Episode Description

What would it take to keep advanced AI systems—and their model weights—secure?

Lisa Thiergart, co-founder and CEO of Security Level Five, joins Rocket Drew to examine the threats facing frontier AI labs, from espionage and insider attacks to model theft through distillation and AI agents that could exploit their own data centers.

They discuss why conventional cybersecurity may not be enough, what a purpose-built secure AI data center might look like, the limits of encryption, and why even extreme security measures may not contain superintelligence.

Chapters

00:00 Intro

02:00 Espionage and insider threats at AI labs

07:55 What is Security Level Five?

11:25 Model theft and distillation

19:46 Securing models before deployment

24:01 When AI becomes the insider threat

27:29 Neocloud security and rogue agents

32:27 Building a secure AI data center

45:26 Can encryption protect AI infrastructure?

53:57 Open models, alignment and existential risk

Guest Correction: PHP is a server-side scripting language, not a protocol. SL5's SENL Framework is for the private industry, not public industry.