Anthropic Details AI Misuse Patterns in September Threat Report

Anthropic has published a new threat-intelligence report covering malicious activity it says it disrupted between December 2025 and August 2026. The September report follows earlier disclosures in March, August and November 2025, and the company frames it as a set of case studies rather than a measure of how its models are typically used.

The report describes cases across cyber operations, influence campaigns, surveillance, scams, biological misuse, conventional weapons development, and illicit distillation. Anthropic says the cases are notable examples rather than a measure of typical use. The actors it describes include suspected state-sponsored groups, financially motivated criminals, commercial spyware vendors, state propaganda institutions and politically motivated individuals, with examples ranging from a network of fake dating apps built to defraud users to surveillance systems designed to identify and monitor dissidents.

Two trends stand out in the cyber section. First, Anthropic says AI has collapsed the labour and tooling gap that once separated well-resourced state operations from individuals: a hacktivist using stolen API keys, loosely connected criminals and a state espionage operator each sustained multi-victim campaigns that would previously have required teams of specialists. Second, the company says a majority of the operations it describes used AI for direct execution or orchestration through multi-agent frameworks, not just as a chatbot answering questions. It notes that publicly available offensive agent frameworks now reproduce much of the scaffolding it first documented in a suspected state-sponsored campaign in November 2025.

Anthropic says the misuse involved its Claude Haiku, Sonnet and Opus models, and that none of the cases involved its Fable or Mythos-class models except one distillation case. In each case, the company says it disrupted the activity, strengthened its safeguards and shared intelligence with authorities and industry partners where appropriate.

For operators and builders, the report is a reminder that capable models change the speed and scale of misuse as well as legitimate work. Read Anthropic's report.