China Investigates DeepSeek and Moonshot Over Data Routed to Anthropic's Claude

China's internet regulator has launched an investigation into DeepSeek and Moonshot AI after Anthropic published a detailed threat intelligence report alleging both companies secretly forwarded millions of user requests, including sensitive police and military data, to Claude for model training. The Cyberspace Administration of China has visited both firms' offices to question executives, according to reporting by The Information. The probe marks a significant escalation in scrutiny over a practice known as illicit distillation—training one's own AI models on another company's outputs through fraudulent means.