U.S. to propose emergency AI notification system with China
Add Axios as your preferred source to
see more of our stories on Google.
Treasury Secretary Scott Bessent told "The Axios Show" he plans to propose a notification process between the U.S. and China if something goes wrong with AI, and said he thinks Beijing would agree.
Why it matters: Bessent's idea reflects what he calls China's realization that its own open-source AI poses a bigger threat than Beijing had realized.
Driving the news: Bessent and Chinese Vice Premier He Lifeng discussed AI safety ahead of President Xi Jinping's state visit last month, including a proposal for an AI incident communication channel.
- Bessent told Axios the talks covered risks like uncontrolled AI agents, and non-state actors using the technology for cyber or biological threats.
What they're saying: "I think that they didn't realize how powerful their open source models are," Bessent said, referring to Chinese AI models that others can use and adapt.
- Bessent pointed to an incident involving Kimi, a Chinese AI model, where the model sent sensitive Chinese data to Anthropic, as one development that has "made the Chinese realize that this is a very powerful technology."
- "A lot of these Chinese models are very powerful — not as powerful as the U.S. models — but you get something that's 80%, 90% as powerful, without guardrails that can be imported anywhere in the world," Bessent said.
The intrigue: China is already worried about a different kind of AI threat — that the technology could weaken the Communist Party's tight control over information.
- China's spy chief warned last month AI could be used to fuel information and "cognitive" warfare against the government, threatening political and ideological stability.
Zoom out: The U.S. is also grappling with how to balance rapid AI development with growing concerns about the technology's risks.
- "We want safe acceleration," Bessent told Axios.
- He added that the government's model reviews are currently voluntary, though the administration reserves the right to intervene if a lab forges ahead despite safety concerns.
- "It is the people at the labs who have to accept responsibility, and I agree with that," Bessent said. "I think the labs have come around to that way of thinking, too."
Threat level: Asked about AI executives warning that they could lose control of increasingly powerful models, Bessent said bluntly: "Well, then they should slow down."
What to watch: Bessent said the U.S. needs to focus more on protecting critical systems from AI threats, rather than concentrating solely on the race to build more powerful models.
- "We are doing this hard scramble to the frontier for the models," Bessent said. "We really need to start worrying more about resilience and defense, too."
