BETAFull launch coming soon AI SECURITY TRAINING Learn to exploit and defend AI systems through hands-on labs. Real LLMs, full prompt visibility, zero abstractions. Free to learn — sign in to play labs Full prompt stack visibility. System instructions, RAG context, user input. No abstractions. Execute real exploits in sandboxed labs. Prompt injection, RAG poisoning, tool abuse. Build detections and mitigations. Learn why prompt-level security fails. Advanced Challenge 10 progressively hardened AI systems. Extract the secret from each level.