Aug 21, 2026 · 35 min
AI agents widen the attack surface from endpoints to PLCs
The Department of Know: Living off Azure, OpenAI's "defender window," and AI-driven PLC attacks
Organizations are being pushed to deploy AI faster while securing agents, cloud services, mobile hardware, and industrial systems that traditional controls cannot fully contain.
- 1AI agents need sandboxing, behavioral detection, and endpoint protections because network controls alone cannot contain their actions.
- 2Zero trust must extend across SaaS, mobile hardware, government systems, and small businesses facing limited security resources.
- 3AI-generated exploitation scripts could accelerate attacks on industrial systems, making centralized governance and guardian agents increasingly important.
Don't miss
The panel examines how threat actors can use AI to generate exploitation scripts for Siemens S7 programmable logic controllers.
The brief
Bil Harmer and David B. Cross join the Department of Know panel to examine the central security problem of rapid AI adoption: organizations must move quickly without surrendering control.
The discussion broadens from government accountability to Unisoc modem risk, social-media oversharing, SaaS exposure, and the cybersecurity gap facing small businesses.
The panel argues that AI agents belong in sandboxes, with behavioral detection and endpoint protections supplementing zero trust rather than relying on network controls alone.
The sharpest turn comes with AI-generated exploitation scripts for Siemens S7 PLCs, linking modern automation to the long-running danger of exposed operational technology.
The proposed response is centralized governance, a multi-model strategy, shared agent rules, and guardian agents that inventory, monitor, alert on, and block unsafe activity.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.