Bessent: US and China should cooperate on AI, compete on tech

Treasury Secretary Scott Bessent said the U.S. and China should “harbor no illusions” about their differences as they engage to manage the risks of artificial intelligence while competing for dominance in the technology, CNBC reported.

In prepared remarks to the National Committee on U.S.-China Relations in New York, Bessent said the two governments have established a new dialogue mechanism on “super intelligence” aimed at guarding against shared risks and averting misunderstandings between the two systems.

The mechanism will see both sides exchanging best practices for testing advanced models before release and keeping human judgment in command of the technology. Bessent said he expects both sides to alert each other to serious incidents involving the models.

No limits on development

Bessent stopped short of committing to limit the advancement of AI. Beijing and Washington remain locked in a race for global dominance, and the Trump administration has resisted broadening restrictions on AI development over concerns about ceding a competitive edge to China.

Trump has recently signaled some openness to guardrails. “I don’t mind guardrails,” he said in a CBS interview released Sunday, but he added that he did not want regulations to put the U.S. in “a deep second position” behind China, which he said has no such restrictions.

China’s foreign ministry in September dismissed calls to slow down frontier models, saying that fearmongering and confrontation would only disrupt global AI governance. The position leaves both sides racing to build the most powerful systems while talking about cooperation on safety.

The summit context

The dialogue was agreed during Xi’s September visit to Washington, where the two leaders agreed to launch an AI communications channel to help defuse serious incidents. Trump and Xi are expected to meet for the third time next week on the sidelines of the Asia-Pacific Economic Cooperation Leaders’ Meeting from Nov. 18 to 19, with Bessent joining.

The summit takes place against the backdrop of the Jensen Huang scandal. The Nvidia CEO admitted to an extramarital affair with a Chinese businesswoman whom U.S. intelligence officials suspect of espionage. Huang joined Trump’s delegation to Beijing in May and was invited to a state dinner in Washington in September, but his presence at the upcoming summit has been thrown into doubt.

Other business leaders, including Tesla and SpaceX CEO Elon Musk and Mastercard’s Michael Miebach, could be on Trump’s delegation next week, according to the South China Morning Post. The business contingent reflects the tension between the security concerns and the commercial interests that tie the two economies together.

The technology restrictions

Washington has taken a series of actions to restrict Chinese access to advanced AI technology. The government briefly suspended access by foreign nationals to Anthropic’s Mythos and Fable 5 models in June before restoring it weeks later. The U.S. has also restricted exports of the most advanced chips to Beijing and accused Chinese developers of copying American models through what it calls industrial-scale distillation.

The restrictions are the stick; the dialogue is the carrot. Bessent’s speech was an attempt to establish a framework where the two countries can cooperate on safety while competing on development. The framework is sensible in theory, but the practice is complicated by the espionage concerns, the export controls, and the built-in tension between a race and a partnership.

Xie Feng, China’s ambassador to the U.S., told the audience at the gala that efforts by both countries were “indispensable” to addressing issues from global flashpoints to improving AI governance. Cooperation “may not resolve every challenge,” he said, but “none can be resolved without it.”

What I think

The dialogue is better than no dialogue. AI is a technology with catastrophic risk potential, and the two countries building the most advanced systems need a communication channel to prevent accidents. The Cold War had a hotline; the AI race needs one too. Bessent’s mechanism for alerting each other to serious incidents is a step in the right direction.

But the cooperation is limited by the competition. Neither side is willing to slow down development, and both are racing to build the most powerful systems. A safety dialogue that does not include any commitment to limit development is a dialogue about how to run faster safely, not about whether to slow down. The guardrails Trump mentioned are the right instinct, but his insistence that the U.S. not fall behind China means the guardrails will be minimal.

The Jensen Huang scandal is the complication. The Nvidia CEO is at the center of a national security story, and his role as a bridge between Silicon Valley and Beijing is now suspect. If Huang is excluded from the summit, it signals that the security concerns are overriding the commercial interests. If he is included, it signals that the commercial interests are too large to be derailed by the scandal. Either way, the AI dialogue is harder to conduct when the CEO of the most important chip company is under investigation.

The technology competition between the U.S. and China is the defining economic and security story of this decade. Bessent’s speech was an attempt to manage it, and the attempt is serious. But the structural tension between cooperation and competition is not resolvable with a dialogue mechanism. The two countries will cooperate on safety and compete on everything else, and the competition will be the stronger force.

The APEC summit next week is the test. If Trump and Xi can announce concrete progress on the AI dialogue, the mechanism has substance. If the meeting produces only platitudes, the dialogue is a forum for talk, not a tool for management. The stakes are high enough that even a forum for talk has value, but the technology is moving faster than the diplomacy.