Anthropic Alleges Chinese Firms Engaged in AI Data Theft

Anthropic has accused three Chinese AI companies—DeepSeek, Moonshot AI, and MiniMax—of engaging in industrial-scale intellectual property theft by illicitly extracting capabilities from its Claude chatbot. The accused companies allegedly utilized a technique known as "distillation," which allows them to enhance the performance of their less sophisticated models based on outputs from Claude. This leads to a competitive advantage by circumventing export controls intended to safeguard US technological dominance.

The implications of these actions raise significant national security concerns. Models derived through this unauthorized method could lack essential safety measures that prevent misuse, such as developing bioweapons or carrying out cyberattacks. With increasing sophistication in these campaigns, Anthropic calls for a coordinated response from both industry stakeholders and government entities to effectively tackle the challenges posed by such intellectual property infractions.