U.S. Open to Discussing AI 'Shared Risks' With China, Treasury Secretary Bessent Says

Treasury Secretary Scott Bessent opened a diplomatic door this week that few expected to find unlocked. In remarks that signal a shift in how Washington might approach artificial intelligence competition with China, Bessent indicated the United States would be willing to sit down and discuss what he called "shared risks" in AI development—a phrase that suggests both nations recognize common ground in the chaos of rapid technological change, even as they race to dominate the field.

The statement marks a notable recalibration. For months, the prevailing tone from U.S. officials has been one of technological competition bordering on confrontation: America must move faster, invest more, and secure its lead in AI before China does. Bessent's comments don't abandon that urgency, but they layer something else on top of it—the idea that some problems in AI safety and governance might be too large for any single nation to solve alone, and that talking to Beijing about those problems could be in everyone's interest.

At the same time, Bessent made clear that openness to dialogue does not mean softening on domestic policy. He pushed back against the notion that AI companies should receive blanket liability protections as they develop and deploy increasingly powerful systems. The Treasury chief's position is straightforward: if an AI lab causes harm, it should face legal consequences like any other business. There will be no special exemptions carved out for the industry, no get-out-of-court-free cards handed to companies simply because they are working in a novel domain.

Bessent also emphasized that the United States needs to accelerate its own development of open-source AI models—systems whose underlying code is publicly available and can be modified by researchers and companies outside the major labs. This is a competitive argument dressed in the language of democratization. Open-source models, he suggested, are essential to ensuring that AI capability is not concentrated in the hands of a few large corporations, and that American innovation remains distributed and resilient. The implication is clear: if China builds open-source models and the U.S. does not, America loses a crucial avenue for maintaining technological leadership.

What Bessent did not say is equally telling. He offered no specifics about what "shared risks" discussions with China might look like, what topics would be on the table, or when such talks might begin. He did not outline what the U.S. would be willing to concede or what it would demand in return. The statement is an opening gesture, not a roadmap. It signals that the door is ajar, but leaves the actual negotiation—if it happens—for later.

The remarks also contain an implicit acknowledgment that AI companies themselves have significant power over how this technology develops. When Bessent suggested that AI labs "could stop any time they want to," he was naming a reality that regulators have been grappling with: these companies are not passive actors waiting for government guidance. They are making choices about what to build, how fast to build it, and what safeguards to include. Holding them accountable through liability law is one mechanism for influencing those choices. But it is not the only one, and it may not be the most effective.

For now, Bessent's comments represent a signal that the U.S. government is thinking about AI governance in multiple registers at once—competing with China, yes, but also preparing for a world in which some aspects of AI development might benefit from international coordination. Whether that coordination actually materializes, and what form it would take, remains an open question.