A joint notice published on September 8 names DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI, and says distillation is the core of their strategy rather than a supplement.
On September 8, 2026, the three US agencies published a joint opinion on six Chinese artificial intelligence companies: DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI. They accuse them of having extracted billions of tokens from variants of Claude, GPT, Gemini and Grok since the end of 2024 in order to recover their own capacities.
Distillation — leading to one model on the output of another — remains a perfectly accepted research technique. What the notice describes as abuse is scale and circumvention: requests would go through hijacked APIs, third-party cloud providers and aggregators that scramble metadata to escape detection, in violation of terms of use.
Two claims are new in relation to the suspicions already expressed in 2025: the agencies consider that the Chinese authorities are likely to be aware of these campaigns, and that this extraction is the basis for the development of these models, not a one-time complement. The notice recommends that U.S. suppliers identify and then discreetly degrade suspicious accounts rather than close them abruptly.
Read the full article on the original source
Go to the source →