AI Safety Talks Advance as U.S.-China Trade Disputes Persist

U.S. and Chinese officials concluded approximately eight hours of talks in New York on September 20, focused on preparing for President Donald Trump and President Xi Jinping’s meeting in Washington later this week. Treasury Secretary Scott Bessent met Vice Premier He Lifeng, joined by U.S. Trade Representative Jamieson Greer and Chinese trade negotiator Li Chenggang. Bessent characterized the discussions as successful, while Chinese state media described them as candid, in-depth, and constructive. Xi is scheduled to visit the United States from September 23 through 25, with the leaders expected to discuss the bilateral relationship, trade, technology, and other strategic issues.

The new proposal was a U.S.-China dialogue on artificial intelligence that would include a notification mechanism for serious AI-related incidents with potential national-security implications. U.S. officials presented the idea as a means to increase transparency and provide a channel for communication on shared threats. Beijing confirmed that AI-related issues were part of the economic and trade discussions, but neither side announced that a mechanism had been agreed, defined its scope, or set a timetable for implementation. The proposal therefore remains an agenda item for further negotiation rather than an established bilateral arrangement.

The negotiations also addressed implementation of a process agreed in May to identify non-strategic goods that could receive lower tariffs. U.S. officials indicated that Chinese consumer and other low-technology goods, as well as U.S. energy, agricultural, and medical products, could be considered. The talks occurred against the backdrop of a tariff truce due to expire on November 10. The truce capped US tariffs on Chinese goods at roughly 20 percent after an earlier escalation drove reciprocal levies into the triple digits. No major agreement was announced on tariff reductions, and no reported progress emerged on US concerns regarding Chinese supplies of rare-earth magnets and other critical minerals.

U.S. officials also deliberately separated the AI safety proposal from the most contentious technology restrictions. Greer said export controls on advanced AI chips and semiconductor manufacturing equipment were not part of the proposed notification mechanism. This distinction indicates that the immediate discussions sought limited cooperation on AI risk management without resolving the broader dispute over technological competition and access to advanced computing capabilities. A working group was expected to continue discussions after the New York meeting.

The proposed notification mechanism would create a narrow form of crisis communication between the two leading AI powers without requiring either government to soften its competitive technology policies. Its potential value lies in reducing the risk that a serious AI incident with national-security consequences is interpreted through strategic suspicion or incomplete information. However, its practical significance will depend on whether the sides can define what constitutes a reportable incident, establish communication procedures, and demonstrate willingness to use the channel during a sensitive event. The absence of an announced agreement means that the proposal should be treated as a diplomatic opening, not as a functioning confidence-building measure.

The separation of AI safety discussions from export controls is strategically consequential. Washington can pursue limited cooperation on risks associated with AI while preserving restrictions on advanced chips and semiconductor manufacturing equipment, which remain outside the proposed mechanism. For Beijing, participation in an AI dialogue may offer a way to engage in governance and transparency without conceding the broader US approach to technology controls. This compartmentalization may make narrow agreements more achievable, but it also limits their ability to reduce the underlying rivalry in advanced technology.

Trade remains the nearer-term constraint on the summit’s prospects. The November 10 expiration of the tariff truce creates a fixed deadline and raises the practical importance of discussions on lower tariffs for non-strategic goods. A limited tariff process could provide both governments with a tangible outcome while leaving politically and strategically sensitive sectors outside any initial arrangement. Yet the lack of reported movement on rare-earth magnets and other critical minerals suggests that the issues carrying greater supply-chain and technology significance remain difficult to resolve.

The summit will test whether the two governments can stabilize selected parts of their relationship while managing disagreements that neither side appears prepared to settle. The clearest indicators will be whether Trump and Xi endorse a formal AI dialogue, extend or replace the tariff truce before November 10, and produce a concrete process for tariff relief or critical-mineral discussions. An outcome limited to continued talks would still preserve channels of communication, but it would leave businesses and investors facing uncertainty over tariffs, technology controls, and strategically important supply chains.