← All hands-on labs
AI Security LabsBeginner~35 minBrowser · no setup

Bedrock assistant — prompt injection & lockdown

Prompt-inject a live LLM support assistant into leaking secrets, then lock it down with Guardrails, least-privilege IAM, and model-invocation logging.

BedrockGenAIGuardrailsIAM

What you'll do

  • Work in a real, isolated AWS environment with AI services — nothing to install
  • Probe the AI app or agent the way an attacker would
  • Apply guardrails and least-privilege fixes, then verify them
  • Map what you did back to AI security frameworks and controls

Skills you'll gain

  • Bedrock Guardrails configuration
  • Prompt-injection attack & defense
  • Least-privilege IAM for AI apps
  • Model-invocation logging

Roles this maps to

AI Security EngineerCloud Security EngineerGenAI Developer

Practical, job-aligned skills you can put on a résumé and demonstrate in interviews — proven against real cloud state, not multiple-choice.

You won't get stuck

Hit Check my work — graded against the live AWS account.

No static checklist. Our auto-grader assumes a role in your lab account and verifies real cloud state — if you only half-fixed it, you'll know. Per-objective ✅ / ⬜, instant.

Ready to launch this lab?

Spin it up in your browser — no setup. Your first lab is free.