• The U.S. and China have agreed to explore a standing AI dialogue and a notification system for AI incidents with national-security implications, according to Treasury Secretary Scott Bessent.
  • Talks in New York on September 20 also produced a "Board of Trade" process to identify non-strategic goods for potential tariff relief, ahead of a Trump–Xi summit on September 24.
  • Beijing's readout acknowledged AI discussions but stopped short of endorsing the U.S. proposal, leaving a formal mechanism unconfirmed. A bilateral tariff truce expires November 10.

Fragile Progress

Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng concluded high-level talks in New York on September 20, with Bessent describing the engagement as successful and saying the two sides discussed a standing U.S.–China AI Dialogue. The U.S. proposed an AI-safety notification system intended to increase transparency around AI-related incidents that reach a national-security threshold, according to people familiar with the matter. Beijing's public readout acknowledged AI discussions but did not explicitly endorse the U.S. proposal, so an agreed mechanism has not yet been confirmed.

"We discussed economics and AI," Bessent told reporters, adding that the talks were constructive. The meeting was framed as preparation for the Trump–Xi summit scheduled for September 24, where leaders could endorse a dialogue, instruct officials to negotiate a notification protocol, or announce limited tariff steps.

Tariff Truce and Trade Board

U.S. Trade Representative Jamieson Greer said the countries have operationalized a "Board of Trade" process to identify a relatively small group of non-strategic products for potential tariff reductions. Examples cited include Chinese consumer and low-tech goods, alongside U.S. energy, agricultural products, and possibly medical devices. The immediate economic pressure point is a bilateral tariff truce due to expire November 10, and markets will watch for concrete language on that deadline.

Rare-earth supply remains a sticking point. U.S. officials say China's critical-minerals flows have remained inadequate despite commitments made under last year's truce. These materials are important for technology, electronics, vehicles, and defense-related supply chains. Greer noted that advanced AI chips and chipmaking-equipment export controls were not on the agenda for the proposed AI-notification mechanism.

Strategic Fault Lines

The proposal sits at the intersection of technological competition and risk management. Washington is pursuing AI cooperation selectively: Bessent has said the U.S. wants to discuss shared risks and avoid a full bifurcation of the two AI ecosystems, including issues involving both open-weight and closed-weight models. At the same time, the U.S. continues to view Chinese capabilities in advanced chips, AI, critical minerals, and industrial capacity through a national-security lens. The current dialogue does not signal that export controls or broader technology restrictions are being rolled back.

China's public response emphasized "frank, in-depth and constructive" economic and trade exchanges and implementation of earlier understandings, but was comparatively noncommittal on the U.S. AI-notification proposal. That difference matters: it suggests agreement to keep talking, not agreement on substantive AI rules. Other unresolved bilateral issues include Chinese commitments to raise purchases of U.S. agricultural products by $17 billion annually and to buy more than 200 Boeing (BA) aircraft; the latest talks did not report progress on either item.

Market and Industry Implications

A credible notification channel would lower risk that a serious AI incident—such as a cyberattack affecting essential infrastructure or a dangerous misuse of a powerful model—escalates into a larger interstate crisis, according to analysts. AI developers and cloud providers could face future expectations for incident reporting, testing, model-security controls, and accountability. Workers and exporters could gain from a steadier trade environment, particularly in agriculture, manufacturing, logistics, and commodity-linked sectors. Consumers may benefit if tariff reductions lower costs for certain consumer products, though no product list or tariff agreement has yet been announced.

Security advocates are likely to welcome direct communication between the two largest AI powers but remain concerned that voluntary dialogue is insufficient without verifiable safeguards. Technology-national-security advocates may object if cooperation is viewed as enabling Chinese AI progress or weakening U.S. technological advantages. A key policy divide is whether AI governance should remain largely voluntary and national-security focused—as the current administration has favored—or move toward stronger binding requirements on AI developers.

Historical Context and Outlook

The countries first discussed possible AI consultations during the Trump–Xi meeting in Beijing in May, but the forum was never formally established. The New York discussions are therefore the most concrete step so far toward an ongoing channel. They come after years of tariff disputes, export controls, investment restrictions, and supply-chain pressure involving rare earths and semiconductors.

Analysts cited by Reuters see limited scope for major breakthroughs on short notice, but note that both governments have a strong incentive to keep trade tensions from reaccelerating. The most likely short-term outcome is continued engagement and incremental measures, rather than resolution of export controls, tariff architecture, rare-earth restrictions, or competitive AI policy. Longer term, a sustained AI channel could eventually address more difficult matters: AI weaponization, safeguards against misuse in biological or nuclear contexts, protection of critical infrastructure, and prevention of AI-enabled cyberattacks. George Chen of The Asia Group cautioned that cooperation will remain constrained by low mutual trust and Beijing's view that U.S. policy seeks to contain China's technological development.

The central takeaway is that the U.S. and China are attempting to build limited guardrails around AI and trade without resolving their strategic rivalry. Whether that produces durable risk reduction will depend on whether the forthcoming summit converts broad discussions into an agreed process with regular meetings, clear thresholds, and implementation commitments.

Correction: An earlier version of this article misstated the date of the Trump–Xi summit. It is scheduled for September 24, not September 20.