Anthropic Says Chinese AI Labs Used Fraudulent Accounts To ‘Replicate’ Claude

Anthropic alleges that three Chinese AI labs—DeepSeek, Moonshot AI and MiniMax—used 24,000 fraudulent accounts and millions of queries to “replicate” Claude’s capabilities through large‑scale distillation. The company says the activity bypassed regional restrictions, violated its terms of service and targeted high‑value reasoning and coding outputs. Metadata and proxy‑network analysis were used to attribute the operation. The disclosure lands amid rising U.S.–China AI tensions and debates over export controls, model security and competitive boundaries. Critics argue the industry still lacks clear norms around data use and capability extraction, leaving questions about how frontier models should be protected.