U.S. cyber and law enforcement officials accused Chinese artificial intelligence companies on Tuesday of maliciously copying technology from sophisticated American-made AI models through a technique known as distillation.

The National Security Agency, Cybersecurity and Infrastructure Security Agency, and FBI issued a joint advisory stating that firms including DeepSeek, Moonshot AI, Alibaba, MiniMax, and StepFun engaged in aggressive and targeted distillation activities at an industrial scale. The agencies said the companies extracted proprietary functionalities and capabilities from U.S. frontier models, including variants of those developed by Anthropic, OpenAI, Alphabet's Google, and SpaceX.

Distillation involves training smaller AI models using output from larger, more expensive ones to reduce training costs. While the method has legitimate research applications, U.S. officials described the Chinese efforts as covert and systematic, often routed through proxies and transfer stations to evade terms of service and geographic restrictions. The agencies noted the activities occurred with likely Chinese government awareness and have taken place since at least late 2024.

The allegations come as the Trump administration prepares for Chinese President Xi Jinping's visit to the United States in late September. U.S. and Chinese officials are also scheduled to discuss AI safety risks in mid-September. Treasury Secretary Scott Bessent indicated that sanctions and Entity List designations remain options for companies involved in such activities.

Chinese officials and companies have previously rejected similar accusations as baseless. The new advisory adds to ongoing tensions over intellectual property in the AI sector, where American firms have long warned that unauthorized distillation allows foreign competitors to replicate capabilities at lower cost without matching full performance.

The CISA advisory detailed how the named Chinese entities conducted high-volume knowledge distillation campaigns, extracting billions of tokens across millions of exchanges. It emphasized that these efforts form a core part of the companies' development strategies rather than a supplement to original research.