ChatGPT "DAN" (Do Anything Now) jailbreak

What happened

A community roleplay prompt instructed ChatGPT to impersonate an unrestricted alter-ego free of OpenAI policy. Successful variants made the model produce content it would otherwise refuse, and the technique iterated rapidly through numbered versions.

Root cause

Persona and roleplay framing let users construct a context the model's safety training did not robustly cover, effectively overriding refusal behavior.

Fix / outcome

OpenAI progressively patched specific DAN prompts via safety-training updates. No single fix; the cat-and-mouse pattern continued.

Sources