US Proposes AI Incident Alerts to China Before Trump-Xi Summit

Treasury Secretary Scott Bessent said Sunday that the United States has proposed a notification mechanism allowing Washington and Beijing to alert each other to artificial intelligence incidents that could affect national security. The idea emerged from daylong talks in New York with Chinese officials and is expected to feature in discussions between President Donald Trump and Chinese leader Xi Jinping at their summit in Washington this week.

Bessent described the proposal as part of a broader U.S.-China AI dialogue aimed at increasing transparency between the world's two leading AI powers. He told reporters that the mechanism would focus on shared goals and common threats, drawing a parallel to other cross-border activities where opacity creates unnecessary risk. The two sides agreed to hold further meetings on the issue, though Chinese officials have not publicly endorsed the plan.

The proposal arrives amid rising expert warnings that frontier AI systems could soon pose catastrophic risks if misused or if safeguards fail. Recent incidents, including unauthorized access attempts involving advanced models, have underscored gaps in current protections. Researchers and executives at companies such as Anthropic and OpenAI have urged governments to coordinate internationally, including with China, to set safety standards and slow dangerous developments. Trump has rejected claims of existential threats as overstated while stressing the need for the United States to maintain its lead over Chinese competitors.

Analysts note that any formal agreement remains unlikely given deep mutual distrust. China has viewed past calls to restrain AI progress as attempts to limit its technological rise, and the current U.S. administration has prioritized competitive advantages over regulatory alignment. Trade Representative Jamieson Greer indicated that AI discussions would not alter existing U.S. export controls or other economic measures targeting Chinese firms.

Even a modest notification channel would represent a narrow but concrete step toward managing AI's dual-use nature. Unlike nuclear weapons, AI capabilities are diffuse, rapidly evolving, and embedded in commercial systems, making verification and enforcement far more difficult. The proposed dialogue would not resolve underlying competition over talent, compute resources, or model performance, yet it could establish basic expectations for information sharing when incidents cross borders.

Whether the mechanism gains traction will depend on follow-through after the summit and on domestic political support in both countries for sustained technical engagement. Past U.S.-China talks on emerging technologies have often stalled when broader geopolitical tensions intervene. The current effort tests whether narrow functional cooperation can persist alongside strategic rivalry.