Anthropic Accuses Three Chinese AI Companies of Data Harvesting Through Fraudulent Accounts
San Francisco-based AI company Anthropic has accused three Chinese AI firms—DeepSeek, Moonshot, and MiniMax—of using approximately 24,000 fraudulent accounts to harvest data from its Claude chatbot, according to reports from The New York Times and TechCrunch.
According to TechCrunch, Anthropic claims the Chinese companies engaged in “distilling” Claude’s AI capabilities, a process that involves extracting knowledge and capabilities from one AI model to train another. The alleged operation used fake accounts to interact with Claude and capture its responses.
The accusations come as U.S. officials debate implementing export controls designed to slow China’s AI development, TechCrunch reports. The timing highlights ongoing tensions between American and Chinese AI companies, particularly around intellectual property and competitive advantages in the rapidly evolving AI sector.
The New York Times characterizes the activity as using fraudulent accounts “to train their own chatbots,” suggesting the harvested data was intended to improve the Chinese companies’ competing AI products.
Neither the accused companies nor U.S. government officials have been quoted in the available reports responding to Anthropic’s claims.