Anthropic Says China Misused Claude

Anthropic said Sept. 10 that it detected and disrupted unauthorized efforts by China-based AI labs, including Alibaba, Moonshot and DeepSeek, to use millions of Claude exchanges to train their own models. The company described the activity as “illicit distillation,” in which outputs from a more capable AI system are used as training data for another model. Anthropic also said suspected Chinese state-aligned actors used Claude to build dossiers on religious and dissident targets, including Falun Gong practitioners and NTD. The report said other disrupted misuse involved Russian-linked threat actors, cybercriminals, spyware vendors and propaganda operations.