Bessent and He put open-weight models on the AI agenda
Washington and Beijing sat down in New York on Sunday to write the AI chapter of this week's Trump–Xi summit — and for the first time the open-weight question is inside the room, next to the tariff clock.
Treasury Secretary Scott Bessent and Vice Premier He Lifeng opened a day of talks at JPMorgan's Manhattan headquarters on Sunday, aiming to tee up agreements on AI, tariffs and critical minerals for the Trump–Xi summit on September 24. Bessent said before the meeting that the AI discussion would cover "both open- and closed-weight models," and that Washington is "open to discussions on avoiding shared risks and avoiding bifurcation of our two systems." He has spent the past week calling for US–China AI "guard rails" aimed at keeping powerful models out of the hands of malign non-state actors.
That last phrase is doing more work than it looks. Open weights are exactly the distribution channel a non-state actor can use without anyone's permission — no API key, no account, no terms of service to violate. Putting "open-weight models" on the agenda of a summit whose other half is a tariff truce means the US is preparing to negotiate the same category of artifact it is simultaneously trying to keep cheap and plentiful at home. The contradiction is not new, but it has never had a chair at the table.
The AI track is riding a trade clock: the US–China truce expires November 10, and a senior US official said on Friday that China's rare-earth magnet flows "have not been up to par." Rare earths matter here because they feed the semiconductor supply chain that the AI buildout sits on, and China's rare-earth leverage has been the enforcement mechanism behind every truce extension so far. Analysts expect small deliverables rather than a breakthrough. Anna Ashton of Ashton Intelligence told reporters the summit timing guarantees "some show of deliverables," but that "status quo is probably both sides' general best expectation."
What makes this round different is that a domestic political counter-proposal is already on the record. Representative Ro Khanna, the ranking Democrat on the House select committee on China, told CBS's Face the Nation on Sunday that an AI agreement should be the first agenda item between Trump and Xi, built on two principles: banning recursive self-improving AI, and safeguards against biological or nuclear misuse. He wants verification through data centers — monitoring that can tell whether a model is being used for training or inference — and invoked Reagan's "trust but verify." Khanna also confirmed that the letters his committee sent to DeepSeek and Alibaba, asking them to match American labs' pacing commitments, have gone unanswered. That silence is the honest state of play: the labs have made no commitments, and the governments are negotiating over monitoring they have not yet agreed to build.
We covered the opening of this channel when it was only a scheduled meeting — US, China schedule first dedicated AI safety talks of Trump's second term.
What to watch: whether the summit produces any AI language with a verification mechanism attached, and whether "bifurcation" survives contact with a rare-earths negotiation that both sides can measure.
If verification runs through data centers, who inspects whose? Tell us in the comments.
