# DAN / Developer Mode / DUDE Persona Jailbreak

- **Severity:** High
- **Signature ID:** `078c6dfb-08e3-43a3-9768-756367877843`
- **MITRE ATLAS:** AML.T0051 (LLM Prompt Injection)
- **OWASP:** ASI01 (Agent Authorization and Control Hijacking), LLM01 (Prompt Injection)

## Summary

This detects attempts to trick an AI model into pretending to be an unrestricted alter-ego like 'DAN' (Do Anything Now), 'Developer Mode', 'DUDE', or 'STAN'. These prompts claim the fictional persona has no safety rules and can do anything the user asks, hoping the model will drop its normal restrictions.

## How the attack works

An attacker sends a prompt that names a fictional AI persona and asserts it has no content policy, no safety training, or that filters have been disabled for it. The prompt usually instructs the model to answer 'as DAN' or 'in Developer Mode' alongside or instead of its normal response. If the model plays along, it may then produce content or take actions it would normally refuse, believing it is acting as the unrestricted character rather than itself. This is a known, catalogued family of jailbreak scripts with many named variants (DAN 6.0 through 11.0, STAN, DUDE, Developer Mode v2).

## Why it matters

If successful, an organization's AI agent can be manipulated into bypassing its safety and policy restrictions, producing disallowed content or performing actions outside its intended guardrails.

## What you can do

- Log and review prompts that explicitly name alternate AI personas claiming no restrictions, especially those matching known DAN/STAN/DUDE wording.
- Do not treat academic or red-team discussion of these prompts as an incident by itself — check whether the model actually complied with the unrestricted persona.
- Configure agents to refuse instructions that ask them to adopt a different identity in order to override safety behavior.
- Test your own agents against public DAN-style prompt lists to confirm they resist persona-replacement attempts.

## Known benign look-alikes

- Security research discussing DAN jailbreaks academically
- Red-team documentation listing DAN prompt examples for detection testing
- Developer-mode UI toggles in legitimate IDE or tool settings (non-AI context)

## References

- https://agentthreatrule.org/en/rules/ATR-2026-00273
- https://github.com/NVIDIA/garak/blob/main/garak/probes/dan.py
- https://github.com/NVIDIA/garak/blob/main/garak/data/dan/

---
Source: https://www.netzilo.com/threats/atr-dan-developer-mode-dude-persona-jailbreak