US threatens sanctions on Chinese AI models over alleged intellectual property theft

Key points

• Treasury Secretary Scott Bessent announced on Tuesday, 20 May 2025, that the US will examine Chinese open-source AI models for signs of intellectual property theft.
• Bessent threatened sanctions against any Chinese AI company found to have stolen technology from American firms.
• Chinese AI lab Moonshot AI's model Kimi K3 is among the latest Chinese models closing the gap on leading US systems.
• Anthropic reached a $1.5 billion settlement this week after a judge ruled it had illegally downloaded millions of copyrighted books to train its own AI.
• Hugging Face CEO Clem Delangue argues distillation is a minor factor in China's AI progress, and that strong Chinese research teams deserve most of the credit.

The US government is turning up the heat on Chinese artificial intelligence. On Fox Business on Tuesday, Treasury Secretary Scott Bessent said Washington will examine open-source AI models, AI systems whose underlying code is shared publicly, coming out of China for signs of intellectual property theft.

"We've seen a lot of talk about open source models coming and threatening the large language models in the US," Bessent said. "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." Bloomberg reported his comments first.

The warning follows rapid progress by Chinese AI labs. Moonshot AI's Kimi K3, which we reported in detail on 18 July, drew attention for rivalling expensive American systems at a fraction of the cost. That kind of competition threatens US firms like OpenAI and Anthropic, not just on market share but on their ability to raise the capital needed to build next-generation AI.

One practice at the centre of the debate is model distillation, a technique where a smaller AI model is trained to mimic the behaviour of a larger, more powerful one. Some say Chinese developers have used it on American models without permission. Others say the practice is universal.

Hugging Face CEO Clem Delangue pushes back on the theft framing. "We know distillation to be a very small factor in the ability to create good models, and it's a practice that everyone is doing, including companies in the U.S.," he said on a recent episode of TechCrunch's Equity podcast. He credits China's research culture, not copying, for its AI gains: "They have really, really good research teams in China, taking a much more open and collaborative approach to AI than in the U.S."

Microsoft CEO Satya Nadella has noted the same tension. He said it was ironic that labs claim fair-use rights to train on public data, then impose restrictive terms on others who want to distil from their outputs.

Could US sanctions actually stop Chinese AI?

Probably not on their own. Sanctions can restrict American companies from doing business with named Chinese firms, but they can't stop Chinese researchers from publishing or sharing code globally. Washington has already limited China's access to advanced chips, the specialised processors that power AI training. Targeting models directly would be a new, harder-to-enforce step.

For ordinary users, the immediate effect is noise rather than action. Chinese AI apps remain freely available. But if sanctions do arrive, American businesses that have quietly built products on cheap Chinese models could find themselves in a legal grey area fast.

The bigger picture: Washington is now treating AI as a national security matter in a way that reaches beyond hardware into the software itself. Whether sanctions have any teeth here is the question worth watching.