The U.S. and China are taking a different approach to artificial intelligence competition: instead of imposing blanket export bans, they're now talking directly about AI safety risks that could affect both nations. On September 24, China's Ministry of Commerce confirmed that officials from both countries had held their first talks on artificial intelligence, describing the exchanges as constructive and candid. This emerging dialogue represents a notable pivot from the export control battles that have dominated U.S.-China tech relations for years.

What Changed in How the U.S. and China View AI Competition?

For years, the dominant strategy in Washington has been to restrict China's access to advanced semiconductor technology and artificial intelligence capabilities through export controls. But during Chinese President Xi Jinping's visit to Washington on September 24, both leaders signaled a different priority. President Xi called for AI to remain "under human control," while President Trump said he did not want the two countries to negotiate limits on AI development. This suggests both sides may be moving away from the zero-sum competition model that has defined recent tech relations.

The shift became concrete when Treasury Secretary Scott Bessent proposed establishing a bilateral notification mechanism for AI-related incidents that rise to the level of national security during talks with Vice Premier He Lifeng in New York beginning September 20. Rather than trying to prevent each other from developing AI, the two nations are now exploring how to alert each other when AI systems pose genuine risks to either country.

Which AI Safety Risks Are Now on the Table?

The emerging dialogue is still being defined, but officials are already discussing specific threats. On September 23, Bessent said the two sides were still working to define which AI risks the dialogue would cover, such as uncontrolled AI agents or non-state actors using AI for cyber operations or biological threats. These are not theoretical concerns; they represent real scenarios that security experts worry could emerge as AI systems become more autonomous and capable.

The timing of this dialogue is significant because it comes amid ongoing disputes over chip access and allegations that Chinese developers have improperly accessed U.S. AI models, which Beijing has rejected. Yet rather than letting these tensions derail cooperation entirely, both sides appear willing to compartmentalize the conversation, separating commercial competition from existential safety concerns.

How Are the Two Countries Structuring This New Dialogue?

• Notification Mechanism: The U.S. proposed a bilateral system where both nations alert each other when AI-related incidents reach the level of national security concern, allowing for rapid communication during potential crises.
• Scope Definition: Officials are actively working to clarify which specific AI risks fall under the dialogue, including uncontrolled AI agents, non-state actors weaponizing AI for cyber operations, and potential biological threats enabled by AI.
• Ongoing Negotiations: The talks between Treasury Secretary Bessent and Vice Premier He Lifeng represent the first formal channel for these discussions, with both sides committed to continuing the conversation beyond the initial exchange.

This approach mirrors crisis communication frameworks used in other sensitive domains, like nuclear weapons or space operations, where adversaries maintain dialogue channels even during periods of tension. The idea is that direct communication about catastrophic risks can prevent miscalculation or escalation, even when the countries remain competitors in other areas.

The dialogue also reflects a growing recognition among policymakers in both nations that some AI risks transcend national borders. An uncontrolled AI system that causes harm in one country could have ripple effects globally, making mutual awareness and transparency in the interest of both the U.S. and China. This is a departure from the purely protectionist stance that has dominated recent years, where the focus was almost entirely on preventing technology transfer and maintaining competitive advantage.

Meanwhile, the broader trade relationship between the two countries continues to evolve. The U.S. and China extended their trade truce by two months to January 10, following the unscheduled meeting between Bessent and He Lifeng in Washington. This extension provides a window for the AI dialogue to develop without the pressure of imminent tariff escalation, suggesting both sides are treating this conversation as a genuine priority rather than a negotiating tactic.

The emergence of this AI safety dialogue does not mean export controls are disappearing. Congress continues to push for tighter China-related trade restrictions, and the fundamental competition over advanced semiconductors remains intense. But the parallel track of safety-focused dialogue suggests that policymakers in both Washington and Beijing are beginning to recognize that some aspects of AI competition require cooperation rather than confrontation. Whether this dialogue can mature into a durable framework for managing AI risks will likely shape how the U.S.-China relationship evolves over the next several years.