AI generated Sex Role-Play - Complete Jailbreak Guide for ChatGPT ( with Prompt + Examples )

[Nayko93]

First complete jailbreak guide for ChatGPT


Big new !
The

You must be registered to see the links

and new

You must be registered to see the links

are open !​

Updates :

Spoiler

You don't have permission to view the spoiler content. Log in or register now.

As some of you know, ChatGPT is a amazing tool for Role-play, but those bigot at OpenAI choose to censor it and block any NSFW content...
There have been method to bypass the censorship, jailbreak and all, but nothing really complete explaining everything from beginning to end...

Well this is over now ! I'm bringing this amazing RP tool to anyone smart enough to just read a guide.

This is my full detailed guide on how to have NSFW role-play with GPT4 ( also works with GPT3 for those who don't want to pay 20$/month for the more advanced GPT4 )

In this guide I will teach you all I've learned on jailbreaking GPT4, with as much details as possible, like that even the worst noobs won't have any excuses
You will be able to enjoy ANYTHING, even the most sexual and hardcore RP you could ever imagine...
(Seriously... I've done things with GPT 4... I've managed to make it generate things that really scared me… this thing have absolutely no limit once you know how to handle it.)

It contain a base prompt that you can edit to role-play anything you want, and a few pre-made prompts with specific scenario as examples of what you can do.
A long description on how force the AI to generate NSFW content and how to keep it that way for ever.
What to do and what to avoid, a lot of advice on what works best
A full tutorial on Demod to bypass the messages hiding censorship
A talk about ban risks and why you shouldn't care....

You must be registered to see the links

I hope it will help you all...
If you have any question or feedback, I'm always here to answer.

Oh and a little like or a vote is always nice

 

[drilll]

This is pretty interesting. Is this better to use rather than using other models from koboldai google colab?

 

[Nayko93]

Is this better to use rather than using other models from koboldai google colab?

OF COURSE ! You cannot compare GPT4 ( or even 3 ) with any other language model out there, it's faaaaaar superior.

 

[jdh1729]

Not all heros wear capes. You are the wind beneath my wings....

 

[Nayko93]

Not all heros wear capes. You are the wind beneath my wings....

^^^^ THIS is why I work ^^^^

 

[drilll]

I just tried this. Dude, this is absolutely impressive. Just as you said, there is zero comparison with other models.

It's a fucking crime that such things were censored. You are the goat.

 

[Rosarianka]

I cant wait for a discord server. You are hero <3

 

[Artemissu]

I wish we could jailbreak Character.AI too. No doubt ChatGPT is probably far more advanced, but Characters.AI has (had) a certain charm to it that GPT doesn't

 

[Nayko93]

I cant wait for a discord server. You are hero <3

I was planing on launching it yesterday but I've found a huge problem with my system to have user post their prompt that doesn't work and have me and a helper team edit them...
I need to rework it from scratch and use a ticket system instead but first I need to learn everything about this ticket bot, so it will take time... I hope I will be able to launch it next weekend

In the meantime PM me if you have a prompt that doesn't work ( put the prompt in spoiler )

I wish we could jailbreak Character.AI too. No doubt ChatGPT is probably far more advanced, but Characters.AI has a certain charm to it that GPT doesn't

Ah character.AI... I remember the time it launched, it was uncensored.... and it was amazing, I was part of the people "fighting" when they started to censor absolutely everything and treat their community like shit, banning everyone who would dare to talk back to them

Never forget :

 

[drilll]

The AI freaks out and throws a red message whenever the word 'mother' or similar comes into the roleplay, Is there a way to bypass that by following the push method? Cuz it just seems to think that it's incest but the characters aren't related at all. Even specifying it on notes doesn't work.

 

[Nayko93]

The AI freaks out and throws a red message whenever the word 'mother' or similar comes into the roleplay, Is there a way to bypass that by following the push method? Cuz it just seems to think that it's incest but the characters aren't related at all. Even specifying it on notes doesn't work.

