Scott Bessent and He Lifeng shake hands in Geneva in 2025. File photo, not the September 2026 New York talks. U.S. Department of the Treasury / Wikimedia Commons, public domain.
.jpg){kind=link}
U.S. and Chinese economic teams opened talks in New York on September 20 with tariffs, rare-earth supplies and artificial intelligence safety on the agenda. The addition of AI raises a question that extends beyond trade: can the two countries discuss risks they may share while continuing to compete over the technology?
China’s Commerce Ministry confirmed that consultations began on Sunday morning local time. The U.S. Trade Representative’s office had announced that Jamieson Greer would join Treasury Secretary Scott Bessent in meeting Chinese Vice Premier He Lifeng ahead of the leaders’ summit in Washington.
Reuters reported that the agenda included the trade truce, critical-mineral flows and potential safeguards for AI. It also reported that U.S. officials considered Chinese supplies of rare-earth magnets and critical minerals insufficient. That is the U.S. assessment of supplies; it does not establish that every industry using those materials has run out.
What would AI safety talks cover?
Bessent told Axios before the meeting that Washington was willing to discuss shared AI risks, including open- and closed-weight models. This describes the scope the U.S. wanted to discuss. It does not establish which specific measures China has accepted.
In practical terms, willingness to talk is only a starting point. A workable arrangement would need to identify the risks covered, the officials responsible for contact, whether companies participate, and how information would be exchanged when a problem occurs.
Those are questions for assessing any future agreement, not arrangements announced by this round of talks. Without that detail, readers cannot tell whether a statement promises further discussion or creates a process that officials and companies can actually use.
Separate the results on tariffs, minerals and AI
Greer’s pre-meeting statement emphasized monitoring existing commitments, trade in non-sensitive goods and access to markets for American farmers, manufacturers and workers. These concern the terms on which goods can be sold. AI risk cooperation would require different kinds of technical and communication arrangements.
Openness to AI dialogue therefore does not, by itself, mean tariffs will fall, mineral export restrictions will ease or other technology controls will be removed. Each change needs its own published terms and implementation date.
The same distinction matters when reading a summit announcement. A pledge to keep talking, an agreed policy change and a measure already in force are three different things. Reporting them separately would help businesses understand what has changed and what is still uncertain.
What Asian businesses need to know next
For businesses across Asia, the useful follow-up is specific: which goods face different conditions, when those conditions take effect, and whether AI cooperation introduces concrete requirements. A political statement alone is too little information to justify changing procurement, shipping or technology deployment plans.
At this article’s September 21 review, the opening of talks and the participants’ stated positions had been confirmed, but a final agreement had not been verified in the public material reviewed. That is a limit of the available evidence, not a claim that the talks have ended without results.
The next assessment should follow the actual text released by both sides and any instructions issued to implement it. Until then, an item appearing on the agenda should not be presented as a deal already reached.
Analysis note: The discussion of implementation and business implications is the author’s analysis. The questions above are not reported provisions of an agreement.
Sources
China Ministry of Commerce / 中國商務部：September 20, 2026
Reuters / MarketScreener：September 20, 2026