US agencies allege industrial-scale extraction of frontier AI model capabilities
A joint cybersecurity advisory from CISA, the National Security Agency and the FBI says China-based AI companies have conducted industrial-scale knowledge-distillation campaigns against US AI companies since at least late 2024. The advisory names DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI, and says the campaigns used billions of tokens across millions of exchanges or requests involving variants of Claude, GPT, Gemini and Grok.
The agencies describe access through native APIs, remote cloud providers, third-party aggregators and gray-market proxy services they call “transfer stations.” The advisory also cites shared premium subscriptions, chain-of-thought extraction, automated failover between access paths and quality-evaluation frameworks intended to detect countermeasures. It recommends that US AI companies monitor anomalous prompts, accounts, networks and usage patterns, alter responses to suspected extraction attempts, and share intelligence across providers, cloud platforms and aggregators.
These are the authoring agencies’ claims and attribution; the advisory does not independently establish each named company’s conduct in this report. The source presents knowledge distillation as a legitimate research technique while alleging that the described campaigns used it maliciously and at industrial scale.