Anthropic Accuses Chinese Labs of Illicitly Training AI Models

Anthropic accused DeepSeek, Moonshot, and MiniMax of misusing Claude to illicitly train their own AI models through over 16 million exchanges.These Chinese companies allegedly used 24,000 fraudulent accounts to conduct distillation attacks that bypass safeguards and undermine export controls.Anthropic is now upgrading its security systems and calling for a coordinated industry response to mitigate these national security risks.