Anthropic accuses Chinese AI labs of mining Claude as US debates AI chip exports

Anthropic has accused Chinese AI companies DeepSeek, Moonshot AI, and MiniMax of using a technique called "distillation" to illicitly train their own models by generating over 16 million exchanges with Claude via more than 24,000 fake accounts. The targeted capabilities included agentic reasoning, tool use, and coding. This accusation surfaces while the US debates the strictness of AI chip export controls to China. Anthropic argues that the scale of these distillation attacks necessitates advanced chips, reinforcing the rationale for export controls to limit illicit extraction. Furthermore, the company warns that models built through this method are unlikely to retain crucial safety safeguards, posing national security risks if deployed by authoritarian governments for malicious activities.

(Source：TechCrunch)