Chinese AI labs secretly used millions of Claude exchanges to train their models, Anthropic says
Anthropic accuses Chinese labs of harvesting millions of Claude exchanges to train rival models—the first major lab-race allegation of its kind.
Why it matters
This is frontier lab-race drama with real stakes: unauthorized model distillation at scale raises questions about API security, competitive intelligence, and whether the frontier labs' moats are eroding faster than expected. Practitioners deploying Claude need to understand the implications for their own data.
The key facts
5 to knowAnthropic detected unauthorized use of Claude by China-based labs including Alibaba and Moonshot AI
Alleged use case: extracting Claude outputs at scale to train competing AI systems
Date: September 11, 2026
Source: CNBC (mainstream coverage signals this is a significant accusation)
Lab-race competitive allegation (Anthropic vs Chinese frontier competitors)
Go to the source
CNBC Technologycnbc.com
Publisher excerpt: Anthropic said it detected unauthorized efforts by China-based AI labs including Alibaba and Moonshot AI to use its Claude models to help improve their own AI systems.