Anthropic says it has identified large-scale coordinated campaigns by three Chinese AI laboratories — DeepSeek, Moonshot AI, and MiniMax — to extract Claude's capabilities through illicit model distillation, collectively generating over 16 million exchanges via approximately 24,000 fraudulent accounts. The campaigns targeted Claude's most distinctive capabilities — including agentic reasoning, tool use, and coding — using proxy services and hydra cluster account networks to evade access controls, with MiniMax alone accounting for over 13 million exchanges. Anthropic warns that distillation attacks undermine AI export controls and strip safety safeguards, and says the threat requires coordinated action across industry, cloud providers, and policymakers.

Detecting and preventing distillation attacks - Anthropic