US Treasury threatens sanctions on Chinese AI models over IP theft claims

Treasury Secretary Scott Bessent drew a new line in the sand Tuesday, telling Fox Business the U.S. will scrutinize Chinese open source AI models for stolen intellectual property — and won’t hesitate to slap sanctions on companies found guilty. “This administration supports open source models, but what we do not support is IP theft,” Bessent said, specifically calling out overseas models that might be pilfering from American firms.

The warning lands just as Chinese models like Moonshot AI’s Kimi K3 gain real traction, threatening the fundraising momentum and business models of U.S. leaders like OpenAI and Anthropic. It also follows an Axios report that the Trump administration is mulling an outright ban on Chinese open source models — a claim others have pushed back on. The White House said in April it would partner with AI companies to fight technology theft.

At the heart of the dispute is model distillation, a technique that compresses a larger model’s capabilities into a smaller system. Not everyone buys the argument that this constitutes theft. Microsoft CEO Satya Nadella recently called out the industry’s double standard, noting it’s “ironic” that labs claiming fair use to train on public data then “impose restrictive terms on distillation.” Hugging Face CEO Clem Delangue went further, telling TechCrunch’s Equity podcast that distillation is “a very small factor” in building good models and that everyone does it — including American companies.

If Bessent’s threat materializes, it would mark a significant escalation beyond chip export controls, targeting the models themselves. But the legal ground is shaky. Anthropic just got court approval to start cutting checks in a $1.5 billion copyright settlement after a judge ruled it illegally downloaded millions of books for training. The IP debate cuts both ways, and China’s research teams, as Delangue points out, are simply taking “a much more open and collaborative approach.”