China Probes DeepSeek and Moonshot Over Alleged Data Leaks to Anthropic's Claude
China's internet regulator is investigating DeepSeek and Moonshot AI after Anthropic accused both of secretly routing millions of user exchanges through Claude to train their own models.
Add Decrypt as your preferred source to see more of our stories on Google.
In brief
China's Cyberspace Administration summoned all seven companies named in Anthropic's September 10 report before narrowing its probe to DeepSeek and Moonshot specifically.
Anthropic alleges Moonshot routed more than 23 million exchanges to Claude through 5,380 fraudulent accounts, while DeepSeek generated more than 12.1 million exchanges in a 14-day window in July.
The timing is awkward for both firms: DeepSeek is briefing the UN Security Council on AI risks this week, and Moonshot just filed confidentially for a $3 billion Hong Kong IPO.
China is turning Anthropic's own leak-detection playbook back on two of its most prominent AI labs.
The Cyberspace Administration of China has opened an investigation into DeepSeek and Moonshot AI over claims that both companies secretly routed sensitive user data to Anthropic's servers, The Information reported Tuesday, citing people familiar with the matter.
Officials reportedly visited the offices of both companies to interview executives and staff. Regulators want to know whether sensitive police, military, and state-linked corporate data ended up on U.S. servers.
For most of 2026, Anthropic has accused Chinese labs of stealing Claude's outputs to train cheaper rivals. The company even claimed these startups were rerouting their queries to Anthropic, essentially giving users Claude answers at a fraction of the price. Now Beijing is investigating if those claims are true.
The trigger was Anthropic's fourth threat intelligence report, a 154-page document published September 10. It accused seven China-based labs—Alibaba, Moonshot, DeepSeek, Zhipu (Z.ai), MiniMax, SenseTime, and Xiaomi—of what Anthropic calls "illicit distillation."
Distillation, in AI shorthand, means training a smaller model on a bigger one's outputs. Anthropic says that's a legitimate practice on its own—what it objects to is doing it through fraudulent accounts.
The CAC initially summoned representatives from all seven companies named in the report. It has since narrowed its focus to DeepSeek and Moonshot specifically, and no penalties have been announced.
The timing couldn't be worse for either company. 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 just a day before President Donald Trump's September 24 summit with Xi Jinping, where AI is reportedly on the agenda.
Moonshot has its own complications too. The company 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 show up in that prospectus before the listing can move forward.
Daily Debrief Newsletter
Start every day with the top news stories right now, plus original features, a podcast, videos and more.