US plans to sanction Chinese AI models if distillation evidence appears: Bessent

U.S. Treasury Secretary Scott Bessent said the Trump Administration is entitled to sanction open-weight models from China if evidence of distillation from American models is verified.

"If we see, especially that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft," Bessent said during an interview on Fox Business on Tuesday morning.

The China-based startup Moonshot AI, which is backed by Alibaba (BABA), released Kimi K3 last week, which it said is the world's largest open-weight AI model. And on Monday, Alibaba released a preview its flagship Qwen 3.8 Max.

Last month, Anthropic issued a letter to White House officials and U.S. Senators claiming that Alibaba was using a technique called distillation to improve the performance of its models. It is a training method where less capable models are improved by utilizing outputs from stronger models. Anthropic said it had evidence showing that operators linked to Alibaba's Qwen AI lab conducted 28.8M exchanges with Claude through about 25,000 fraudulent accounts.   

"We are finding watermarks of our U.S. large language models on many of the Chinese models, and that's unacceptable," Bessent said. "We're going to be looking at that in the coming days or weeks." 

U.S. and China officials are reportedly planning to hold meetings on AI in September. The AI talks, a significant outcome of ‌the U.S. President Donald Trump and Chinese President Xi Jinping's meeting in May, will likely take place before Xi's planned Sept. 24 visit to the U.S.

Executives from OpenAI and Anthropic have warned for years  about the potential of China's open-weight models impacting U.S. AI labs. The Chinese government subsidizes computing components and electricity consumption for AI labs, which allows these companies to release them for free to the public.