Washington accuses six Chinese companies of industrial-scale theft of AI capabilities.

The U.S. government has officially accused six Chinese companies of conducting, at least since late 2024, a systematic campaign to extract the capabilities of leading U.S. artificial intelligence models on an industrial scale. The joint advisory was issued by the NSA, CISA, and FBI.

The companies involved are DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI. According to the accusations, these companies allegedly sent millions of requests to variants of American models like Claude, GPT, Gemini, and Grok, obtaining billions of tokens to use in knowledge distillation processes. This activity is said to have been a central component of China's development strategy.

The distillation technique, normally legitimate, allows for training a smaller model to replicate the responses of a larger one. However, the U.S. accusation focuses on the industrial scale, organization, and methods used to acquire proprietary functionalities without fully bearing the enormous development costs.

The global competition in artificial intelligence has become one of the main areas of strategic confrontation between the United States and China. While American companies like OpenAI (creator of GPT and ChatGPT), Anthropic (Claude), and Google (Gemini) have driven innovation with billion-dollar investments, China has sought to close the technological gap through various means.

The technique of "knowledge distillation" is an AI training method where a smaller, more efficient model learns to mimic the behavior of a larger, more powerful model by accessing its responses. Normally, this is an accepted practice in research, but it becomes controversial when conducted on an industrial scale to replicate proprietary capabilities without authorization.

Today's accusations represent a significant escalation: it is no longer just isolated incidents of industrial espionage, but an organized operation that, according to Washington, would involve Chinese companies with the likely awareness of the Beijing government. This case emerges at a time when AI regulation is at the center of political debate both in Europe and the United States, and could have consequences for trade and technological relations between the two superpowers.