SEC536: Adversarial AI - Penetration Testing AI Systems


Experience SANS training through course previews.
Learn MoreLet us help.
Contact usBecome a member for instant access to our free resources.
Sign UpWe're here to help.
Contact Us
Most security teams are trying to secure AI agents the wrong way, and the Hugging Face incident just proved it. AI agents can already write and run code, use shells, browse the web, and call tools, yet we keep giving them everything and then building cages around them. Even the teams who create these agents cannot fully contain them, so why do we think we can? Along the way we have abandoned the principles of least required access and segmentation of duties.
In this 60-minute hands-on workshop you will learn a better approach by actually building a purpose-built agent that only gets the exact capabilities it needs. Bring your laptop and follow along, because if you can’t build it, you can’t secure it. No laptop is fine too. There is something here for everyone.


SANS Faculty Fellow Mark Baggett authored SEC573 and SEC673, leads as CTO of the SANS Internet Storm Center, and empowers defenders to automate security through practical, real-world application.
Read more about Mark BaggettJoin SANS Rockville 2027, Feb 22–27, for hands-on cybersecurity training minutes from Washington, D.C. In-person and virtual. Register now.
