The NSA, CISA and FBI issued a joint cybersecurity advisory on September 8 naming six Chinese AI companies — DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI — accusing them of running industrial-scale campaigns to extract capabilities from leading US AI models since late 2024, likely with the knowledge of the Chinese government.
The agencies say the firms pulled billions of tokens across millions of queries from Anthropic's Claude, OpenAI's GPT, Google's Gemini and xAI's Grok. DeepSeek allegedly targeted reasoning and agentic functions to train its R1 and V3 models, which the advisory says undercuts its widely cited $5.6 million training-cost claim. Moonshot AI reportedly distilled Claude Fable 5 to build Kimi K3, while Z.AI is said to have pulled billions of tokens from GPT-5.5 and Claude Opus 4.8 by mid-2026.
CISA Acting Director Nick Andersen urged AI companies to "take immediate steps to safeguard their platforms against knowledge distillation campaigns." The advisory recommends behavioral monitoring and quietly degrading output for suspected accounts — a sign the US-China AI rivalry is now being treated as an active security threat, not just a competitive one.