US threatens sanctions on Chinese AI models over IP theft
Washington is threatening to sanction Chinese open-source AI developers for intellectual property theft, a move that risks fragmenting the global market and creating compliance headaches for European firms.
The United States is preparing to sanction Chinese artificial intelligence companies if it determines their open-source models were built using stolen intellectual property. Treasury Secretary Scott Bessent announced the crackdown on Tuesday. “If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft,” Bessent said.
For European developers and investors, the threat marks a potential escalation in the fracturing of global technology markets. European startups increasingly rely on open-source models to build applications without the massive capital required to train systems from scratch. Sanctions on Chinese models would force these companies to navigate complex compliance rules or abandon capable, cost-effective tools.
Washington's intervention comes as Chinese models like Moonshot AI’s Kimi K3 rapidly close the capability gap with leading American systems. The rise of these alternatives threatens the business models of US firms like OpenAI and Anthropic, making it harder for them to raise capital. Targeting the models themselves represents a significant step beyond existing US restrictions on China's access to advanced microchips.
At the centre of the dispute is "distillation," a technique where a smaller model is trained to replicate the capabilities of a larger one. US AI labs claim foreign actors are using this method to copy their technology and release it freely. However, the legal status of distillation remains highly contested. “I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation,” said Microsoft CEO Satya Nadella.
Critics argue Washington is ignoring the real drivers of Chinese AI progress. “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.,” said Clem Delangue, CEO of Hugging Face. “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.”
The legal ambiguity surrounding AI training extends beyond this geopolitical dispute. Anthropic this week received approval to begin paying authors as part of a $1.5 billion settlement after a judge found it illegally downloaded copyrighted books for training. For European regulators enforcing the AI Act, the US push to penalise specific training methods will complicate efforts to establish a clear, unified rulebook for the industry.
