The Chinese AI sanctions threat took concrete shape on Tuesday when Treasury Secretary Scott Bessent said Washington would examine Chinese open-source models for signs of intellectual property theft, warning that sanctions could follow if violations are confirmed.
Bessent made the remarks on Fox Business’s Mornings with Maria, hosted by Maria Bartiromo, according to Benzinga. His comments were first reported by Bloomberg.
‘This administration supports open source models, but what we do not support is IP theft,’ Bessent said. ‘If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft.’
What Bessent Actually Alleged
The Treasury Secretary was not speaking in generalities. He said the US had detected ‘watermarks’ from American large language models inside several Chinese AI systems, calling the activity unacceptable, Vision Times reported.
He went further: Chinese models had sent ‘hundreds of millions of calls to US models in an attempt to reconstruct the model’s code and processes,’ and he pledged an investigation would conclude ‘within days or weeks,’ according to The Herald Business.
Bessent also floated a disclosure question: should US companies using Chinese AI models be required to tell their customers? He framed it bluntly, saying consumers ‘should not be using counterfeit products.’
On the same day, US Trade Representative Jamieson Greer told CNBC that Washington would ‘look very carefully at how China is proliferating AI development’ and work to ensure American companies compete on a level playing field.
The Chinese AI Sanctions Threat and the Distillation Debate
The backdrop is a genuine competitive anxiety in Washington. Chinese models, including Moonshot AI’s Kimi K3 most recently, are gaining capability and pulling users away from American firms like OpenAI and Anthropic, which rely on continued fundraising to finance frontier research.
After restricting China’s access to advanced chips and tightening export controls, the administration is now signalling it may target AI models directly. On Monday, Axios reported that the Trump administration is considering a wholesale ban on Chinese open-source models, though others have disputed that reporting.
Here is where the legal and technical ground gets genuinely complicated. The mechanism Washington appears most concerned about is model distillation: a technique that translates some capabilities of a larger model into a smaller, more portable system. AI labs have argued for months that foreign actors are using distillation to replicate their technology. But not everyone agrees that distilling another company’s model constitutes theft.
Microsoft CEO Satya Nadella put it directly: ‘While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation.’
Hugging Face CEO Clem Delangue was blunter still on a recent episode of TechCrunch’s Equity podcast. ‘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. ‘If it were easy just to do distillation to get good at building AI models, there would be many other countries, including in the U.S., with much better open source AI. The reality is they have really, really good research teams in China… taking a much more open and collaborative approach to AI than in the U.S.’
Anthropic’s Copyright Settlement Adds Context
The IP-theft framing lands awkwardly given what happened in a San Francisco federal court last week. On 20 July 2026, US District Judge William Alsup signed off on Anthropic’s $1.5 billion copyright settlement after ruling the company had illegally downloaded and stored copyrighted books to train its AI, according to Reuters.
The settlement covers approximately 465,000 books, with authors and publishers receiving around $3,000 per work covered by the agreement, AP News reported. Attorneys were awarded more than $101 million of the $187.5 million they had requested in fees.
Judge Alsup dismissed objections to the settlement’s size, saying complaints were ‘not grounded in a realistic assessment of the overall risks and rewards of a trial.’ Anthropic can now begin cutting cheques to authors.
The irony is hard to miss. Washington is threatening to sanction Chinese firms for copying US AI models at precisely the moment a leading US AI firm is settling claims that it copied human creators to build those models in the first place. The legal ground that ‘IP theft is IP theft’ turns out to be considerably murkier in practice.
Bessent’s promised investigation, due ‘within days or weeks,’ will be the first real test of whether this is a sanctions policy or an opening bid in the next round of trade talks.
