China investigates DeepSeek, Moonshot over secret data routing to Anthropic’s Claude

China's Cyberspace Administration is investigating DeepSeek and Moonshot AI after Anthropic alleged both companies secretly routed sensitive user data to its Claude models, The Information reported on Monday. The probe follows a 154-page threat intelligence report Anthropic published on September 10, which detailed how seven Chinese AI companies conducted what it described as an industrial-scale campaign to extract Claude's capabilities without authorization.anthropic+2

What Anthropic Found

Anthropic's report alleged that DeepSeek rerouted more than 12.1 million exchanges to Claude over a 14-day period in July, with users believing they were interacting with DeepSeek's own models. The company said DeepSeek used a technique involving Claude's "thinking signature" to obtain reasoning traces that would otherwise have been protected by the model's safeguards. Moonshot AI, the developer of the Kimi family of models, allegedly routed nearly 300,000 customer requests to Claude over a 10-day period through 5,380 fraudulent accounts, most appearing to be located in Singapore and Japan.ndtvprofit

The allegations go beyond conventional model distillation. Anthropic said some rerouted requests contained sensitive information, including data connected to a Chinese municipal police system, a Russian government agency, and internal corporate credentials. The report also named Alibaba ↗ 0.48%, Xiaomi, Zhipu, SenseTime, and MiniMax among firms that engaged in unauthorized distillation, with Alibaba alone accounting for more than 151 million exchanges between May and July.ndtvprofit

Regulatory Response and Geopolitical Context

After the report's release, the Cyberspace Administration summoned representatives from all seven named companies for questioning, though the probe has zeroed in on DeepSeek and Moonshot. U.S. intelligence agencies, including the FBI, issued a joint advisory on September 8 urging American AI labs to safeguard their intellectual property.tradingview+2

The investigation comes at a politically sensitive moment. DeepSeek is scheduled to brief the United Nations Security Council this week on AI risks, alongside representatives from OpenAI and Anthropic. President Trump and Chinese President Xi Jinping are also set to meet this week in Washington, where artificial intelligence is expected to be among the topics discussed. U.S. Treasury Secretary Scott Bessent said in July that the administration would consider sanctioning open-weight models from China if evidence of distillation from American models is verified.reuters+1