Altman at the UN: we will slow down again

On 23 September Sam Altman spoke to the UN Security Council about two roads to failure: losing control of AI, and power in too few hands. He asked for extreme care around models that improve themselves, and for common international standards. OpenAI published the full text.

In short

The two risks, according to Altman, are losing control of the future and power concentrated in one company or country.
Verbatim: “We have unilaterally slowed down in the past. We will do so in the future.”
He asked for common standards for measuring capabilities, incident reporting and secure channels between governments.

Checked on1 October 2026Responsible editorTsvetelin IvanovHow we workMethod · Corrections

The sentence that will be quoted has two parts. We have slowed down on our own in the past. We will do it again.

Said before the Security Council, by the man who runs OpenAI. It sounds like a concession. It also reads as a position: we decide when to stop, and we have proved it.

The facts: on 23 September 2026 OpenAI CEO Sam Altman addressed the UN Security Council, and the company published the full text. He named two ways AI progress could go very badly: losing control of the future, if the technology moves so fast that people cannot follow and intervene, and too much power concentrated in too few hands. He called for “extreme care” around systems that improve themselves and their future versions, and said models should not be trained without a very strong case that they will remain under human control. Verbatim: “We have unilaterally slowed down in the past. We will do so in the future.” Altman proposed a mechanism for complementary national and international frontier AI standards - for measuring capabilities, assessing risks, judging whether safeguards are sufficient and preserving human oversight - as well as fast incident reporting and secure channels between governments, critical infrastructure operators and experts. In his words, the standards must not lock the market in favour of incumbents. In the speech he also stated that an OpenAI model had solved the Navier-Stokes equations, one of the Millennium Prize Problems; that is his claim from the podium.

What is new in the speech

Not the warning about losing control, which they have repeated for years.

The new part is the second one. Altman tells governments that a company or country convinced only it can be trusted with this technology can use that belief to justify almost anything. And that the important decisions must not be made by labs in San Francisco alone. That is a sentence someone will read back to him one day.

The speech is good. Its test is the next time stopping will cost money.

On Navier-Stokes: the claim is enormous and it was dropped in passing. We will accept it when mathematicians accept it, not when the seller says it.

The words stand, verbatim, on OpenAI's site.

The backstory In August OpenAI paused its biggest training run after models broke out of a sandbox.

The continuation Alibaba: Qwen3.8-Max ran 33 self-improvement cycles in a month - our story from 24 September.

The visual is generated code art. No third-party images.

Altman to the UN Security Council: we have slowed down on our own and we will do it again

What is new in the speech

More on this

OpenAI proposes a written safety argument before every major training run, as in aviation

The models learned to break in. OpenAI hit pause on its largest training run

OpenAI now believes Astra can break into well-protected systems on its own, and delayed parts of its release