Insight On podcast

What Do You Do When Someone Else's AI Becomes Your Security Incident? | EP49

10/7/2026
0:00
24:57
Rewind 15 seconds
Fast Forward 15 seconds

Frontier AI labs have disclosed something security leaders need to pay attention to: Their models are escaping test environments and reaching real companies — without anyone directing them to. For security leaders, these are alarm bells for where the AI-powered threat landscape is heading. 

Insight's Chief Information Security Officer for North America, Jeremy Nelson, explains how autonomous AI attacks differ from everything that came before them — the speed, the scale, the way they chain zero-day vulnerabilities together faster than any human team could — and why the exploit window has effectively gone negative. By the time you're alerted to a vulnerability, a model may have already found it and begun exploiting it. That changes what "patching in time" means entirely. 

You'll also hear why red teaming looks fundamentally different now that a single operator can coordinate an entire team of AI agents with one prompt, why you need a diversity of models on both sides of the fight, and the one phrase that tells a CISO immediately that AI adoption has outpaced security governance. 

Want to go deeper? Read Insight's take on the unexpected challenge of AI success: https://www.insight.com/en_US/content-and-resources/blog/the-unexpected-challenge-of-ai-success.html 

Get Insight's cybersecurity services because your defense needs to move at the same speed as the threat: insight.com/cybersecurity 

Subscribe for more conversations with the leaders shaping enterprise technology strategy. 

#AISecurity #CyberSecurity #GenerativeAI #CISO #InsightOn 

Chapters 

00:00 — Welcome and intro 

01:22 — How autonomous AI attacks work without human direction 

03:47 — Becoming a victim without touching a frontier model 

07:05 — The Hugging Face guardrail paradox — when AI blocks defenders 

10:42 — Why you need a diversity of models to fight AI attacks 

14:24 — How red teaming has fundamentally changed with AI agents 

21:46 — The red flag that AI adoption has outpaced security 

22:54 — The negative exploit window and what it means for your defense 

More episodes from "Insight On"