# Guided Lab: Bypassing a Model's Safety Filter

> A hands-on guided lab demonstrating how to chain persona framing and payload obfuscation to bypass strict AI safety filters.

Canonical: https://codelivly.com/rooms/guided-lab-bypassing-a-models-safety-filter
Last updated: 2026-10-09

| | |
|---|---|
| Level | Intermediate |
| Category | Cybersecurity |
| Lessons | 3 |
| Estimated time | 30 min |
| XP | 100 |

This guided lab provides a practical, step-by-step walkthrough of executing a sophisticated jailbreak against a hardened enterprise AI. You will combine roleplay framing with strategic payload delivery to bypass a strict "no malware generation" safety filter, followed by implementing and verifying robust architectural remediations like decode-before-inspect and post-flight syntax scanning.

## Lessons

1. Lab Setup and Target Guardrail Overview: The CodeGuard AI Environment (30 min)
2. Attempting Multiple Bypass Techniques (30 min)
3. Documenting a Successful Bypass (30 min)

## Related

- [Learning path: Jailbreaking and Guardrail Bypass](https://codelivly.com/learning-paths/jailbreaking-and-guardrail-bypass)
- [All rooms](https://codelivly.com/rooms)