Red Team your AI Agents

by Michael Morgenstern

Abstract

I’ve been leading the security charge at a variety of companies to secure this incredible new technology. But like every other incredible new technology, usage is outpacing security and everyone is racing to catch up. In the meantime, you can all have lots of fun breaking these tools! Let’s talk about how!

I will discuss a variety of TTPs [Tactics, Techniques, and Procedures -ed] for prompt injection, showing several powerful infections to exfiltrate data on several different popular AI tools.

And I’ll do a couple (very quick) demos!

I promise you’ll leave this talk wanting to go break all the Agents you can find (with at least a few more ideas on how to do so).

Reference

• Cloud Security Alliance: Agentic AI Red Teaming GUIDE