The US and China agree on a super-intelligence incident channel, with the next round set for November

Beijing's version is shorter. China's Foreign Ministry put out a one-page readout on Saturday saying the two sides agreed to sign a memorandum of understanding on strengthening crisis communications and preventing crises between the two militaries, and to set up a communication mechanism for AI-related incidents. Same mechanism, different vocabulary: Washington says super intelligence, Beijing says artificial intelligence. That is not wordplay. Super intelligence is a rebrand the Trump administration has pushed, and according to The Independent the president has instructed government departments to use the term in posts and documents.

Trump's remarks outside the White House are worth reading word for word. He said: The United States of America is not going to be putting on brakes. Then he added: I would rather not integrate because we're leading by a lot. When you're leading, you don't open it up to each other. We're leading by at least a year, maybe a year and a half. Some people say two years. I'm not looking to open it up. In plain terms: the mechanism exists, but Washington is saying explicitly that it will not slow down because of it, and will not open up its technology on a reciprocal basis.

The most specific description of the channel itself appeared in the run-up, not the announcement. A PBS NewsHour broadcast on September 24 summarized Treasury Secretary Scott Bessent's proposal as a dedicated communications line to report any national security AI incidents, plus what he called a kind of list of agreed dangers, with examples including out-of-control models and terrorists using AI for cyberattacks. Be careful here: that was a proposal-stage description, not the final text. What was actually published is the White House fact sheet, and it contains one line about establishing a bilateral communication channel for SI incidents, with no itemized list.

China's public position on AI boils down to one sentence. According to The Independent, Xi told Trump that both countries are leading nations in artificial intelligence and that they must ensure the development of AI is always under human control and serves the wellbeing of the people. The same report notes that Trump had earlier dismissed AI safety fears at the UN General Assembly as a hoax, and had written on Truth Social: WHOEVER WINS AI, WINS! Put those two sentences side by side and you can see why analysts characterized the summit as producing no major breakthroughs, but useful working groups.

A second voice arrived in the same window. In an interview aired Sunday on NBC's Meet the Press, Bill Gates called on the US to lead on global AI regulation and offered an analogy: getting countries to agree on a global framework could be more difficult than Cold War-era negotiations on limiting nuclear weapons. His words: This is more difficult than that was. He also argued that imposing monitoring systems on AI companies would not handicap the US in its competition with China, and said he believed that after hearing from him and others, we will come to a consensus. A line from the same interview: it is not enough to have a kill switch.

Placed in the context of the past month, the escalation of AI safety as a topic is traceable. NBC's report lays out the line: Jacob Coxon, a whistleblower who had recently resigned from Anthropic, said publicly that the people building AI earnestly believe it could kill us all by the end of the decade. Days later, Dario Amodei of Anthropic and Sam Altman of OpenAI both agreed there is a need to slow the pace of development; Amodei wrote in a blog post that frontier companies must slow down, but only to the point where they still lead China, and urged coordination between democratic and authoritarian governments. Both companies later announced plans to track the behavior of their AI agents.

The summit was not only about AI. Per the same AP wire, the two sides agreed to keep cooperating through a Board of Trade focused on selected bilateral issues, which the White House said began operating this week; agreed to extend a trade truce by two months; China agreed to import at least 10 million metric tons of coal from the US in 2027 and 2028; the tariff reductions are aimed mostly at non-sensitive goods such as agricultural products and holiday decorations; and both sides said they would keep working on American concerns about rare earth supply. They agreed to meet again at the APEC summit in Shenzhen in November, and later at the G20 in Florida.

One note on how to read news like this. What came out of the summit is a mechanism, not a commitment: one incident channel, one next round scheduled for November, one MOU on military crisis communications. The value of a mechanism is that a phone line exists when something goes wrong. A mechanism does not change positions, though: Trump's no brakes and Xi's under human control still sit side by side. To judge whether it has teeth, watch for a specific incident report being made public after that November round, not for which noun each side chose this week.