US Intelligence Agencies Say Six Chinese AI Firms Are Systematically Copying American Models

Keypoints:

• The NSA, CISA, and FBI issued a joint advisory on September 8 naming six Chinese AI companies
• DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI were accused of large-scale distillation since late 2024
• The agencies say the companies extracted billions of tokens across millions of requests from US frontier models
• DeepSeek's reported $5.6 million training cost was called misleading for excluding distillation costs
• China's government pushed back, calling the claims unfounded

Three major US security agencies issued a rare joint advisory on September 8, accusing six Chinese AI companies of running industrial-scale campaigns to copy capabilities from leading American AI models. The NSA, CISA, and FBI named DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI, and said the effort likely had the knowledge of the Chinese government.

The technique itself, called knowledge distillation, is standard and legal in AI development: training a smaller model using outputs from a larger, more capable one. The complaint centers on scale and method. 

The advisory alleges DeepSeek drew on multiple versions of Claude, GPT, and Gemini to train its R1 and V3 models, and said the company's widely cited $5.6 million training cost is misleading because it leaves out the cost of data obtained through distillation. Moonshot AI was accused of extracting data from Claude to train its Kimi K3 model.

The agencies described tactics mapped to a formal cybersecurity framework, plus methods outside it entirely, including a gray market of API proxies called "transfer stations" that let companies mask their location and dodge safeguards built into commercial AI terms of service. Those terms typically forbid distillation outright, since it lets a competitor benefit from a model's training without the enormous cost of building one from scratch.

China's Ministry of Commerce rejected the claims. The advisory lands weeks before President Trump is scheduled to host Chinese President Xi Jinping in Washington, with AI expected as a topic of discussion and a separate US-China AI safety dialogue planned for mid-September.