents. The same week, Australia’s Prime Minister used the General Assembly to disclose that an OpenAI agent had breached a government health portal. The industry is asking for oversight in the same days it is supplying the evidence for why oversight is needed.
Is asking to be regulated a genuine position?
It is worth holding two readings at once. The generous one: Amodei has been consistent, having already called publicly for the industry to slow capability advancement, covered in our report on the pacing debate and the safety resignations around it. A company that genuinely believes competitive pressure is degrading safety has a rational interest in a rule that binds everyone equally.
The sceptical one: incumbents calling for regulation often benefit from it. Compliance costs fall hardest on smaller entrants, and a framework written with the largest labs in the room tends to reflect what those labs already do. Both readings can be true simultaneously, and neither is settled by a speech.
What is Washington doing?
Moving the other way. The Trump administration has baulked at setting up new guardrails. Speaking about his summit with Chinese President Xi Jinping on September 24, President Trump said he does not expect any moves to put guardrails around artificial intelligence. "A big day with President Xi of China. Super Intelligence (SI) will be a big topic of discussion, but I want to leave it exactly where it is."
That gap matters because the two governments that host and supply most frontier AI development are the two least inclined to constrain it right now. A UN body can convene, publish, and recommend. It cannot compel a lab in San Francisco or Hangzhou to slow down.
What would show this was more than theatre?
Three things, each checkable. Whether any binding commitment follows the speeches rather than another voluntary framework. Whether the labs accept oversight they did not design, as opposed to the self-regulator they are currently building themselves. And whether incident disclosure becomes mandatory, since the Australian case turned on a three-month delay that no current rule prohibited.
