Bessent and He Lifeng trade AI safety for hard leverage
By Tom Beckwith · Reporting from Washington ·
Treasury Secretary Scott Bessent and Vice Premier He Lifeng meet in New York to discuss AI and trade, a technocratic effort to prevent escalation while the U.S. maintains punitive leverage.
The upcoming meeting between Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng is not a diplomatic thaw. It is a cold-blooded exercise in risk management. Washington is correct to treat this as a transaction rather than a partnership. By pairing talks on AI safety with hard punitive measures, the United States is attempting to build circuit breakers into a relationship that remains fundamentally adversarial.
The technocratic circuit breaker
Bessent and He Lifeng will meet Sunday, September 20, 2026. According to CNA, the venue is JPMorgan Chase headquarters in Manhattan. This choice is a matter of security, not symbolism. The two men are there to discuss AI security and trade before President Donald Trump hosts President Xi Jinping at the White House.
The goal is to prevent a technical accident from triggering a geopolitical catastrophe. Time Magazine reports proposals for technical exchanges on AI models that could help amateurs develop bioweapons. There is also a push for a government-to-government crisis-communication protocol. These channels would address incidents like AI-powered drones attacking a U.S. Navy vessel. The proposal suggests written communications to fit the Chinese system of committees.
This approach mirrors the Cybersecurity Agreement of 2015. That agreement used targeted, executive-level deals to mitigate high-tech risks and stop intellectual property theft. Both capitals now face runaway technical capabilities they cannot fully control. This creates a shared, self-interested need for rudimentary guardrails. It is a return to the logic of the U.S.-China Strategic and Economic Dialogue, where technical frictions were managed to prevent systemic collapse.
The price of admission
Beijing will hope these talks signal a softening of U.S. policy. They are mistaken. The New York huddle occurs while the U.S. maintains export controls on advanced chips and chipmaking equipment. More importantly, the talks are fenced by hard leverage.
On September 11, 2026, President Trump signed Presidential Determination No. 2026-23. This document names the PRC as the world's largest producer of precursor chemicals used for synthetic drugs. This is not a mere talking point; it is a formal designation of a drug-producing country. Furthermore, the U.S. suspended duty-free de minimis treatment for all countries on February 24, 2026. This move stripped Chinese e-commerce shippers of their primary route to bypass enforcement.
He Lifeng is the right man for Beijing to send. He is a member of the Political Bureau of the Communist Party of China Central Committee. As noted by news.cgtn.com, he leads the delegation for consultations on "economic and trade issues of mutual concern." Because of his close historical ties to Xi Jinping, He carries the direct authority of the party leader. He is not there to brainstorm; he is there to see what the U.S. will accept in exchange for more aggressive action on drug precursors.
The cost of miscalculation
The danger here is the gap between technocratic goals and political theater. President Donald Trump has dismissed runaway AI as a "hoax." Meanwhile, a state-owned China Daily editorial accused American firms of using safety calls to "blunt China’s AI advance." When the leaders speak in slogans, the professionals in the room are often overruled.
We have seen this pattern before. The U.S.-China trade war of 2018-2019 showed that tariffs and export controls are durable, while agreements are fragile. If Washington treats AI safety as a bargaining chip for drug precursors, it risks a miscalculation. If Beijing views safety protocols as a Trojan horse for espionage, the channels will freeze.
Five years from now, the cost of failure will be an automated escalation. An unauthorized cyber operation or a drone skirmish could spiral because the "written protocols" were ignored in favor of political posturing. The professionals are trying to build a floor under the relationship. The politicians are still arguing over who owns the building.
The New York talks will likely produce a narrow, technical framework for AI safety that is carefully decoupled from semiconductor controls. This is the only viable path. Any attempt to link the two would be a mistake. Washington must continue to use its punitive architecture to force concessions on narcotics while maintaining the technical valves that prevent a digital spark from starting a kinetic fire.