Three US government bodies – the NSA, FBI and cybersecurity agency CISA – have jointly accused Chinese AI firms of running "industrial-scale" distillation campaigns designed to extract proprietary capabilities from leading American AI models. The advisory names DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI, and claims Beijing is "likely" aware of the activity, framing distillation as the central strategy of Chinese AI development rather than a peripheral technique. The accusation matters because it feeds into broader US concerns about intellectual property theft and technological competition with China in the AI race.

Distillation involves a smaller model learning by querying a larger one, and while it has legitimate uses, most commercial providers ban it in their terms of service to protect their investment. The agencies allege Chinese firms circumvent these restrictions via APIs, cloud providers and "transfer stations" – grey-market proxies that resell access to frontier models cheaply while masking user identity. Specific claims include that DeepSeek used distillation to generate synthetic training data, undermining its assertion of building models cheaply, and that Alibaba used the technique to enhance its free Qwen models. The agencies have urged AI companies to watch for warning signs, such as sudden heavy usage from new accounts, and to subtly alter outputs to suspected distillers, amid an ongoing war of words in which China has accused US firms of doing the same.