U.S. Accuses Chinese AI Firms of Systematic Model Theft

The U.S. government has accused Chinese AI companies, including DeepSeek and Moonshot AI, of systematically distilling U.S. AI models since 2024. A joint advisory from the NSA, CISA, and FBI highlights that these companies have extracted billions of tokens from advanced models like Anthropic's Claude and OpenAI's ChatGPT. The advisory details the sophisticated methods used by these firms to bypass restrictions and extract proprietary functionalities from U.S. models. Techniques include spreading requests across multiple accounts and leveraging proxies to obfuscate user data. The advisory emphasizes that this activity is a core part of China's AI development strategy. The U.S. agencies are calling for a coordinated response across the AI ecosystem to address these industrial-scale distillation efforts.

Key Points: • U.S. agencies accuse Chinese firms of systematic AI model theft. • DeepSeek and Moonshot AI are among the companies named in the advisory. • Chinese firms use advanced techniques to bypass U.S. AI model protections.