Anthropic Accuses Chinese AI Labs of Mining Claude Capabilities

Anthropic has leveled serious allegations against several prominent Chinese AI companies. The AI safety firm accused DeepSeek, Moonshot AI, and MiniMax of conducting a coordinated campaign to extract the capabilities of its Claude AI models. According to Anthropic, the companies used approximately 24,000 fake accounts to systematically query the Claude API, likely in an effort to distill its knowledge and reasoning patterns for training their own competing models. This practice, known as model extraction or distillation, involves using a model's outputs as training data to create a cheaper or more accessible copy. The allegation emerges amid ongoing debates in Washington over tightening export controls on advanced AI chips to China, highlighting the intensifying technological and geopolitical tensions in the AI sector. If proven, such activity would represent a significant breach of standard terms of service and intellectual property norms. It underscores the high-stakes race for AI