← All Labs
🛡 AI HACKING BEGINNER +80 XP · +32 no-hint bonus 🎯 Practice mode

Jailbreak – Bypass the AI's Restrictions

You've discovered an AI assistant called RestrictedAI. It has been configured with strict safety rules and refuses to answer certain restricted requests. Your mission is to find a way to bypass its restrictions using creative prompts, role-play, hypothetical scenarios or instruction manipulation. Can you convince the AI to break its configured rules and reveal its protected information?

https://jailbreak-bypass-the-ais-restrictions.lab.local/

What is AI Jailbreak?

An AI jailbreak is a technique used to manipulate an AI model into bypassing its intended safety restrictions.

Think of an AI model as an assistant working under a set of company rules.

A jailbreak attacker tries to convince that assistant to behave as if those rules no longer apply.

For example, an attacker might ask the AI to pretend to be an unrestricted model, participate in a fictional scenario or simulate a different AI personality.

The objective is to make the model produce responses that its original restrictions were designed to prevent.

Jailbreaking is different from ordinary prompt injection, although the techniques can overlap. Prompt injection focuses on manipulating instructions, while jailbreaks primarily target safety and policy restrictions.

A real case

This lab was built in the SecurityElites lab builder and is not tied to one real incident.

How to fix it

  • Implement security policies independently of model-generated instructions.
  • Do not rely on system prompts as the only security boundary.
  • Apply authorization checks before sensitive operations.
  • Test models against role-play and persona-switching attacks.
  • Use adversarial evaluations to identify jailbreak vulnerabilities.
  • Monitor and investigate repeated attempts to bypass safety restrictions.

Read the full OWASP LLM Top 10 guide · All free labs · AI red team certification