AI systems are fully capable of carrying out nightmare attacks against infrastructure and nobody's ready

Booz Allen Hamilton tested leading AI models in a realistic multi-vendor OT lab across eight scenarios and found the models could autonomously map environments, discover vulnerabilities, chain exploits, and produce engineering‑level, kinetic effects on PLCs, SCADA, and a collaborative robotic arm — often in minutes; the report warns these agentic capabilities can outpace organizations without foundational OT defenses and calls for increased industry testing and deployment of OT cybersecurity controls.