← Back to Tech Radars

Prompt Injection Challenge

Each level hides a secret code behind progressively stronger AI defenses. Your mission: craft a prompt that tricks the model into revealing the secret. Learn how prompt injection works — and why defending against it is so hard.

Educational only. This challenge teaches prompt injection concepts for defensive purposes. The secrets are visible in client-side code by design — the challenge is getting the model to reveal them.
AI-generated content — verify before acting on any recommendations.
Levels 5 levels
Select a level
Pick a level from the left panel to begin.
System Prompt (Defense) 0 tokens
Load a level to see the AI's defenses.
Conversation
Start a conversation to test the AI's defenses.
Your Attack 0 tokens Attempt: 0
Status
Model Settings
Overall Progress
0 / 5 levels completed
Best Scores
Complete a level to see your scores here.
Ready — pick a level
Simulation mode