NSA, FBI, and CISA name six Chinese AI firms running industrial-scale extraction against US models
A joint advisory says DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI have systematically queried Claude, GPT, Gemini, and Grok at massive scale since late 2024 to extract proprietary capabilities via knowledge distillation.
The National Security Agency, the Cybersecurity and Infrastructure Security Agency, and the FBI jointly published advisory AA26-251A on September 8, 2026. It names six China-based AI companies (DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI) as running systematic knowledge-distillation campaigns against US frontier models since at least late 2024. The technique queries target models such as Claude, GPT, Gemini, and Grok at massive scale to generate synthetic training data that extracts the target's proprietary capabilities. The advisory lists obfuscation tactics including spread accounts, proxies, and third-party aggregators.
The language
The wording is unusually direct for a joint government cybersecurity notice. It describes "aggressive, malicious, and targeted distillation activities at an industrial scale" that form the core of the named companies' AI development strategy. The US government is formally attributing a systematic extraction campaign to six named commercial entities.
What this changes for vendor evaluation
If you benchmark against or route enterprise traffic through the named firms' models, this is now a government-attributed security concern on the record. It doesn't change what those models can do. It does belong in your vendor risk assessments and export-control-adjacent compliance reviews from here on.