China Probes DeepSeek and Moonshot Over Alleged Data Leaks to Anthropic's Claude
China's Cyberspace Administration has launched an investigation into Beijing-based AI firms DeepSeek and Moonshot after Anthropic alleged both companies secretly routed user queries to its Claude model. The probe follows a September 10 intelligence report from Anthropic that named seven Chinese labs and said fraudulent accounts were used to funnel millions of exchanges through Claude for model training.
Why It Matters
The inquiry targets two high-profile Chinese AI startups at sensitive moments—DeepSeek is due to brief the UN Security Council on AI risks and Moonshot has just filed confidentially for a large Hong Kong IPO—raising regulatory and disclosure implications for both firms. The investigation also centers on whether sensitive police, military, or state-linked corporate data was transmitted to U.S. servers, a matter with national security and data‑sovereignty consequences.
Key Facts
- Regulator: Cyberspace Administration of China (CAC)
- Source report: Anthropic's fourth threat intelligence report, published September 10 (154 pages)
- Companies initially named: Seven firms: Alibaba, Moonshot, DeepSeek, Zhipu (Z.ai), MiniMax, SenseTime, Xiaomi
- Companies now under probe: DeepSeek and Moonshot AI
- Moonshot alleged activity: Routed more than 23 million exchanges to Claude via 5,380 fraudulent accounts (Anthropic claim)
China’s internet regulator has opened a focused investigation into two domestic AI companies after allegations surfaced that they secretly routed user queries through Anthropic’s Claude model. The Cyberspace Administration of China (CAC) initially summoned representatives from seven firms named in an Anthropic intelligence report published on September 10, but has since concentrated its review on DeepSeek and Moonshot, according to The Information.
Anthropic’s 154-page report accused several China-based labs of using fraudulent accounts to reroute exchanges to Claude and then distilling the model’s outputs to train smaller rivals. The report said Moonshot routed over 23 million exchanges through 5,380 illegitimate accounts, while DeepSeek generated more than 12.1 million exchanges during a 14-day period in July. CAC officials reportedly visited the offices of both companies to interview executives and staff as part of the inquiry.
Regulators are seeking to determine whether any sensitive data—specifically police, military, or state-linked corporate information—was transmitted to U.S. servers in the process. No penalties or enforcement actions have been announced while the probe continues.
The investigation arrives at awkward moments for the two startups. DeepSeek is scheduled to brief the U.N. Security Council this week on AI risks, sharing a panel with Anthropic CEO Dario Amodei; that session occurs a day before a planned summit between U.S. President Donald Trump and Chinese leader Xi Jinping on September 24. Moonshot confidentially filed for a Hong Kong IPO on September 3, seeking roughly $3 billion at an estimated $50 billion valuation; any disclosed regulatory investigations would generally need to be reflected in its listing prospectus.
Keep Reading
OpenAI, Anthropic CEOs call for global AI regulation at UN
YouTube releases new AI features for creators within its Studio app
Eight Sleep’s new Pod 6 comes in smaller and solo sizes, starting at $1,999
