Trump and Xi agreed to a bilateral AI ‘incident’ channel -- and to start calling it ‘super intelligence’

The Trump administration published its own account of the summit's AI outcome on Friday, a day after the event itself closed with only a state dinner and China's separate readout. The White House said the two governments agreed to set up a “bilateral communication channel” for AI-related incidents, alongside a dedicated AI dialogue scheduled for November. That fills in the gap this newsroom flagged when the summit ended Thursday: a vague dialogue pledge with no U.S.-side account to check it against. There still isn't a document defining what counts as a reportable incident, who sits on either side of the channel, or what the November session will actually cover.

The more concrete detail to come out of the readout is a naming choice. Trump told reporters the two leaders agreed to call the technology “super intelligence” instead of “artificial intelligence” going forward — SI, in his own shorthand. “I call it SI because it's a much better name,” he said. “It's much more accurate,” since “artificial”, in his framing, implies something false. It is a rebranding, not a technical redefinition, and the White House's account is the only place the term “super intelligence” has actually been used to describe what the two countries agreed to — nothing attributed to Xi or a Chinese government statement uses it.

“I call it SI because it's a much better name. It's much more accurate.” — President Trump, on renaming AI to ‘super intelligence’ in the joint framing with Xi

The cooperative framing sits next to a competitive one Trump stated just as plainly. He said the U.S. would limit what it shares with China under the new channel and would not slow its own AI development to match any joint understanding: “The United States of America is not going to be putting on brakes,” he told reporters, pointing to American technological leadership as the reason. Xi's own public framing, by contrast, called for “a healthy competition” and said both countries share “the capability and responsibility to develop and manage AI for good” — language that reads as aspirational rather than as agreement to any specific limit.

That silence matters more here than in most joint-statement stories, because the summit's own reporting already primed skepticism about how much was actually agreed. Al Jazeera's correspondent covering the three-day visit described it as “a lot of ceremony and a lot of pomp” with “a lot more pomp than progress,” and framed the continuation of dialogue itself — not any specific deliverable — as “perhaps the biggest deliverable for both sides.” That reading lines up with what didn't move: Washington's export limits on advanced Nvidia chips bound for China are unchanged, and the three chip-export bills Senate Democrats spent the same week publicly pushing remain stuck inside a defense-bill rider with no vote scheduled before the midterms.

(The two governments also used the same summit to extend their trade truce by two months, agree a military crisis-communications memorandum, and settle a coal-import commitment from China — the AI channel is one line item in a broader normalization push, not a standalone breakthrough.) None of that broader deal-making resolved the underlying disagreement driving most of this year's AI-policy friction between the two governments: export controls on the advanced chips that train frontier models. A loophole this newsroom traced earlier this month, running through a renamed, blacklisted server maker, remains open regardless of any incident-reporting channel — the two are separate tracks, and Friday's readout does not claim to touch the second one.

That's also the domestic backdrop this agreement lands on. Senate Democrats spent the same week publicly pushing three separate chip-export bills, all still sitting inside a defense-authorization rider that isn't scheduled for a floor vote until after the November midterms — a timeline this newsroom traced in detail after the earlier NDAA gridlock. A bilateral AI-incident channel does not touch that legislative fight at all; it is an executive-to-executive arrangement between two governments, not a change to any of the export-control statutes Congress is separately trying to tighten or loosen. Reading the two as connected — as if a cooperative gesture on incident-reporting signals movement on chips — is exactly the kind of inference Friday's readout gives no basis for.

None of that makes the channel meaningless — a standing line for reporting AI incidents between the two governments most responsible for frontier development is a real, if modest, guardrail against the kind of miscalculation both sides say they want to avoid. It just isn't the breakthrough either government's language implies, and the gap between the two is worth holding onto until November actually produces an agenda.

What actually changed this week, in other words, is narrower than either government's language suggests: a channel exists on paper, a date is set for November, and one side has attached a new name to the technology it covers. Whether an “AI incident” gets reported through it before November, and whether Beijing ever calls the technology anything but 人工智能, are the two concrete tests this agreement now has to clear — neither has been set up yet, on the public record.