The substance of the accusations
US law enforcement and intelligence agencies accused six Chinese AI companies of aggressive, targeted distillation on an industrial scale. Reuters names DeepSeek, Moonshot AI and Alibaba among the accused but does not provide the full list in the publicly accessible portion of its report. [1 · Reuters]
Distillation trains a smaller model on the responses of a larger system, reducing costs and reproducing some of its capabilities more quickly. The authorities said Chinese developers accessed versions of models from Anthropic, OpenAI, Google and SpaceX, probably with the knowledge of the Chinese government. [1 · Reuters]
What remains unconfirmed
Reuters does not report the publication of technical evidence, the volume of responses obtained or the legal characterization of each incident. The Chinese embassy in Washington did not respond to the publication’s inquiry, and the article does not include the named companies’ positions. [1 · Reuters]
US officials further argue that these methods reduce Chinese companies’ costs and could enhance China’s military and cyber capabilities. This is an assessment by US authorities, rather than an independently confirmed conclusion. [1 · Reuters]
Sources
- Reuters — US accusations against Chinese AI companies — September 8, 2026
Expert commentary
The statement moves the distillation dispute from API terms of use into the realm of national security. For US laboratories, this could mean tighter access restrictions, checks on the origin of requests and increased monitoring of large-scale collection of responses. [1 · Reuters]
Until the authorities disclose evidence and the legal basis for their claims, neither the scale of copying nor the Chinese state’s involvement can be considered established. Developments to watch include official investigation materials, company responses and possible sanctions or export restrictions. [1 · Reuters]