U.S. Accuses Chinese Firms of AI Token Extraction Campaign

• •Six Chinese AI firms are accused of mass token extraction from U.S. models.
• •The operations are believed to be state-sponsored and systematic, not opportunistic.
• •Anthropic blocked similar extraction attempts from seven Chinese labs, including Alibaba and DeepSeek.

The U.S. government has accused six Chinese AI companies of conducting large-scale token extraction campaigns against American AI models since late 2024. The companies named are DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI. The NSA, CISA, and FBI issued a joint advisory detailing that these operations were systematic and likely state-sponsored, aimed at capturing advanced capabilities from models like Claude, GPT, Gemini, and Grok. Anthropic also reported blocking illicit extraction attempts from seven Chinese labs, including some of the same companies. The extraction methods involved sending millions of requests through APIs, utilizing a sophisticated infrastructure designed to evade detection. This activity represents a significant threat to U.S. AI development and raises concerns about intellectual property theft.