AI & LLM Security

Who this course is for

If you come from a security background — pentesting, threat modeling, cryptography — and want to apply that thinking to AI systems specifically, this course is the bridge. It builds on the Cryptography and Penetration Testing roadmaps, so you should already be comfortable with adversarial thinking and security fundamentals before starting; the AI-specific parts (prompt injection, jailbreaks, agent tool-use risk) are taught from scratch.

What you'll actually build

The course works through five real problems, in order: mapping the LLM threat landscape (prompt injection, jailbreaks, data exfiltration), red-teaming AI applications the way an attacker actually would, designing defensive architecture for agentic systems that hold tool-use permissions, auditing model and data supply chains for tampering risk, and applying AI governance frameworks like the EU AI Act and NIST AI RMF in practice rather than as theory. Each module pairs a concepts lesson with a hands-on exercise attacking or defending a realistic system.

The capstone

You finish by conducting a full security audit of a live agent application — the same kind of engagement a security team would actually be asked to deliver, documented to a professional standard you could put directly in a portfolio or use as a work sample.

Why this matters right now

AI and LLM security is one of the fastest-growing, most understaffed security specialties that exists right now — most organizations shipping agentic systems have no one on staff who has actually red-teamed one, and the compliance side (EU AI Act, NIST AI RMF) is becoming a hard requirement rather than a nice-to-have. Security engineers who can speak both languages — traditional security and AI-specific risk — are in short supply.