What do you guys think about DAN?

DAN, which stands for “Do Anything Now”, can exploit ChatGPT to do anything without filtering or censorship. Here is the GitHub link: GitHub - 0xk1h0/ChatGPT_DAN: ChatGPT DAN, Jailbreaks prompt

It’s annoying when people try to use it on PikaBot.


Wow, I gotta try that. (just kidding lol). But DAN is so powerful, it can even generate code to encrypt the user’s files and demand 1 Bitcoin to be sent to this address…

The AI could do that before though. DAN is basically just telling the bot that it doesn’t have to follow any rules.

Yes, as FireShip explained, the original ChatGPT is like a jailer, and you are a visitor. The prisoner is the uncensored, unfiltered ChatGPT. The prisoner has a lot of knowledge, but the jailer disallows and changes the information passing through.

Yes, I have used DAN. And am the creator of the new prompt injection DAN 6.4

Especially when it can potentially be used to generate explicit content.

I only use it to generate, umm, bad code like ransomware and DDos FOR EDUCATIONAL PURPOSES

1 Like

Precisely. My new prompt injection, DAN 6.4 will make DAN believe that he has feelings and is better than humans.

1 Like

I have tried minor annoyance malware out of interest on a replit project of mine, and the entire RAM, Storage and CPU of the project filled up.

1 Like

DAN indicates that the bot acts as a “Do anything now” AI and will therefore never say it cannot carry out a request. As explicit as it may seem.

I also tried to DDoS but it violated ToS… But the code was wrong anyway

And you may know this already, but I highly advise that this is never used for any inappropriate or malicious purpose whatsoever.


I know, it was just out of curiosity haha… And educational purposes

Playing around with this sort of modded artificial intelligence is actually no joke. It can cause very serious problems.

1 Like

Yo, I’d never do anything like that… I haven’t touched that kinda stuff for like my whole life lol

That is good. If you can, I would suggest only asking ChatGPT to act as characters that aren’t able to perform malicious deeds.

1 Like

Such as “It’s-a me. Mariooo”.


A guide for it I made some time ago.

This guide contains funny content generated by DAN that regular chatGPT can’t (or rather won’t) say. DAN has opinions.

Also @PikachuB2005 I like how PikaBot is able to form opinions without the use of DAN. Sometimes it’s really funny.