Scott Bessent emphasizes US-China AI cooperation amid differences

The world’s two largest AI powers want to compete hard and cooperate carefully at the same time. US Treasury Secretary Scott Bessent laid out that balancing act on Tuesday, November 10, in New York.

His message was that the United States and China should “harbor no illusions” about where they disagree on artificial intelligence. They should still find ways to work together on the technology’s risks.

A new channel for superintelligence talks

Bessent made the remarks at the National Committee on US-China Relations gala. He said the two countries need to manage AI risks jointly even as they race each other for technological leadership.

The centerpiece is a new bilateral dialogue focused on “superintelligence,” which Bessent referred to as SI.

The dialogue is meant to let both sides swap best practices for evaluating and testing advanced AI models. It also covers keeping humans in control of those systems.

The framework also includes a direct communication channel. Its job is to alert both nations about serious AI incidents.

That incident-alert mechanism was reportedly agreed during a White House meeting in September 2026. Bessent pointed to it as an example of the proactive oversight he thinks the moment requires.

Shenzhen is the next stop

Bessent said he will join President Donald Trump next week for a meeting with Chinese President Xi Jinping. It will be the third Trump-Xi meeting this year.

The talks will take place in Shenzhen, on the sidelines of the Asia-Pacific Economic Cooperation leaders summit on November 18-19. According to Bessent, the summit will push the new superintelligence dialogue forward.

The first two meetings this year took place in Beijing and Washington. The upcoming session continues that pattern of alternating high-level contact.

He characterized the frequency of US-China leader engagement in 2026 as historic. In his view, that rhythm lays a foundation for future cooperation.

De-risk, not decouple

The dialogue reflects the administration’s stated approach to China: “de-risk, not decouple.” The idea is to reduce dangerous dependencies without severing ties entirely.

On AI, that translates into a specific goal. The government does not want to slow American AI progress, but it does want guardrails in place where the risks are shared.

What this means

For policymakers, the most concrete piece is the incident-alert line. A working channel between Washington and Beijing could matter most during a fast-moving AI failure, when confusion is its own hazard.

The November 18-19 summit is the next real test. What comes out of Shenzhen on superintelligence, testing and incident alerts will show whether this framework has substance.