In a bombshell revelation, the leading U.S. artificial intelligence company Anthropic has accused several prominent Chinese AI firms of illegally siphoning massive amounts of data from its vaunted chatbot, Claude. What this really means is that the long-simmering AI rivalry between the U.S. and China has erupted into an all-out war over data and intellectual property.
China Accused of "Industrialized" Data Theft
In a scathing blog post, Anthropic claimed that three Chinese AI companies - DeepSeek, Moonshot, and MiniMax - have created thousands of fraudulent accounts to conduct millions of interactions with Claude, all in an effort to harvest the data and use it to train their own chatbots. The bigger picture here is that this appears to be part of a broader, "industrialized" campaign by Chinese firms to siphon off valuable training data from leading U.S. AI models.
According to Anthropic, DeepSeek alone engaged in 150,000 interactions with Claude, while Moonshot had over 3.4 million and MiniMax a staggering 13 million. This practice, known as "distillation," is a common technique in the AI field, but Anthropic's terms of service explicitly prohibit using its technology in this way.
Potential National Security Risks
Anthropic warned that this data theft could pose significant national security risks, as the extracted knowledge could be used by China to develop more advanced military AI or surveillance tools. As Reuters reports, a senior Trump administration official claimed that DeepSeek's latest AI model "likely relied heavily on distilling models from major U.S. AI companies."
The implications are far-reaching. As Anthropic digs in against the Pentagon's AI demands, this data theft saga underscores the high stakes in the global AI race and the growing tensions between the U.S. and China over technological supremacy.