Bessent demands China get tougher on trade, gets refused, proposes exchanging AI safety alerts with them anyway.

…(cont) On September 10th, Beijing sent a friendly note ahead of this week's Trump-Xi meeting, saying it hoped the trade talks would resolve soon. No number attached. No concession attached. Just a hope, on the record, costing nothing to issue.

Then, this Sunday, Bessent sat down with Vice Premier He Lifeng at the new JPMorgan Chase headquarters in midtown Manhattan, three weeks after the tougher-on-China speech, to talk trade ahead of the summit. What came out of that room was not a surplus target and it was not a tariff number. It was a proposal: the United States and China would exchange notifications about AI safety risks, a shared alert system, so that each side tells the other when an artificial intelligence model does something worth worrying about.

Run the two filings side by side and the gap has a size. The opening ask was denominated in trillions of dollars a year, owed by one country to the rest of the world's ledger. The closing offer is denominated in phone calls, or their AI equivalent, texts, about which model said what. An auditor asked to reconcile those two documents would tell you they don't share a currency, and they don't close each other out.

Nobody here misspoke. Bessent said what he said in Asheville; the meeting at JPMorgan happened as described; the alert proposal is real and on the table for the summit. The gap isn't a lie, it's a swap: a fight over a trillion-dollar imbalance became, three weeks later, a conversation about who calls whom first when a model misbehaves. That's a real thing worth having. It is not the thing that was asked for on August 31st. The surplus, last we checked the file, is still $1.2 trillion.