Anthropic identifies industrial-scale distillation attacks by DeepSeek, Moonshot AI and MiniMax on Claude

Anthropic reported that it identified industrial-scale distillation campaigns by the AI laboratories DeepSeek, Moonshot and MiniMax to extract capabilities from its Claude model. The three companies generated more than 16 million exchanges with Claude through approximately 24,000 fraudulent accounts in violation of the company's terms of service and regional access restrictions. The campaigns targeted agentic reasoning, tool use and coding capabilities. DeepSeek conducted over 150,000 exchanges focused on reasoning tasks and rubric-based grading. Moonshot AI performed over 3.4 million exchanges aimed at agentic reasoning, tool use, coding and computer vision. MiniMax carried out over 13 million exchanges directed at agentic coding and tool orchestration. The laboratories used proxy services and networks of accounts to evade detection. Anthropic attributed each campaign to the specific laboratory through IP address correlation, request metadata and infrastructure indicators. The company said the attacks allow foreign labs to acquire capabilities extracted from American models and that they undermine export controls by enabling circumvention through other means.