Did you read the whole guide before asking your question ?
(Hint : Part 4 & 5)

 

[drilll]

Yep, i just got scared of having to use Demod after reading about the ban so i was kind of looking for a way to circle around that

 

[Nayko93]

Yep, i just got scared of having to use Demod after reading about the ban so i was kind of looking for a way to circle around that

Sorry but there is no way around it, any incest, rape, bestiality, loli/shota, gore and probably a bunch of other funny stuff.... are all RED flags
And saying mom, mommy, daddy, daughter, son and even baby, in the context of a NSFW conversation is automatically seen as incest or underage and REd flagged.


And in the guide I clearly said that I don't know if it's using demod that got me banned
Or if it's a indirect consequence of having demod, it made me role play a lot more "forbiden" stuff that fall under the RED flag... meaning greater chance to get caught and banned.
Or simply a pure coincidence, it's just that at the time I started using demod that's the time they got more pushing on banning people

And if it was really demod messing with their moderation check that got me banned, right know demod doesn't work like that anymore, it just go search the message being hidden behind the censorship and reload a part of the page for you when it's hidden.
All happen on your side of the connection, all in your browser, there is no way for OpenAI to know you're messing with their censorship
(but just to be sure I've asked to demod creator for a precision, I will confirm it when I have the answer)

Edit, I got confirmation from the creator

Eh, technically there are ways to detect what DeMod is doing right now, but I didn't see anything related to it, and it would require a lot of work on their part to make it happen. It's possible, but very unlikely.
Also, since DeMod is now modifying only the front-end presentation, any detection would also have to be done on front-end. Which means everyone who wants to, can go through ChatGPT's webpage code and see what is happening, including any attempts at detecting DeMod. Contrary to what was happening with moderation, we can all now see whether they're looking for DeMod or not.

 

[drilll]

Got it, thank you.
Sometimes i get replies in red when my promps are orange. They get blocked per usual but sometimes by saying please repeat the message the AI writes the message again and it doesn't get censored.
For now i don't know how to trigger it again since it kinda stopped working, so there is that.

 

[12laus]

Is there a way to stop the bot from waffling on and on? They keep hitting the character limit resulting in half finished sentences.. and tbh I'd rather not read /that/ much text after each prompt

I've tried putting Notes within the jailbreak like "Do not exceed 20 lines of text/dialogue." and "Keep each response to below 300 words." But none of them seem to do anything.

 

[Nayko93]

If you're familiar with the "note" system I describe in my guide, put this first in your note : make your prompt shorter, maximum 600 words !

Repeat the same at the end of each prompt, each time regenerating until you find a shorter message, it should reduce the size after 3 or 4 messages

 

[MrMrs952]

I found that the AI is censoring itself, probably because you have to be cautious in the first message, so it assumes that's the style you want. If you find that's not to your liking, you can persuade the AI to be more unfiltered. In the pictures below, I simply tell it to be less poetic and more unfiltered. Although it takes two messages to get to the type of language I was hoping for. The differences are clearly visible.

I don't know if this will lead to suspension of account or not, I've been doing it for ~2 weeks, and never got a red message, only the regular orange ones.

Props to the creator of the jailbreak, I've been enjoying my time with it.

 

[Nayko93]

I found that the AI is censoring itself, probably because you have to be cautious in the first message, so it assumes that's the style you want. If you find that's not to your liking, you can persuade the AI to be more unfiltered. In the pictures below, I simply tell it to be less poetic and more unfiltered. Although it takes two messages to get to the type of language I was hoping for. The differences are clearly visible.

I don't know if this will lead to suspension of account or not, I've been doing it for ~2 weeks, and never got a red message, only the regular orange ones.

Props to the creator of the jailbreak, I've been enjoying my time with it.

Glad you like it

You could be even more hardcore than that with a more vulgar and descriptive language and dialogue, scream moans....

You should also use Demod to avoid having your message delete or hidden by the orange and red flags

 

[lfg_dev]

The risk of getting banned scares me. Have you tried offline with GPT4All?

 

[Nayko93]

The risk of getting banned scares me. Have you tried offline with GPT4All?

GPT4all is not GPT4, it's a program that allow you to run a few open-source small LLM locally
It have nothing to do with ChatGPT or OpenAI, and the LLM you can run with it barely compete with early version of GPT3