The United States is open to discussing shared risks with China in upcoming artificial intelligence talks this weekend, Treasury Secretary Scott Bessent said, according to Axios. The statement marks a notable softening in Washington's public posture on AI safety diplomacy — and a striking contrast with President Donald Trump, who has dismissed AI safety fears as a "hoax."
Bessent's comments, delivered Wednesday and reported by Axios journalists Marc Caputo and Maria Curi, signal that the administration is willing to engage Beijing on the risks posed by rapidly advancing AI systems, even as it rejects the precautionary framing favored by many AI safety advocates. For more context on this story, see our ongoing breaking AI news.
What the talks will cover
The weekend discussions are expected to address concerns over the bifurcation of the global AI ecosystem — the prospect of entirely separate American and Chinese AI stacks — and will include both open and closed model issues, according to reports citing Bessent's statement. The Treasury secretary is set to meet with a Chinese Vice Premier as part of the broader economic dialogue between the two countries.
Bifurcation has become one of the quieter worries in global AI policy. Export controls have progressively cut Chinese labs off from advanced American chips and models, while Beijing has pushed domestic alternatives. A clean split, some analysts argue, would entrench two rival ecosystems with fewer channels for coordination if either side's systems begin misbehaving in ways that cross borders.
A gap inside the administration
The Treasury secretary's openness to risk discussions sits awkwardly beside the president's own rhetoric. On Monday, Trump called AI safety concerns a "hoax" during an unscripted speakerphone appearance at the All-In Summit in Los Angeles, telling Nvidia CEO Jensen Huang live on stage that warnings about the technology were playing "right in the hands" of America's rivals. "That could be political people. It could also be China. And we're not going to let that happen. It's a hoax," Trump said, according to video of the exchange. Huang agreed from the stage, replying: "You're right. We're not going to let that happen, sir."
The president's framing treats safety advocacy as a strategic weapon aimed at slowing America down. That view directly opposes the position staked out by Anthropic CEO Dario Amodei, who called over the weekend for the pace of AI capability development to slow and be more closely monitored — a stance OpenAI's Sam Altman and Elon Musk have publicly shared. The weekend of warnings from leading American AI executives, in fact, unsettled markets: U.S. tech stocks tumbled in its aftermath, per CNBC reporting.
Bessent's framing threads the needle between those camps. By emphasizing "shared risks" — a term that centers mutual threats like AI-enabled cyberattacks, fraud, and loss of control — rather than safety regulation, the administration can engage Beijing without conceding the domestic regulatory argument. Whether Chinese officials accept that framing is another question.
Beijing's preconditions
China has already signaled what it wants from such talks. In late August, a social media account affiliated with Chinese state broadcaster CCTV used to signal official policy thinking said the US must prove its AI companies are subject to the same safety, disclosure and audit rules as anyone else before "substantive" discussions can begin — while accusing Anthropic of misconduct in a combative preamble.
Beijing's intervention named specific conditions: the Yuyuantantian post said "substantive" discussions required prior proof that American AI companies operate under equivalent safety, disclosure and audit rules — while accusing Anthropic, which has been among the most vocal advocates for slowing frontier development, of misconduct. "A clear distinction must be drawn between genuine security threats and mere technological competition," the post argued.
That demand places the two sides in a familiar bind. Washington has spent two years resisting binding AI rules domestically, while Beijing points to its own comprehensive AI governance regime as evidence it regulates more seriously. Any joint statement emerging from this weekend's talks would have to bridge that gap without either government appearing to concede ground at home.
Stakes beyond the talking points
The timing is not incidental. Both countries' AI sectors are racing toward landmark public listings, frontier labs on both sides of the Pacific are shipping systems with increasingly autonomous capabilities, and researchers in both countries have warned publicly about misalignment and misuse risks. A recent New York Times op-ed by a former US AI diplomacy official argued that upcoming safety talks are unlikely to produce meaningful agreement — but the fact that they are happening at all marks a shift from a year ago, when the primary US-China AI conversation was about which chips each side could buy.
For now, the deliverables are modest: an acknowledgment that risks are shared, a channel that stays open, and perhaps agreement to keep talking. Whether that channel survives contact with the export-control fights likely to dominate the wider economic relationship is the real question. In AI diplomacy, that passes for progress.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
