US Accuses Six Chinese AI Firms of Industrial-Scale Model Theft

On Tuesday, the NSA, CISA, and FBI charged six Chinese AI firms—DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI—with carrying out large-scale distillation attacks on US AI models since late 2024. These companies are accused of extracting proprietary features from models such as Claude, GPT, Gemini, and Grok, likely with the knowledge of the Chinese government. The assaults involved the use of fake accounts to exploit AI model inference APIs, executing coordinated queries, and employing prompt injection techniques to uncover concealed reasoning. Reportedly, these actions have expedited Chinese AI development and lowered training expenses. US agencies are encouraging American AI companies to work with the government to counter these threats and enhance the detection of advanced campaigns.