China’s Cyberspace Administration has opened a probe into DeepSeek and Moonshot AI over whether they sent sensitive Chinese data to Anthropic’s Claude models, according to Qianer Liu and Jing Yang at The Information. The probe follows a September 10 Anthropic report alleging that Moonshot quietly passed over 23 million user requests to Claude from May through July and DeepSeek sent over 12 million in two weeks of July, presenting the answers as their own. Anthropic also said Moonshot forwarded footage from hundreds of Chengdu police cameras on behalf of a user it believes is tied to the People’s Liberation Army. The regulator questioned all seven companies named in the report before narrowing in on the two, and it has yet to decide on penalties.
