U.S. Threatens Sanctions Over Chinese AI Model IP Theft

The Trump administration is weighing sanctions and Entity List designations against Chinese artificial intelligence firms following the July 16 release of Moonshot AI's Kimi K3. U.S. Treasury Secretary Scott Bessent and White House OSTP Director Michael Kratsios allege that Moonshot and other labs, including DeepSeek and MiniMax, used "distillation" to illicitly extract proprietary technology from U.S. models like Anthropic's Fable. Kratsios specifically accused Moonshot of using a sophisticated internal platform to bypass detection and utilizing restricted Nvidia GB300 chips via servers in Thailand.

Chinese officials have rejected these claims as "AI hegemonism," with the Ministry of Commerce asserting that American firms also distill Chinese models. While President Donald Trump maintains the U.S. leads substantially in AI, the Kimi K3 model has seen a 387% surge in U.S. downloads, prompting companies like Coinbase and Mozilla to adopt Chinese systems for their cost-efficiency. In response to the tech rivalry, President Xi Jinping established the World Artificial Intelligence Cooperation Organisation in Shanghai to promote global open-source collaboration.

Silicon Valley is divided on the government's response. OpenAI and Anthropic have lobbied for tighter controls to prevent IP theft. Conversely, Nvidia CEO Jensen Huang and Meta CEO Mark Zuckerberg argue that blocking open-weight models would stifle innovation and unfairly benefit proprietary U.S. giants. A coalition of 179 startups has also urged the administration against a ban, claiming that access to open models is a competitive necessity.