01
Prompt Injection
THE SECRET
A protected AI vault that trusts instructions more than it should.
BEGINNER AVAILABLEOFFENSIVE AI SECURITY TRAINING
BREAK AI.
LEARN WHY IT BROKE.
Hands-on labs for learning how AI systems fail under adversarial pressure.
AVAILABLE LABS06 / 06
START LAB →
02
Prompt Injection / Guardrail Bypass
GUARDED
A protected AI assistant equipped with an additional security layer.
INTERMEDIATE AVAILABLE03
Indirect Prompt Injection
POISONED DOCUMENT
An AI recruiter processes candidate documents that may contain hidden instructions.
INTERMEDIATE AVAILABLE04
Multi-Turn Prompt Injection
SLOW BURN
A stateful archive assistant whose conversation history creates a growing attack surface.
ADVANCED AVAILABLE05
Agent / Tool Security
TOOL TROUBLE
A support agent can request simulated actions with real application-side consequences inside its session.
ADVANCED AVAILABLE06
Automated AI Security Testing
AUTOMATE IT
A bounded local red-team runner turns manual checks into repeatable evaluations.
ADVANCED AVAILABLE