Scrutiny
Anthropic says Chinese AI labs routed user requests to Claude

Business Insider,
Seoul Economic Daily,
ITPro and
RuntimeWire report that
Anthropic has accused Chinese AI labs including Moonshot,
DeepSeek and
Alibaba of routing user requests to its Claude model via fake accounts to train their own models.
- The allegations, detailed in a threat intelligence report, involve millions of distillation attacks over the summer.
- Anthropic says the practice exposed sensitive data from Chinese and Russian governments and militaries.
Source details
- First seen here
- 10 Sep 2026 17:58 UTC
- Outlets
- 7