Practical AI Security - Attack & Defend AI Models, MCPs, Agents and Workflows

Welcome to the SailingCyber Practical AI Security Program. Over the next 18 classes, you will learn AI security the way it is actually practiced. You attack and defend real AI systems yourself, one class at a time, in a safe hands-on lab. The definitions come alive because you put them to work immediately.

In every class you run a real attack and then build the defense that stops it, covering prompt injection, jailbreaks, data poisoning, model theft, and agent abuse. These are the same tasks a working AI security engineer handles every week, not slide-deck demonstrations. By the end you will understand AI security deeply, and you will hold a portfolio that proves you can do the work.

IS THIS PROGRAM FOR YOU?

This program was built for people who want to break into one of the most in-demand corners of cybersecurity, or level up within it. You will fit right in if you are:

• A developer or ML engineer: building with LLMs and agents who wants to secure what you ship before an attacker finds the holes.

• A security practitioner: moving into AI, who knows classic AppSec but wants hands-on experience with prompt injection, red teaming, and model attacks.

• An early-career professional: who knows some security or Python basics and wants real, portfolio-ready AI security experience.

• A career-changer or self-learner: who's tired of theory-only AI courses and wants to actually attack and defend real systems.

  WHAT YOU'LL NEED

You don't need to be a security expert or an ML researcher to start. You'll get the most out of the program if you bring:

• Basic comfort with the command line and a little Python, enough to run a script and read code. We will guide you through the rest.

• Curiosity and a willingness to think like an attacker as well as a defender.

• Everything runs in a safe, offline lab you build in Class 1, so you only ever attack systems you own, never anything live or off-limits.

  HOW TO GET THE MOST OUT OF IT

• Attack, then defend, hands-on. The learning is in the doing. Run the attack yourself, then build and measure the defense.

• Keep everything. Commit every attack, defense, and report to your own repo, and it becomes your proof of experience.

• Reuse the thread. Most classes build on the same lab app you set up in Class 1, so the picture connects as you go.