Treasury Secretary Scott Bessent said the US would examine open-source models from China for signs of intellectual property theft, threatening sanctions against Chinese AI companies if IP theft is established.

"We've seen a lot of talk about open source models coming and threatening the large language models in the U.S.," Bessent said on Fox Business Tuesday. "This administration supports open source models, but what we do not support is IP theft. If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft."

The statement comes as Chinese models are gaining capabilities and popularity, threatening American AI firms like OpenAI and Anthropic. Moonshot AI's Kimi K3 recently emerged as a potential competitor to frontier models from US labs.

On Monday, Axios reported the Trump administration is considering a wholesale ban on Chinese open-source models, though others have disputed that claim.

Growing tensions over AI competition

AI companies have warned for months against campaigns by foreign actors to copy their technology and redeploy it as open-source. In April, the White House said it would work closely with AI firms to combat the theft.

Sanctions would add to Washington's strategies to maintain its AI lead. After restricting China's access to advanced chips and tightening export controls, the government now signals it may target AI models themselves.

This marks a significant escalation in technological competition between frontier labs and Chinese open-source alternatives.

Model distillation allows larger model capabilities to be translated into smaller systems that are easier to run. However, not everyone agrees that distilling another company's model constitutes theft.

Earlier this month, Microsoft CEO Satya Nadella criticised large labs for making this assumption, defending fair use rights to train on publicly available data.

The Treasury's threat represents the most direct government intervention yet in the AI model space, moving beyond hardware restrictions to target the software itself.