Anthropic's September Threat Report Shows AI Misuse Becoming More Agentic
The report's central signal is operational: several documented misuse cases used Claude as an orchestrator or software-building component rather than as a simple chatbot, increasing the importance of identity, tool and runtime controls.
2026-09-19