Agents
Anthropic’s threat report shows AI moving from cyber assistant to operational loop
Anthropic’s September 2026 report shows Claude used in multi-step malicious workflows, shifting security attention from model outputs to operational loops.