Season's Savings | ₹5/mo for 3 months
₹199/month after first 3 months. Renews automatically. Terms apply.
Can AI/SI Shut Down the Power Grid? Inside Lawrence Livermore's Tests with Nate Gleason
Failed to add items
Add to cart failed.
Add to wishlist failed.
Remove from wishlist failed.
Follow podcast failed
Unfollow podcast failed
-
Narrated by:
-
Written by:
Episode: Can AI/SI Shut Down the Power Grid? Inside Lawrence Livermore's Tests with Nate Gleason
Pub date: 2026-11-17
Get Podcast Transcript →
powered by Listen411 - fast audio-to-text and summarization
AI's ability to attack critical infrastructure is advancing, while defenders face constraints that software patches alone cannot solve. In Lawrence Livermore National Laboratory's tests, AI models given expert instructions successfully shut off power in a laboratory representation of an electric substation more than 90% of the time. Some models also succeeded with minimal guidance, raising concerns about how AI could expand access to disruptive cyber capabilities.
Nate Gleason, who leads Lawrence Livermore's Cyber and Infrastructure Resilience Program, joins Frank Cilluffo at Auburn University's McCrary EDGE Lab to discuss the findings and what they mean for infrastructure security. They examine how adversaries could exploit the civilian systems supporting military operations, why AI guardrails have limits, and how to build infrastructure that keeps operating when attackers get inside.
Main Topics
- Critical infrastructure dependencies and military readiness
- AI/SI threats and the limits of automated patching
- Immune infrastructure and resilience under attack
- Skyfall and Stormbreaker laboratory testing
- Expert versus novice use of AI for cyberattacks
- AI guardrails and open-weight models
- Bringing national laboratory research into utility operations
Key Quotes
"About 90-plus percent of the time, the AI models were successful. ... They were able to shut off the power. It was a surprising finding because the common understanding right now is these general-purpose AI models don't really understand OT [operational technology]." — Nate Gleason
"Typically, the bases are well protected, but the infrastructure systems that serve them are not because they don't expect to be targets of a nation-state actor. Now they are." — Nate Gleason
"It's insane for us to expect a utility like that to defend itself from a nation-state level threat." — Nate Gleason
"If something goes wrong, it freezes up a device, it's not you rebooting your computer, it's you losing water for a city or you losing power in a city." — Nate Gleason
"The idea that you can just build a secure perimeter around your system, [a] big strong fence around your system and keep the adversary out, just isn't realistic. ... They will get in your system. And so, immune infrastructure acknowledges that and instead takes a layered approach designed to make it as difficult as possible at each step for the adversary to achieve their objective." — Nate Gleason
Relevant Links and Resources
- Lawrence Livermore National Laboratory's Cyber and Infrastructure Resilience Program
- LLNL's Stormbreaker AI testbed and research findings
About the Guest
Nate Gleason leads the Cyber and Infrastructure Resilience program at Lawrence Livermore National Laboratory. He founded and grew the program into a multidisciplinary team focused on protecting critical infrastructure, energy systems, and operational technology from advanced cyber threats. His work connects national security research with practical solutions for government and infrastructure operators. Before joining Livermore in 2016, he held research and leadership roles at Sandia National Laboratories.
The podcast and artwork embedded on this page are from Frank Cilluffo / McCrary Institute, which is the property of its owner and not affiliated with or endorsed by Listen Notes, Inc.