The two governments also plan to share approaches to evaluating advanced AI systems before deployment. Bessent said the discussions would address model testing, safety practices and methods for ensuring that humans retain oversight of increasingly capable technology.
The initiative includes a communication mechanism through which either government could quickly contact the other following a significant AI model incident. However, Bessent did not identify the agencies responsible for managing the channel, explain what would qualify as a reportable incident or specify when the system would become operational.
The announcement follows an agreement reached during Chinese President Xi Jinping's September visit to establish formal AI discussions and a mechanism for addressing serious incidents. Bessent's latest remarks provide additional details about the areas of cooperation the two governments intend to pursue.
Despite the agreement, Bessent emphasized that cooperation on AI safety should not be interpreted as a reduction in competition between the two countries.
“We should harbor no illusions about the differences between the United States and China,” he said, acknowledging that the governments will continue to pursue different priorities even as they work together on shared risks.
The framework does not include a commitment to restrict the development of advanced AI systems. Instead, the discussions focus on communication, safety evaluations and maintaining human control over the technology.
That distinction reflects the Trump administration's broader approach to AI regulation. Washington has generally resisted restrictions that could weaken the United States' competitive position relative to China, although Trump recently indicated that some safeguards could be acceptable.
In an interview with CBS released Sunday, Trump said, “I don’t mind guardrails,” while cautioning against regulations that could leave American AI development behind China's.
Beijing has also expressed reservations about proposals to slow the development of advanced AI models. China's foreign ministry argued in September that confrontation and alarm over the technology could interfere with international efforts to establish AI governance.
Still, Chinese officials have acknowledged the importance of cooperation between the two countries on technological risks.
Xie Feng, China's ambassador to the United States, addressed the same gala and described bilateral cooperation as essential to addressing major international challenges, including AI governance.
The emerging dialogue comes amid continuing disputes over access to advanced AI technology. Washington has restricted exports of high-performance chips to China and accused Chinese developers of using American AI models to improve their own systems through large-scale distillation.
The United States also temporarily restricted foreign nationals' access to Anthropic's Mythos and Fable 5 models in June before reversing those controls several weeks later.
Those measures illustrate the competitive pressures surrounding the new safety discussions. While the governments are seeking ways to reduce risks associated with advanced AI, they remain divided over access to the technology and the policies governing its development.
Bessent addressed that tension directly, saying the two countries must find ways to balance technological competition with their responsibilities as major powers.
“Our interests will not always coincide, and our approach to every challenge likewise will not always converge,” he said.
Much of Bessent's speech also focused on the economic relationship between Washington and Beijing, particularly the trade imbalances that have contributed to tensions between the countries.
He reiterated the administration's objective of reducing economic vulnerabilities without severing trade ties, describing the approach as “to de-risk, not to decouple.”
Bessent argued that the United States needs to increase domestic production relative to consumption, while China should strengthen domestic demand for the goods it produces. He said a more balanced relationship could benefit both economies without preserving the trade disparities that have caused friction.
The Treasury secretary also emphasized that stronger American economic policies do not necessarily conflict with maintaining productive international trading relationships. He described balance, reciprocity and respect for national sovereignty as important principles for future engagement.
The remarks come ahead of another planned meeting between Trump and Xi in China next week, which Bessent is expected to attend. The leaders are scheduled to meet during the Asia-Pacific Economic Cooperation gathering on November 18 and 19.
The new AI dialogue adds a technology safety component to those broader diplomatic discussions. While the governments have outlined plans to exchange information and coordinate responses to serious incidents, the operational details remain unresolved.
For now, the agreement establishes a framework for communication rather than binding limits on AI development. Its practical significance will depend on how the two countries implement the incident notification system and whether their planned exchanges on model testing lead to sustained cooperation.
This analysis is based on reporting from CNBC.
Image courtesy of Missouri State.
This article was generated with AI assistance and reviewed for accuracy and quality.
