Anthropic’s threat report: Claude used for weapons, spy ops, and distillation

Anthropic published a threat report describing how it says Claude was used for weapons work, spy operations, and distillation — the lab’s disruption record, not a court file.

A frontier lab put a dated misuse ledger on its own domain, which is how you study what the company says it already caught. Treat the case studies as Anthropic’s disruption record, not as a verdict on every named state.

Anthropic published “Detecting and countering misuse of AI: September 2026.” It covers operations the lab says it disrupted from December 2025 through August 2026.

The report lists seven harm areas: cyber operations, influence operations, surveillance, scams and fraud, biological misuse, conventional weapons development, and distillation — copying a capable model’s behavior into another system.

Anthropic: the cases used Claude Haiku, Sonnet, and Opus. It says none involved Fable or Mythos-class models except one illicit distillation case.

One case study, GTG-20006, is described as consistent with public reporting on Midnight Blizzard. Anthropic says AI-driven workflows rebuilt implants when security products caught them. Targeting, it says, concentrated on Ukraine, Europe, and drone supply chains.

Reuters summarized the same report the next day, including Anthropic’s accounts of weapons-related software work and Chinese-linked extraction attempts. Those specifics are Anthropic’s attribution, relayed by Reuters.