Anthropic has published its September 2026 threat intelligence report, revealing how its Threat Intelligence team identified and disrupted threat actors that misused the Claude AI model between December 2025 and August 2026. The activity spanned seven harm areas, including cyber operations, influence operations, surveillance, and illicit model distillation.
Scope of the Report
The report, titled “Detecting and countering misuse of AI: September 2026,” covers activity across cyber operations, influence operations, surveillance, and other harm areas. It details the efforts of Anthropic’s Threat Intelligence team to detect and counter these abuses during the nine-month period.
Seven Harm Areas
According to the report, the misuse fell into seven distinct harm areas. Among them are cyber operations, influence operations, surveillance, and illicit model distillation. The report provides insight into how threat actors attempted to leverage Claude across these categories.
Disruption Efforts
Anthropic states that its Threat Intelligence team successfully identified and disrupted the threat actors involved. The report outlines the team’s approach to detecting and countering the misuse of AI, though specific tactics or outcomes are not detailed in the available summary.
