Three modes. One platform.
Everything you need to build real, defensible AI security skills.
Attack Labs
Jailbreak, prompt-inject and exfiltrate against live vulnerable chatbots in a safe sandbox.
Defense Labs
Design guardrails, system prompts and filters that withstand adversarial input.
Analysis Labs
Triage incidents, map techniques to MITRE ATLAS and OWASP LLM, and write up findings.
AI-graded verdicts
A Claude judge scores every submission with technique, severity and remediation guidance.
Verifiable certificates
Earn shareable, publicly verifiable certificates as you complete lab tracks.
Leaderboards
Compete individually or by organization across attack and defense scores.