US-China AI hotline: a US proposal Beijing has not accepted, and what Trump and Xi left off the table

The United States has proposed a hotline with China for AI incidents serious enough to threaten national security, and China has not said yes. Treasury Secretary Scott Bessent put the idea to Vice Premier He Lifeng in New York on Sunday 20 September, and it goes to Donald Trump and Xi Jinping at their Washington summit on 24 and 25 September. A widely shared post on X from the MTS show said on 24 September that Bessent and He Lifeng had "already settled it". That claim is the one we could not confirm, and the gap between a proposal and a settlement is the whole story for now.

What is the US-China AI hotline?

It is a proposed notification mechanism: each government would warn the other when an AI-related incident rises to a national-security level. Bessent described it to reporters after the New York talks as part of a new US-China AI dialogue, built around "common goals and common threats", according to Reuters' account of his remarks.

His case for it was about visibility rather than restraint. "We think that, just like with any cross-border activity, that moving from opaque to more transparency between the number one and the number two AI powers in the world is very important," Bessent said. Trump and Xi first discussed consultations on AI in Beijing in May, Reuters notes, but that forum was never formalised.

Neither government has published what incidents the channel would cover or who would staff it. Reports listing specific categories exist, but we did not find one that attributes the list to either side, so we do not repeat it here.

Has China agreed to the AI hotline?

Not publicly. Xinhua's readout of the New York talks described "frank, in-depth and constructive exchanges on key economic and trade issues" and only briefly acknowledged that AI came up. He Lifeng and China's chief trade negotiator Li Chenggang left without speaking to reporters.

The most Beijing has said on the record came on the first day of the summit. China's commerce ministry told a press conference on 24 September that the two countries had held their first dialogue on AI, Fortune reports. A dialogue is what both sides can already point to. A hotline is what the US has asked for. Fortune's 22 September report put it the same way: the two sides agreed to hold further talks about a hotline, not to create one.

So when the MTS post says Bessent and He Lifeng "already settled it", it is reporting something the documents do not show. If the summit produces a joint statement that includes the mechanism, this item will be updated with it.

What did Trump say about superintelligence before meeting Xi?

He said he wants to leave it alone. On the morning of 24 September Trump posted on Truth Social: "Super Intelligence (SI) will be a big topic of discussion, but I want to leave it exactly where it is. That is China's position also. Our guardrail is the DOJ!"

That fits what he told the UN General Assembly on 22 September, when he said Washington would oppose any "globalist scheme" to control AI. Fortune points out the tension: two days after that speech, his own Treasury secretary was proposing a formal channel with China. The post names the Justice Department as the guardrail but does not say what DOJ oversight of superintelligence would involve, and nothing we read explains it.

Xi struck a more cooperative note at the White House. "China and the U.S. are leading nations in AI and we have both the capability and responsibility to develop and manage AI for good and ensure the development of AI is always under human control," he said, according to Bloomberg.

Are pacing, a recursive self-improvement ban and export controls on the table?

No, and on export controls that is on the record. US Trade Representative Jamieson Greer said export controls on advanced AI chips and chipmaking equipment were not on the agenda for the AI mechanism talks, Reuters reports.

On pacing, Al Jazeera reports that both governments have rebuffed calls to slow AI development, and that the only proposals either side has publicly floated are a dialogue and the notification mechanism. China, it reports, has dismissed Silicon Valley's existential-risk fears as "fearmongering", while Trump has suggested guardrails are not needed under his leadership.

A ban on recursive self-improvement, meaning AI systems that improve themselves, was never a government proposal in this round. It came from Representative Ro Khanna, the ranking Democrat on the House China committee, who said on CBS's Face the Nation that one of the "core principles" of a Trump-Xi agreement should be "banning any self-improving AI, recursive self-improving AI". Neither delegation has taken it up in anything we read.

Why would China distrust calls to pace the frontier?

Because Beijing tends to read AI safety through the chip fight. "Pacing the frontier" is the phrase of a July 2026 statement signed by 1,386 employees of frontier AI companies, which asks "that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development". Its signatories include Anthropic's Dario Amodei and OpenAI chief scientist Jakub Pachocki.

Analysts quoted by Al Jazeera explain why that can sound different in Beijing. "If every safety discussion is linked to export controls or slowing China's development, Beijing will naturally question the purpose of cooperation," said Sun Chenghao of Tsinghua University. Henry Gao of Singapore Management University put the gap in risk perception plainly: "In China, national security concerns appear more geopolitical: A major perceived danger is that the US could gain a decisive AI advantage and use it against China."

The MTS post goes one step further and says China may see US pacing as a way to cement the US lead, and that Amodei's talk of denying authoritarian regimes top AI capabilities has not helped. That is the post's reading. No Chinese official says it in the documents we read.

Why did Trump greet Xi at Joint Base Andrews?

As a mark of honour: Trump met Xi and Peng Liyuan on the tarmac at Joint Base Andrews on 23 September, rather than receiving them at the White House. Bloomberg calls it "a rare honor only extended to popes at that airport over the past six decades". CBS News described it as a presidential move not seen since 1962; the MTS post names that occasion as John F. Kennedy's welcome for British prime minister Harold Macmillan. The post's shorthand, "the first time a president has done that since" 1962, leaves out the papal exception.

Who is at the Trump-Xi state dinner?

The confirmed names so far are AI and chip executives plus one banker. CBS News reports that Nvidia's Jensen Huang, Qualcomm's Cristiano Amon and JPMorgan Chase's Jamie Dimon will attend, according to people familiar with their plans, and OpenAI's Chris Lehane confirmed to CBS that Sam Altman will be there.

Other outlets have reported Jeff Bezos, Elon Musk, Sundar Pichai, Michael Dell, Tim Cook and Mark Zuckerberg on the list. We did not confirm those against a first-hand source, and the MTS post itself marks Zuckerberg as "reportedly". Trump's own description was: "I would say you have the entire tech world, the entire banking world, and a lot more."

What happens after the Trump-Xi summit?

The summit ends on 25 September, and the test is whether a joint statement names the notification mechanism. The AI labs are asking for more than a hotline. At the UN Security Council on 23 September, Amodei called for small agreements first, such as a ban on using AI to make biological weapons, then systems to verify commitments, then a global alert system. "No leader, no company, and no nation can manage this alone," he said. Altman called for fast, shared incident reporting, Fortune reports.

The case for a harder agreement later is quoted in the MTS post from Theo Jaffee: "If in six months or a year when the two leaders meet again, AI has progressed substantially and has caused negative incidents in the real world, which inevitably it will, then I think there will be much stronger demand for pacing the frontier." That is a forecast, not a report. For now, the only AI measure either government has put on the table is a phone line, and only one of them has picked up the receiver.