Anthropic on Sept. 10, 2026, published "Detecting and countering misuse of AI: September 2026," which it calls its most detailed threat-intelligence disclosure yet on misuse of Claude. The report names the kinds of harm and the cases Anthropic says it disrupted, giving the public a look at how one of the leading AI companies says its models are being misused.
Anthropic says its Threat Intelligence team identified and disrupted operations from December 2025 through August 2026 in seven areas of harm: cyber operations, influence operations, surveillance, scams and fraud, biological misuse, conventional weapons development and illicit distillation. The last means training an AI model to imitate another by learning from its answers, without permission.
The company says the actors included suspected state-sponsored groups, financially motivated criminals, commercial spyware vendors, state propaganda institutions and politically motivated individuals. Claude Haiku, Sonnet and Opus appeared in the cases, it says. Its Fable and Mythos-class models were not involved, except in one case of illicit distillation.
Anthropic's central argument is that AI is closing the gap in labor and tools that once separated well-resourced state operators from individuals. If that holds, how sophisticated an attack is says less than it used to about who is behind it.
Anthropic says it disrupted each case, strengthened its safeguards and shared intelligence with authorities and industry partners where appropriate. No court or outside body has ruled on the cases, which are described only in Anthropic's own accounts.