U.S. Agencies Accuse Chinese AI Firms of Model Distillation

The National Security Agency, the Cybersecurity and Infrastructure Security Agency, and the Federal Bureau of Investigation issued a joint advisory accusing six China-based AI companies of conducting industrial-scale distillation campaigns. The agencies allege that firms including DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI used millions of fraudulent requests and bulk subscriptions to extract capabilities worth billions from U.S. models such as GPT, Claude, Gemini, and Grok.

U.S. officials claim this process allowed Chinese firms to close the capability gap and enabled DeepSeek to understate its actual training costs. The Government of China rejected these allegations as groundless, asserting that its progress is a result of domestic innovation and scientific self-reliance. Some analysts suggest China's competitiveness also stems from algorithmic efficiency and optimizations of the attention mechanism, necessitated by U.S. restrictions on high-end Nvidia chips.

Treasury Secretary Scott Bessent warned that the United States cannot afford to lose the AI race, emphasizing the urgent need for rapid data center and infrastructure expansion. Bessent argued that the U.S. cannot pause its efforts because Chinese competitors will not do so, noting that if China pulls ahead in AI, other strategic considerations become irrelevant.