Anthropic Accuses Three Chinese Startups of Harvesting Its A.I. Data

The San Francisco artificial intelligence start-up Anthropic says three Chinese companies — DeepSeek, Moonshot and MiniMax — used about 24,000 fraudulent accounts to generate more than 16 million conversations with its Claude chatbot, material the companies could use to train their own systems.

The practice, known as distillation, is common in A.I. work, but Anthropic’s terms bar surreptitious harvesting and prohibit use of its technologies in China. OpenAI has made similar accusations, and in a memo to the House Select Committee on China warned that some Chinese start-ups are using new, obfuscated distillation methods to free-ride on U.S.

technologies. Anthropic called the campaigns a national security risk, saying distillation can strip away guardrails and enable A.I. systems to be repurposed for bioweapons or mass surveillance, and urged coordinated action by industry and policymakers. DeepSeek, Moonshot and MiniMax did not immediately respond to requests for comment.

China, San Francisco

anthropic, claude, deepseek, moonshot, minimax, distillation, data harvesting, fraudulent accounts, openai, national security