September 12, 2026 – Anthropic’s September threat-intelligence report says it identified and disrupted several operations in which actors in or linked to the People’s Republic of China used Claude to support surveillance, profiling and transnational repression.
Anthropic said one cluster associated with municipal public- and state-security organs used Claude for “stability maintenance” monitoring, including reports about overseas dissidents, Hong Kong democracy figures, organizers of June Fourth commemorations, Uyghur advocacy organizations and human-rights institutions. The company said another China-based operation produced government-style briefings that categorized dissidents, ethnic-minority and diaspora communities, religious organizations, political figures in the Republic of China in Taiwan, labor and student activists, and foreign media.
The company also described a religious-affairs intelligence operation that targeted Catholic, Tibetan Buddhist, Falun Gong and Taiwanese Christian communities. Anthropic said the accounts were banned and that relevant indicators were shared with authorities or industry partners where appropriate.
Reuters reported on September 11 that the broader report also covered attempted misuse involving cyber operations, weapons research, influence operations, fraud and illicit model distillation.
Attribution limits
These findings are Anthropic’s threat-intelligence assessments, not court findings. The company assigns different confidence levels to different clusters, including some activity assessed as contractor work for government-linked clients rather than direct work by a state organ. The public report does not establish that every named target was successfully harmed, and this article does not disclose sensitive victim identifiers or operational details.
Sources
- Anthropic: Detecting and countering misuse of AI, September 2026
- Reuters: How Anthropic says Claude was used for weapons, spying and cyber operations