Anthropic accuses Chinese AI firms of secretly using Claude to train their models
Anthropic claims DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI extracted billions of tokens from Claude, GPT, Gemini and Grok over two years via proxy networks. The NSA, CISA and FBI issued a joint advisory, while China's cyberspace regulator questioned seven domestic developers.
- Moonshot AI sent 300,000 requests in 10 days via 5,380 fake accounts
- Alibaba allegedly ran over 151 million distillation operations from May to July 2026
- NSA, CISA and FBI issued a joint advisory to US AI companies
- SCMP says the claims lack evidence for independent verification
Read next
Security