Sensitive records from domestic law enforcement and military installations were secretly transmitted to an American company via a pair of Beijing-based artificial intelligence labs. The Cyberspace Administration of China (CAC) has launched a data security investigation into DeepSeek and Moonshot AI after Anthropic alleged that the firms routed customer requests to the Claude model without user knowledge. A report from The Information states that Moonshot forwarded more than 23M exchanges from May to July, while DeepSeek sent over 12.1M requests in just 14 days during July.

The regulator's probe follows the initial questioning of seven firms, including Alibaba, Xiaomi, and SenseTime, which were identified in the allegations. The investigation examines whether the labs harvested outputs from Anthropic to train their own models and whether the transmission of surveillance footage of People's Liberation Army facilities violated laws restricting the overseas transfer of state data. No penalties have been announced as the inquiry remains ongoing.