The CEOs of two of America’s leading AI companies told the UN Security Council on Wednesday that artificial intelligence has become a threat to global security and called on world governments to write common rules for it.
But the White House won’t jump. On Thursday morning, ahead of his summit with Chinese President Xi Jinping, Trump announced that “superintelligence” — his administration’s new term for AI — would be a big topic of discussion, “but I want to leave it right where it is. That’s China’s position, too.” He added: “Our guardrail is the DOJ!”
And hours later, standing next to Trump in the White House, Xi struck a different, more cooperative tone. “We have both the ability and responsibility to develop and manage AI sustainably and to ensure that the development of AI is always under human control,” he said.
Behind the scenes, the two nations have started talking. China’s Commerce Ministry said at a news conference on Thursday that the countries had their first dialogue on AI after Finance Minister Scott Bessent met Chinese Vice Premier He Lifeng in New York earlier this week. Bessent said that the US has proposed a notification mechanism between the two countries for AI incidents related to national security.
This is similar to what Anthropic boss Dario Amodei asked the Security Council to do a day earlier. “No leader, no company or nation can do this alone,” he said.
It is also an early sign that the government’s public stance and private diplomacy are not entirely in sync. Two days earlier, Trump told the UN General Assembly that the US rejected any “globalist plan” to control AI. But now his finance minister is proposing a formal channel with China, the country that Trump and American VCs call the main villain in the AI race against the US
But Amodei’s proposal in the Security Council went even further. He called for starting with small agreements that each country could agree on, such as a ban on the use of AI to create biological weapons, then building systems that allow governments to verify each other’s commitments, and finally the global alert system. Sam Altman, CEO of OpenAI, also called for rapid, collaborative reporting of incidents. And he said his company has slowed its own AI development before and will do so again in the future. In doing so, he undermines the industry’s usual argument that they are in a race from which they cannot escape.
The language borrowed from the UN
Xi’s line about keeping AI “under human control” echoed almost verbatim what speakers had said at the Security Council a day earlier. Yoshua Bengio, an AI scientist and co-chair of the U.N. Scientific Panel on AI, said the goal is an AI that “demonstrably remains under our control.” Pakistan’s foreign minister also said the top priority was to retain “human judgment and control” over AI.
This allows Beijing to present itself as the responsible party on AI during its visit to the White House, while Trump claims China shares his hands-off approach.
China is also building its own coalition for safe AI. In July, the World Artificial Intelligence Cooperation Organization was launched with 29 countries (and no major Western democracies), offering developing countries a way to share AI rather than let a few American companies dominate. At the Security Council, Pakistan, a founding member of that organization, warned that measures “pursued in the name of security must not become tools of technology denial,” a thinly veiled attack on U.S. export controls on advanced chips or on a regulatory system that prioritizes some top models at the expense of open source. Easing chip controls is one of Xi’s top demands this week.
Clement Delangue, the French CEO of Hugging Face, the company whose OpenAI models were hacked this summer, agrees with Xi that open models make all companies safer. He told the council that when his team tried to use closed American AI models to defend themselves, they were blocked by the models’ security. So they fought the attack with an open source model from China. The greater risk than powerful AI is that it is controlled by only a few companies and countries.
That’s a tricky proposition for OpenAi and Athropic, both of which sell closed models. The security system they propose instead, with extensive testing, verification and a coordinated pace at the border, is also the kind that few well-funded companies can most easily deliver.
The American technology industry is not necessarily united behind the CEOs either. Meta CEO Mark Zuckerberg told NBC News This week he explained that the industry does not need “some kind of industry-wide coordination”, arguing that there are “a lot of commercial incentives to get this right”. Nvidia CEO Jensen Huang, who is expected at the state dinner with Xi on Thursday evening, describes the AI fears as exaggerated. Trump’s position is much closer to hers than Altman’s or Amodei’s.