Detecting and countering misuse of AI: September 2026

Details

Anthropic's Threat Intelligence team published its September 2026 report covering operations disrupted between December 2025 and August 2026 — the broadest scope of any of its threat reports to date. Seven harm categories are documented: cyber operations, influence operations, surveillance, scams and fraud, biological misuse, conventional weapons development, and illicit distillation. Threat actors span suspected state-sponsored groups, financially motivated criminals, commercial spyware vendors, state propaganda institutions, and politically motivated individuals. The most notable specific disclosures include a network of fake dating apps engineered to defraud users and surveillance systems built to identify and monitor dissidents. The report introduces 'Generative Threat Groups' (GTGs) — Anthropic's new classification framework for persistent AI-augmented threat actors. Claude Haiku, Sonnet, and Opus models were all exploited. Notably, Claude Fable and Mythos-class models were not involved in any confirmed misuse except one illicit distillation case, suggesting access controls on frontier models are holding. Intelligence was shared with government authorities and industry partners. Anthropic frames the disclosure as a transparency obligation and calls on other developers to use the case studies to recognize similar patterns. The company explicitly warns that AI misuse risks will increase as models become more capable unless developers and society's defenders act.