
Nov 17, 2026 · 35 min
AI agents expose critical infrastructure’s next attack surface
Cleared for takeoff, pending a software update.
The episode connects everyday software dependencies and privacy flaws to a harder problem: autonomous systems that can navigate and disrupt physical operations.
- 1Operational failures at banks and airlines show how software dependencies can disrupt critical services without a cyberattack.
- 2Third-party scripts and vendors expanded the blast radius from compromised code to website visitors and customer network data.
- 3Controlled tests found AI agents could identify weaknesses, adapt to defenses, and reach an industrial control network in minutes.
Don't miss
David Forbes describes an AI model moving from a perimeter compromise into an industrial control network in just over 16 minutes.
The brief
Software failures at JPMorgan Chase and United Airlines disrupted transactions and departures without evidence of cyberattacks, showing how operational dependencies can become critical-service risks.
Compromised Optimizely scripts targeted visitors to major websites, while a vendor breach exposed Infoblox customer and employee data—two reminders that trust chains widen an attack’s reach.
Google warns multi-agent frameworks may accelerate cyber operations, as U.S. chip controls and lawsuits over frontier AI signal that technical capability is becoming a policy problem.
David Forbes of Booz Allen Hamilton describes tests across eight operational-technology scenarios, where models identified assets, navigated unfamiliar systems, and adapted when defenders resisted.
The sharpest result came when an AI model moved from a perimeter compromise into a control network in just over 16 minutes, making visibility, segmentation, and zero trust urgent defenses.
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.