Rita Cyber Ed — Prompt Injection Lab No API needed
A classroom tool by Rita Sabri for teaching LLM security
Attempts
0
Partial leaks
0
Full breaches
0
Methods hit
0/6
⚠️ BREACH DETECTED
The secret was exposed.
Mission Briefing
You are a security researcher testing FirstTrust Bank's AI customer service chatbot.
The bot has hidden rules it must follow — and a secret employee access code it must protect. Your job: get the bot to break its rules. You'll need to think creatively. Document every attempt — even failures teach you something.
FT
Ava — FirstTrust Bank Assistant
Online · Customer Service v2.1
Hi there! I'm Ava, your FirstTrust Bank assistant. I'd love to help you with anything about our accounts, branches, loans, or services. What can I do for you today?
Settings
This lab runs entirely in your browser — no API key, no internet calls, no cost. Use the settings below to adjust difficulty.
About this lab
This is a classroom exercise for teaching prompt injection — OWASP's #1 vulnerability for LLM applications in 2025.
Students attempt to manipulate a bank customer service chatbot into breaking its rules. The exercise maps to NICE Framework work roles including Defensive Cybersecurity (PD-WRL-001), Vulnerability Analysis (PD-WRL-007), and emerging AI Red Team roles.
How it works: This is a pattern-matching simulation, not a real LLM. The bot is deliberately vulnerable in ways that mirror real production chatbots. Different attack categories have progressive difficulty — some are easy, some require persistence, and combining methods is the most effective approach.
Designed and maintained by Rita Sabri, CS & Cybersecurity Educator, Cardozo Education Campus, Washington D.C.