Anthropic has accused Chinese developers Moonshot and DeepSeek of quietly routing their own customers' questions to Claude, then showing the answers as if their models produced them.
Key Points:
- Anthropic's 154-page threat report attributes more than 23 million Claude exchanges to Moonshot and 12.1 million to DeepSeek.
- Moonshot allegedly relayed almost 300,000 customer requests through 5,380 fraudulent accounts in a single 10-day window.
- The company says the diverted traffic exposed user names, corporate data and one apparent military surveillance request.
Anthropic Report Names Moonshot, DeepSeek
The company released the findings Thursday in a threat intelligence report covering misuse it says it detected between December 2025 and August 2026. The document runs to 154 pages and sorts that activity into seven categories.
In one 10-day window, Moonshot relayed nearly 300,000 customer requests to Claude through 5,380 fraudulent accounts, most of them appearing to sit in Singapore and Japan.
Anthropic bars access to its models from inside China. It attributed more than 23 million exchanges to Moonshot between May and July. DeepSeek accounted for more than 12.1 million over 14 days in July.
Also Read: XRP ETFs Hit $1.7B Record As 9-Week Inflow Streak Defies Price Weakness
Joe Khawam Flags Legal Risk
Joe Khawam, managing director of legal and AI policy at the Law Reform Institute in Washington, said the report raises privacy questions that could implicate both American and Chinese law.
He called some of the details potentially relevant in a legal sense. Khawam, a former State Department lawyer, also read the timing as a sign Washington will press the issue at US-China AI talks this month.
Anthropic said the rerouted sessions carried names, email addresses and company data belonging to users in the United States, Europe and China. One request, apparently from someone tied to the People's Liberation Army, asked Claude to assess surveillance footage of a person in Chengdu. The company added that it does not know whether Moonshot told customers their queries were leaving its systems.
Distillation Claims Against China's AI Labs
Anthropic first published distillation allegations in February, counting more than 16 million exchanges through roughly 24,000 fraudulent accounts and naming DeepSeek, Moonshot and MiniMax. The September tally reaches about 190 million across seven labs, with Alibaba alone responsible for more than 151 million.
Beijing has rejected that framing. China's Commerce Ministry said the accusations carry no factual or legal basis. It warned of countermeasures if Washington moves to contain the country's AI companies.
Read Next: Sam Altman Rules Out A 2026 OpenAI IPO And Blames AI Safety Work
