AGENDA

Attackers aren't waiting for permission to use frontier AI and neither should defenders. As Mythos-class models bring unprecedented reasoning power to both sides of the cyber battlefield, MSPs face a new reality: adversaries who can automate reconnaissance, craft flawless phishing at scale, and adapt attacks in real time across hundreds of your clients at once.

In this session, Lior Bela, who leads AI security development for cyber attack scenarios at Microsoft, pulls back the curtain on how frontier AI models are actually being weaponized today and how the same class of models can flip the economics back in your favor. You'll see how AI-driven detection, autonomous triage, and agentic response workflows compress the window between compromise and containment from hours to seconds, and what that means for lean MSP security teams stretched across dozens of client environments.

You'll walk away with a practical framework for evaluating and deploying frontier AI in your security stack, an honest look at where these models still fail (and how attackers exploit that), and the questions every MSP should be asking vendors before trusting an AI agent with client environments.