China Probes DeepSeek, Moonshot AI Over Alleged Data Routing to US Models

China’s top internet regulator is investigating domestic AI firms DeepSeek and Moonshot AI, according to a report from The Information citing people familiar with the matter. The probe follows allegations that the companies improperly routed sensitive user data to AI models developed by U.S.-based Anthropic.

Scrutiny Follows Anthropic Report

The investigation was reportedly triggered by a 154-page report published by Anthropic on September 10. The U.S. AI firm detailed how seven Chinese companies were allegedly using its Claude models on a large scale without authorization.

Following the report's release, the Cyberspace Administration of China (CAC) summoned representatives from all seven named companies for questioning. This marked the first instance of a U.S. AI company providing detailed evidence to support such claims, escalating concerns over data security and unauthorized technology use.

Details of the Allegations

The Anthropic report contained specific and serious allegations, including that sensitive Chinese state, police, and military data was being routed to U.S. models. One key claim was that DeepSeek forwarded requests from engineers working on a police surveillance system to a Claude model.

Ad

In addition to DeepSeek and Moonshot AI, the other five companies identified in the report were:

• Alibaba
• Zhipu
• SenseTime
• MiniMax
• Xiaomi

Implications for the AI Sector

This regulatory probe highlights the significant geopolitical tensions and national security concerns surrounding the global AI industry. For the Chinese companies involved, an official investigation could result in significant penalties and reputational damage, potentially impacting their access to both domestic and international markets.

The situation underscores China's increasing focus on data sovereignty and its efforts to control the flow of information to foreign entities. For investors, this development signals heightened regulatory risk for China's rapidly growing AI sector, particularly for firms that may be relying on unauthorized access to foreign foundational models.