The United States may impose sanctions on China if an ongoing investigation determines that there has been “intellectual property theft” to U.S. artificial intelligence companies (IA) by Chinese rivals, threatened Treasury Secretary Scott Bessent on Tuesday.
© Michael Nagle/Bloomberg via Getty Images
22/07/2026
by
Lusa
“ Sand we see, mainly, that foreign models are stealing our big companies, we have the ability to sanction them for this,” Bessent said in an interview with Fox Business today.
The U.S. Treasury Secretary stated that Washington will not tolerate the theft of intellectual property and that evidence was found of the presence of U.S. A. model capabilities in Chinese systems, something that will be analyzed “in the coming days or weeks.”.
Bessent referred to the so-called “distillation” technique, by which a smaller AI model learns to reproduce the behavior of a more advanced model using its responses.
The US companies, such as Anthropic, claim that some Chinese companies have used this technique to replicate the capabilities of their unauthorised models, including DeepSeek, Moonshot AI and MiniMax.
Bessent's statements arise at a time of increasing competition between the two largest economies in the world for leadership in AI, after Chinese companies have recently presented models that rival those developed by Anthropic or OpenAI in terms of performance and cost-benefit ratio.
Among them is the Moonshot AI startup, which last Friday presented the Kimi K3, a model that, according to the company, surpasses those of OpenAI and Anthropic in some performance tests used by the industry.
Its launch caused falls on Wall Street, particularly among semiconductor companies, among fears that Chinese models might push the margins of the sector.
Moonshot AI has announced that Kimi K3 reaches 2.8 trillion parameters – as it indicates the complexity of an AI system during its training – a higher number than other Chinese open source models such as DeepSeek V4 Pro (1.6 trillion) or Zhipu AI GLM-5 series (744 billion).
The Artificial Analysis consultant classified K3 at a level comparable to Claude Opus 4.8 and GPT-5.5, considered the second best models of Anthropic and OpenAI, respectively, although behind Claude Fable 5 and GPT-5.6 Sun in terms of global performance.
Although US companies continue to lead the development of more advanced models, Chinese companies such as DeepSeek, Alibaba, Tencent and Moonshot AI have accelerated competition with systems – some of them open source – that stand out for their cost-benefit ratio and allow companies and researchers to adapt them to their needs.