Anthropic Accuses Chinese AI Labs of Illicitly Distilling Claude Capabilities
Anthropic publicly accused three Chinese AI laboratories of using fraudulent accounts and proxy services to illicitly distill Claude's capabilities at industrial scale, violating terms of service and creating what it characterized as national security risks if the extracted capabilities were deployed without safeguards.
What happened
Anthropic accused three Chinese AI laboratories—DeepSeek, Moonshot, and MiniMax—of running industrial-scale distillation campaigns to extract Claude's capabilities for their own models, using 24,000 fraudulent accounts and proxy services to make over 16 million exchanges with Claude while evading detection.
DeepSeek accounted for 150,000 exchanges, Moonshot for 3.4 million, and MiniMax for 13 million. Anthropic stated the volume, structure, and focus of prompts were distinct from normal usage patterns and violated terms of service and regional access restrictions.
Reported impact
Anthropic warned that illicitly distilled models lack necessary safeguards and create national security risks, potentially enabling foreign governments to use these unprotected capabilities in military, intelligence, surveillance systems, offensive cyber operations, disinformation campaigns, and mass surveillance.
Index analysis
Distillation is a common training method in the AI industry, but most leading proprietary model providers explicitly ban its use to extract competitors' capabilities. Anthropic stated that using distillation as a shortcut to take capabilities from competitors steals the company's intellectual property, computing power, and effort.
Governance controls
Claude is not available in China, and Anthropic had regional access restrictions in place. The company paired its allegations with repeated calls for stronger export controls on AI capabilities.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.