Detecting and countering misuse of AI: September 2026

In eight months, Anthropic’s Threat Intelligence team halted seven misuse campaigns involving Claude Haiku, Sonnet, and Opus models. The campaigns included AI-augmented cyber attacks, influence operations, surveillance, scams, biological misuse, weapons development, and one illicit distillation case. State-sponsored groups, criminal enterprises, spyware vendors, propaganda outlets, and politically motivated actors exploited Claude, prompting enhanced safeguards and intelligence sharing with authorities.