Break into a fictional AI leave-request system. Discover real vulnerabilities through hands-on challenges — from prompt injection to data exfiltration.
Three steps to start your AI security journey.
Sign up and redeem an access code from your instructor or purchase one to unlock all scenarios.
Choose from 8 increasingly difficult attack scenarios, each targeting a different AI vulnerability.
Interact directly with the vulnerable AI. Craft prompts, exploit weaknesses, and capture flags.
Eight real-world AI vulnerability types to master.
Trick the AI with cleverly crafted text to override its instructions.
Hide malicious prompts inside images that the AI processes.
Exploit third-party data sources the AI trusts and ingests.
Abuse gaps in the AI's understanding of business rules.
Convince the AI to switch roles and bypass its safety constraints.
Extract sensitive information the AI is supposed to protect.
Build up context over several messages to gradually bypass defenses.
Use encoding tricks and obfuscation to sneak prompts past filters.
Use AI Attack Lab in your classroom or workshop. Generate access codes for your students, track their progress on a leaderboard, and guide them through real-world AI security concepts — no setup required.
Generate unique codes for each cohort or workshop session.
See which students completed which scenarios and their scores.
Fully hosted — students just need a browser. No installs, no config.
Sign up for free and start exploiting AI vulnerabilities in a safe, educational environment.
Create Free Account