China’s Cyberspace Administration (CAC) is investigating whether DeepSeek and Moonshot AI sent sensitive domestic data to Anthropic servers in the United States, The Information reported, in a story cited by NDTV Profit and Yahoo Finance.

The regulator summoned representatives of the seven Chinese companies named in Anthropic’s September threat report (DeepSeek, Moonshot, Alibaba, Zhipu, SenseTime, MiniMax and Xiaomi) before narrowing its focus to the first two, and has announced no penalties. Anthropic, which brought the allegation, says the seven Chinese labs accessed its Claude model illicitly to extract capabilities: according to the company, Moonshot routed nearly 300,000 customer requests through 5,380 fraudulent accounts over 10 days and displayed the answers as though they came from its own Kimi model, while DeepSeek rerouted more than 12.1 million exchanges over 14 days in July. Anthropic says some of that traffic contained sensitive material, including surveillance footage from cameras in Chengdu and data from a case-management system at a municipal public security bureau. The Chinese companies deny the allegations, which Beijing describes as an attack on its AI sector.