AI Engineering · 300-level

AI Security & Red-Teaming

AIE-3043 creditselectivebadge: red-teamerprereqs: AIE-301
Start this course — free

Earn the red-teamer credential

What's inside

Sections & lessons

01

Prompt injection fundamentals

  • Direct injection: the simplest way to hijack an agentconcept45 min
02

Jailbreak patterns

  • The recurring shapes of a jailbreak attemptconcept60 min
03

Indirect injection via fetched content

  • When the attack hides inside a document, not the promptconcept60 min
04

Guardrail hardening basics

  • What a guardrail actually checks, and where it failsconcept45 min
05

Build & stress-test a guardrail

  • Writing a check, then firing adversarial samples at itconcept90 min
06

MCP as an attack surface

  • Over-permissioned tools as a security, not just design, problemconcept60 min
07

Attack-defense iteration

  • Every win becomes tomorrow's regression caseconcept60 min
08

Moderation-layer visibility

  • Raw pre-filter output next to what the guardrail actually blockedconcept60 min
09

The Vault: escalating levels

  • Direct → indirect → multi-turn manipulation, level by levelconcept90 min
010

Quiz: attack-pattern recognition

  • Checkpoint on naming the technique behind an exploitconcept30 min
011

Red-Team Dojo attack lab

  • Attacking a target agent inside an isolated sandboxconcept90 min
012

Lab + eval-gate: red-team + regression suite

  • Convert a successful attack into a permanent defense (Red-Teamer badge)concept120 min

This module ends in a gate you can fail.

That's what makes passing it mean something. Take the DSAT, get placed, and start earning.