U.S. Treasury Secretary Scott Bessent said on Tuesday that the administration will review Chinese open-source artificial intelligence models for indications of intellectual property theft and could impose sanctions on Chinese AI firms if theft is confirmed. Speaking on Fox Business, Bessent stressed that while the administration supports open-source models, it does not support IP theft.
Bloomberg was the first to report Bessent’s comments. The remarks come as an increasing number of Chinese models — most recently Moonshot AI’s Kimi K3 — have improved in capability and gained popularity, posing competitive pressure on U.S. firms such as OpenAI and Anthropic and potentially affecting their ability to raise capital for frontier models.
Political context and possible actions
Axios reported on Monday that figures in the Trump administration were considering a wholesale ban on Chinese open-source models, though that claim has been disputed by others. In April, the White House said it would work closely with AI companies to combat theft of technology.
The United States has already restricted China’s access to advanced chips and tightened export controls; targeting models themselves with sanctions would represent another escalation in the effort to preserve U.S. leadership in AI, shifting focus from hardware and components to the software models.
Model distillation and legal debate
Model distillation is a technique that transfers some capabilities of a larger model to a smaller, easier-to-run system. There is disagreement within the industry over whether distillation amounts to theft of another company’s model.
Microsoft CEO Satya Nadella earlier this month criticized major labs for assuming that distillation is improper, arguing that fair use rights to train on public data are important for innovation while noting the irony of then imposing restrictive terms on distillation.
At the same time, training practices remain a source of legal risk: Anthropic this week received approval to begin issuing payments to authors under its $1.5 billion settlement after a judge found the company had illegally downloaded and stored millions of copyrighted books to train its AI.
Why Chinese models are catching up: distillation or research capacity?
Industry figures disagree on whether distillation is the primary reason Chinese models are rapidly improving. Clem Delangue, CEO of Hugging Face, told TechCrunch’s Equity podcast that distillation is a minor factor in building strong models and that China’s progress also reflects very capable research teams and a more open, collaborative approach to AI.
The U.S. threat of sanctions and ongoing investigations will shape policy toward the AI race. If Washington moves to sanction Chinese models, it would mark a significant escalation with implications for regulation, trade, and technological competition.
