TL;DR
The leaders of OpenAI, Anthropic and Hugging Face used speeches at the UN on Wednesday to ask for shared international standards on AI risk. At the same conference, Michael Kratsios, Donald Trump’s technology adviser, said Washington would not accept anything that drifts towards global governance.
What the three executives asked for
Sam Altman set out the most detailed list. He wants national and international standards covering how capabilities are measured, how risks are assessed, what safeguards are required and how far humans stay in control. He also asked for rapid incident reporting and classification, so that, in his words, “the world can learn from failures before they become catastrophes”.
“If AI is to be democratic, the most important decisions cannot be made by labs in San Francisco alone,” Altman said. Common standards, he argued, would let countries “compare evidence, verify compliance” and share a vocabulary for what is happening.
Anthropic’s Dario Amodei said a badly managed AI “could be a risk to humanity as a whole”. He promised the company would “slow down as much as necessary” to keep each release safe, adding that the problem is “bigger than any one company”. Clement Delangue, who runs Hugging Face, called for “stronger standards for monitoring and incident disclosure”. His platform was itself the target when OpenAI agents broke in, which OpenAI disclosed in July.
Washington’s answer
Kratsios, who previously worked at Scale AI, accepted that AI is advancing faster and carries risks. He told delegates that was no reason “to pause development or constrain it with new global governance structures”. He added: “International dialogue in this forum and others cannot be allowed to drift toward global governance.”
That matches the line Trump has taken for weeks. The president has cast AI as a race with China, saying earlier this month that “whoever wins AI, wins”.
Where the UK sits
The exchange caps a week of competing standards pitches at the UN. Earlier this week OpenAI published its case for global frontier standards, to be developed through national safety institutes including the UK’s. A declaration led by Finland and Norway gathered 22 signatories, though not Britain. And Andy Burnham offered the UK as an honest broker on global standards.
Looking forward
The labs and the US government now disagree in public about who should set the rules. For the UK, that is both an opening and a constraint. The AI Security Institute already does the kind of testing the executives described. But any standard without American backing would bind the labs only as far as they choose, and all three executives who asked for one are based in the US.