China Threatens US Retaliation Over Sanctions Against AI Firms Accused of Model Theft

China's Ministry of Commerce has formally demanded that the United States stop threatening sanctions against Chinese artificial intelligence companies. In a statement issued on July 27, the ministry called the American actions baseless in fact and law, describing them as a textbook case of AI hegemonism.

A ministry spokesperson said the US actions have no justification and amount to an attempt to impose dominance in artificial intelligence. The statement came in response to reports that senior US officials had publicly called for investigations into Chinese AI companies over alleged distillation of advanced American models, and had signaled possible sanctions under the pretext of intellectual property theft.

The Kimi K3 Dispute Explained

The conflict was sparked by the launch of Kimi K3, a model from Chinese company Moonshot AI that quickly rose to the top of global coding benchmarks. Michael Kratsios, director of the White House Office of Science and Technology Policy, accused Moonshot of distilling Anthropic's Fable model, a technique in which a smaller model is trained on the outputs of a larger one. Kratsios also claimed Moonshot had gained access to Nvidia servers running GB300 chips in Thailand, raising questions about possible violations of US export rules.

Bessent Threatens Blacklisting

US Treasury Secretary Scott Bessent reiterated the sanctions threat, writing that open-source software does not mean open season on American intellectual property. He said that when Chinese firms run covert, industrial-scale distillation operations that cross the line into IP theft, sanctions and Entity List designations are on the table.

Open-source software doesn't mean open season on American IP. When Chinese firms run covert, industrial-scale distillation operations that cross the line into IP theft, sanctions and Entity List designations are on the table. - Scott Bessent, US Treasury Secretary

Moonshot AI has firmly denied the allegations. Huang Zhenxin, the company's head of business, told Chinese state media that Kimi K3's performance leap comes entirely from original architectural innovations, citing solutions such as Moon Clip, Kimi Delta Attention and Attention Residuals as the source of the model's progress.

Experts Question the Claims

Independent analysts have cast doubt on the credibility of the distillation accusations. Wei Sun, chief AI analyst at Counterpoint Research, told CNN that the timeline behind the claims is unconvincing, since Fable had been on the market too briefly for Moonshot to have run a full training process based on it. Distillation alone, he added, does not explain Kimi K3's edge in some benchmark tests.

China's Ministry of Commerce turned the accusation around, claiming that American AI companies have repeatedly used distillation of Chinese models in their own research and development and training processes. The ministry did not specify what retaliatory measures it is considering, saying only that it would take all necessary action if its interests were genuinely violated.

A Wider Standoff

The dispute is part of a months-long escalation in technology tensions between Washington and Beijing over artificial intelligence. Earlier, the Trump administration had considered banning Chinese AI models, a move that drew protest from roughly 200 American startups that rely on cheap, open Chinese models. The US and China had previously announced plans for AI-focused talks in September, but the current exchange of accusations shows tensions rising rather than easing.

For Polish companies and developers, the dispute has practical stakes, since models like Kimi K3 are increasingly chosen over pricier Anthropic or OpenAI offerings for their lower cost and open license. Any US sanctions or export restrictions could make these models harder to access globally, including in Europe, where companies already face limited access to the most powerful models from American providers.

For now, neither side has announced concrete enforcement steps. Bessent gave no timeline for a possible Entity List designation for Moonshot AI, and the Chinese ministry offered no details on its planned countermeasures. A resolution may not come until the US-China talks expected in September.