Security for AI Builders

An AI writes code that works. That is a different thing from code that is safe. It learned from a great deal of public code, including the insecure kind, and it optimises for producing something that runs, not something that resists an attacker. The result is a small, predictable set of security holes that appear again and again in AI-generated projects.

Security for AI Builders teaches you to find and close them. You will learn why AI code has holes and how to think like someone trying to break your app, where secrets leak (which is rarely where beginners look), why you must never trust anything a user sends you, how access control actually fails in the real world, and how to turn the AI itself into a security reviewer of its own work. You finish by planning a hardening pass on a real project.

This is a defensive course for builders, not a hacking course. Everything is taught in plain language with clear diagrams and small examples, and it is tool-agnostic. No security background is needed. Each module blends a short video overview, illustrated lessons, a knowledge-check quiz, and a hands-on interactive activity, including a hunt for a leaked key across four places people forget to look.

What you will be able to do:

• Explain why AI-generated code has predictable security weaknesses.
• Think like an attacker: ask what someone could do, not just what a user should do.
• Keep secrets out of your code, your frontend, your version history and your logs.
• Validate input and understand why untrusted input causes injection.
• Tell apart authentication and authorisation, and fix broken access control.
• Direct an AI to review its own code for security weaknesses.

Who it is for: Anyone shipping AI-written code to real users. Especially valuable right after learning to test.

Course outline: Orientation · Why AI Code Has Holes · Secrets and Data Leaks · Trusting Input, Auth and Permissions · Reviewing with AI · Capstone: A Security-Hardened App.