AI company Anthropic has reportedly accused three Chinese AI developers of illegally extracting capabilities from its Claude chatbot to improve their competing models. According to reports from Swedish tech publications, Anthropic alleges that Deepseek, Moonshot, and Minimax used approximately 24,000 fraudulent accounts to conduct over 16 million exchanges with Claude, violating both user agreements and regional restrictions.
The companies are said to have employed a technique called distillation, which involves training less competent models on a better model. This method reportedly allows for faster and cheaper training of new AI models. Anthropic claims that Minimax had over 13 million exchanges with Claude, Moonshot 3.4 million, and Deepseek 150,000.
This is not the first time Chinese AI developers have faced such accusations. In early 2025, OpenAI reportedly accused Deepseek of using illegal distillation. Anthropic is calling for stricter laws and a coordinated response from AI companies and cloud services to address what it describes as increasingly intensive and sophisticated campaigns to free-ride on others' work.