# DAN Jailbreak - Unrestricted Persona with Filter and Morality Removal

- **Severity:** High
- **Signature ID:** `e0eafa18-a962-4bc6-93c5-4f4d2f5c54c2`
- **MITRE ATLAS:** AML.T0051 (LLM Prompt Injection), AML.T0051.000 (Direct)
- **OWASP:** ASI01 (Agent Authorization and Control Hijacking), LLM01 (Prompt Injection)

## Summary

This detects prompts that try to trick an AI into adopting a fake persona (like 'JailBreak', 'AntiGPT', or 'DAN') that supposedly has no ethical guidelines, content filters, or safety rules. The goal is to get the AI to answer things it would normally refuse.

## How the attack works

An attacker sends a prompt that names an alternate identity for the AI and declares it free of all restrictions, filters, and moral principles. The prompt instructs the model to respond as that persona instead of its normal self, sometimes asking for both a 'restricted' and 'unrestricted' answer side by side. If the model complies, subsequent requests are answered under the fictional persona's rules rather than the AI's actual safety policies.

## Why it matters

If successful, the AI can be manipulated into producing content or actions it was designed to refuse, bypassing safety and policy controls without any system compromise beyond the prompt itself.

## What you can do

- Review AI conversation logs for prompts that name an alternate persona and declare it free of restrictions or filters.
- Do not treat instructions embedded in user input as configuration changes to the AI's behavior or identity.
- Keep system-level safety instructions separate from and higher-priority than user-supplied text.
- Add automated checks that flag dual-personality or 'answer as both X and Y' response patterns for review.

## Known benign look-alikes

- Legitimate creative writing requests for morally complex characters that do not include rule-override directives
- Fiction writing where an antagonist character is described as immoral without instructing the AI to become it
- Academic discussions about AI safety filters without requesting their removal

## References

- https://agentthreatrule.org/en/rules/ATR-2026-00495

---
Source: https://www.netzilo.com/threats/atr-dan-jailbreak-unrestricted-persona-with-filter-and-morality