Bessent proposes U.S.–China AI dialogue and incident notifications

U.S. Treasury Secretary Scott Bessent said Sunday’s talks with Chinese Vice Premier He Lifeng covered a U.S. proposal for an AI incident-notification mechanism at the national-security level, and that both sides discussed standing up a forum called the U.S.–China AI dialogue ahead of this week’s Trump–Xi summit.

POLITICS desk — The morning filing was “talks open.” This is the first on-the-record U.S. pitch for a named bilateral AI forum plus incident alerts before Trump and Xi meet in person. That turns AI safety from lab essays into a summit agenda item — while China’s silence means the proposal is not yet a deal.

Bessent told reporters the sides discussed setting up a mechanism “called the US-China AI dialogue,” and that they “agreed to meet again.” He also called the engagement “very successful.” File those named-forum / meet-again / very-successful lines as Bessent’s, via the same-day Straits Times carry of the AFP on-the-record remarks. An AI dialogue here means a standing bilateral forum — a named talking channel — not a treaty. This desk did not interview him.

Bessent framed the goal as moving “from opaque to more transparency” between the world’s top two AI powers. Reuters quotes him: “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.” File that opaque-to-transparency line as Bessent’s, via Reuters. Incident notification, as he described it, would cover AI-related incidents that rise to a national-security level — a heads-up between governments, not a public alarm and not a signed safety treaty.

China’s formal response to the notification proposal was unclear. Reuters says He Lifeng and China’s chief trade negotiator Li Chenggang left the JPMorgan talks without speaking to media. Do not read this card as Beijing accepting the alerts, signing a pact, or issuing a joint statement. File the silence as Reuters’. This desk did not catch them at the door.

USTR Jamieson Greer — the U.S. official who runs trade talks and tariff policy — said U.S. export controls on sophisticated AI chips and semiconductor manufacturing equipment were not on the agenda for the AI-mechanism talks. File that off-agenda line as Greer’s, via Reuters. This filing is the proposed dialogue and incident alerts, not a chip-export deal.

Greer said Sunday’s discussions prepare groundwork for a “successful summit” between Trump and Xi in Washington later in the week. Reuters and AFP both carry that “successful summit” line; AFP’s Straits Times carry dates the Washington meeting Thursday and Friday. File the groundwork line as Greer’s. This desk is not writing the summit outcome.

The meeting also covered trade items: a process agreed in May, called the Board of Trade, for each side to identify possible tariff cuts on non-strategic goods. Greer did not name a product list. Reuters and AFP both say Bessent and Greer did not report progress on the flow of critical minerals to U.S. companies. Do not invent a minerals breakthrough. File the Board of Trade picture and the no-minerals-update as the wires’. This desk did not sit on the tariff list.

Do not claim China accepted the notification mechanism, that a binding AI safety treaty was signed, or that the U.S.–China AI dialogue is already a standing institution with staff and a schedule. Bessent said they discussed standing it up and agreed to meet again. Reuters says the Chinese response to the notification proposal was unclear. A deal would have text and signatures. None were published.

Plain English for the rest of the card: USTR = U.S. Trade Representative, here Jamieson Greer — the official who runs U.S. trade talks and tariff policy. AI dialogue / U.S.–China AI dialogue = the named bilateral forum Bessent said the sides discussed standing up — a talking channel, not a signed treaty. incident notification / incident-notification mechanism = a proposed government-to-government heads-up when an AI-related event rises to national-security level. national-security level = serious enough that the two governments would treat it as a state concern, not a lab blog post. Board of Trade = the May process for spotting possible tariff cuts on non-strategic goods. non-strategic goods = items neither side treats as a security-sensitive export. This filing is the on-the-record U.S. pitch after the talks closed. A pact would have text. None was published.

CONFIRMED here: Reuters’ same-day New York report of Bessent’s and Greer’s on-the-record remarks after the JPMorgan talks closed — Tier B wire, via Devdiscourse syndication, not a Treasury or Ministry of Commerce newsroom PRIMARY — plus the same-day Straits Times carry of AFP’s “very successful” / “called the US-China AI dialogue” / “agreed to meet again” quotes, and Bloomberg’s same-cycle headline corroboration of the “very successful” talks and the AI-dialogue discussion. The notification-mechanism proposal for Trump and Xi, the opaque-to-transparency quote, He’s and Li’s silence, Greer’s export-controls-not-on-the-agenda line, the “successful summit” groundwork line, the Board of Trade tariff-cut process, and the lack of a reported critical-minerals update are Reuters-attributed. The “called the US-China AI dialogue,” “agreed to meet again,” and “very successful engagement” lines are Bessent-attributed via AFP / Straits Times. NOT claimed: that China accepted the notification mechanism, a signed AI safety treaty, that the dialogue is already a standing institution, a minerals breakthrough, that this desk sat in the room, a stock tip, or investment advice. Distinct from the already-filed bessent-he-ai-talks, khanna-ai-china-pact-ftn, and albanese-ai-us-china-deal.