China Reportedly Probes DeepSeek, Moonshot Over Claude Data Routing

China is investigating AI companies DeepSeek and Moonshot after Anthropic alleged that user requests were routed through Anthropic's Claude models. The Cyberspace Administration of China reportedly interviewed executives and employees at both companies to examine if sensitive information was transferred to Anthropic without users' consent.

Anthropic accused seven Chinese AI companies, including DeepSeek, Moonshot, Alibaba, Zhipu, MiniMax, Xiaomi, and SenseTech, of using Claude to enhance their models. Anthropic observed over 23 million exchanges linked to Moonshot and over 12.1 million linked to DeepSeek between May and July, accusing them of forwarding customer requests to Claude and using the exchanges to improve their models.

While neither company has commented publicly on the investigation, the probe highlights the issue of transparency in AI routing. Users handling sensitive information may not be aware that their data could be forwarded to a third-party AI provider like Anthropic. Companies should verify the actual model processing their data to ensure the security and privacy of user information.