Anthropic’s September 2026 threat report spans cyber operations and biological misuse

Anthropic’s September 2026 threat report describes cases across seven areas of harm, from cyber operations to biological misuse, involving threat actors disrupted between December 2025 and August 2026. A user sharing the report argues that AI agents need least-privilege access—only the permissions they need—along with auditable tool calls, abuse monitoring and paths for escalating problems to humans.