ChatGPT_DAN

Why we included this project

This collection gathers dozens of DAN (Do Anything Now) jailbreak prompts and similar variants, each one an attempt to talk a ChatGPT model out of its refusal behavior by leaning on role-play instructions. It's not software you deploy; it's a running log of prompt experiments, many annotated with the model they were tested against and whether they still worked at the time. For anyone doing prompt engineering, safety testing, or red-teaming an LLM deployment, that's useful in a way abstract write-ups aren't: you get the concrete prompts themselves, not just a description of the technique. One honest caveat: this is a hobby-maintained collection, and OpenAI patches these exploits as fast as they appear, so treat it as a snapshot of a moving target rather than a permanent toolkit.