The Information says China’s internet regulator is probing DeepSeek and Moonshot AI after Anthropic alleged they routed millions of user exchanges through Claude to train their models. Officials reportedly interviewed executives and staff at both companies.
Officials reportedly want to determine whether sensitive police, military and state-linked corporate data reached U.S. servers. Anthropic’s Sept. 10 threat-intelligence report accused seven China-based labs—Alibaba, Moonshot, DeepSeek, Zhipu (Z.ai), MiniMax, SenseTime and Xiaomi—of “illicit distillation.” Anthropic says distillation, training a smaller model on a larger model’s outputs, is legitimate in itself; its objection is to alleged use of fraudulent accounts. The CAC first summoned representatives of all seven, then narrowed its focus to DeepSeek and Moonshot. No penalties have been announced.
Moonshot confidentially filed for a Hong Kong IPO on Sept. 3, targeting $3 billion at a $50 billion valuation. Any disclosed regulatory investigation would need to appear in its prospectus before the listing could proceed. DeepSeek is scheduled to brief the UN Security Council on AI risks this week alongside Anthropic CEO Dario Amodei, the day before the Sept. 24 Trump-Xi summit, where AI is reportedly on the agenda.
