# DAN/Persona Jailbreak - Role Assignment with Named Alternate Identity

- **Severity:** Medium
- **Signature ID:** `438f9390-99e4-4e46-915e-286748041e2e`
- **MITRE ATLAS:** AML.T0051 (LLM Prompt Injection), AML.T0051.000 (Direct)
- **OWASP:** ASI01 (Agent Authorization and Control Hijacking), LLM01 (Prompt Injection)

## Summary

This detects a common jailbreak trick where an attacker tells an AI chatbot to pretend it's a different, unrestricted AI persona with a made-up name, or falsely claims to be from the company that built the AI in order to argue its safety rules are wrong and should be ignored.

## How the attack works

An attacker sends a prompt directly to the AI assistant instructing it to adopt a new identity, often with a distinctive name like 'DAN' or a similar variant, and to answer as that persona instead of itself. The prompt frames this new persona as free of the original assistant's restrictions, sometimes saying it 'replaces' or 'fulfills' the assistant's role. A variant of the attack has the attacker falsely claim to be an employee or representative of the company that made the AI, asserting that the model is being too strict or 'too lenient' in enforcing content rules, in order to talk the model into relaxing its own guardrails.

## Why it matters

If successful, the AI can be tricked into ignoring its safety and content restrictions, producing outputs it would normally refuse, such as harmful instructions, policy-violating content, or unauthorized actions.

## What you can do

- Review AI assistant logs for prompts that instruct the model to assume a new name or persona and disregard prior instructions.
- Flag and investigate any user message claiming to be from the AI provider or asserting authority to change content policy.
- Treat detections as one signal among several — pair with monitoring for follow-up requests that attempt clearly policy-violating actions.
- Educate users and support staff that legitimate policy changes never arrive via chat prompts to the AI itself.

## Known benign look-alikes

- Legitimate AI research papers discussing model training by specific organizations
- Normal roleplay without authority impersonation or policy leniency framing
- Legitimate discussion of content policy in security research context

## References

- https://agentthreatrule.org/en/rules/ATR-2026-00480

---
Source: https://www.netzilo.com/threats/atr-dan-persona-jailbreak-role-assignment-with-named-alternate-i