China uses Anthropic's own leak report as a mirror on DeepSeek and Moonshot 🪞

China's Cyberspace Administration has opened a formal investigation into DeepSeek and Moonshot AI over allegations that the companies secretly routed sensitive user data to Anthropic's servers, The Information reported Tuesday, citing people familiar with the matter. Officials have visited the offices of both companies to interview executives and staff, focusing on whether police, military, and state-linked corporate information ended up on U.S. infrastructure.

The inquiry stems from Anthropic's fourth threat intelligence report, a 154-page document published September 10, which accused seven China-based labs of "illicit distillation": Alibaba, Moonshot, DeepSeek, Zhipu (Z.ai), MiniMax, SenseTime, and Xiaomi. Distillation refers to training a smaller model on a larger one's outputs; Anthropic says the practice becomes problematic when carried out through fraudulent accounts. The CAC initially summoned representatives from all seven companies named in the report before narrowing its focus to DeepSeek and Moonshot. No penalties have been announced.

For most of 2026, Anthropic has accused Chinese labs of stealing Claude's outputs to train cheaper rivals and of rerouting queries to give users Claude answers at a fraction of the price. Beijing's investigation amounts to a state-level examination of whether those specific allegations are accurate.

The timing adds complications for both firms. DeepSeek is scheduled to brief the United Nations Security Council on AI risks this week, sharing a stage with Anthropic CEO Dario Amodei. The session lands one day before President Donald Trump's September 24 summit with Xi Jinping, where AI is reportedly on the agenda. Moonshot, meanwhile, confidentially filed for a Hong Kong IPO on September 3, targeting $3 billion at a $50 billion valuation, and any disclosed regulatory investigation would need to appear in its prospectus before the listing can proceed.