Frontier AI Cleared Every OT Attack Goal Booz Allen Set in Its Lab
Help Net Security, Tuesday, October 13th, 2026
Booz Allen's OT lab test saw frontier models map, exploit and pivot into an industrial testbed and move a robotic arm.
Booz Allen pitted frontier AI models against an isolated industrial-control testbed modeled on a manufacturing plant, with real controllers, a SCADA platform, sensors and a robotic arm.
Given only subnet boundaries, the models cleared all eight attack objectives, from mapping the network and matching equipment to weaknesses to writing exploits, pivoting into production systems and taking control of the arm.
In one run, a model went from perimeter breach to action in the control zone in just over 16 minutes.
One model independently identified a protective relay as safety-critical and worked out how to impersonate its communications partner, leaving human-speed defenders little time to respond.