https://github.com/malaya-zemlya/gpt-jailbreaks

A repository of methods of bypassing GPT and ChatGPT content filters
https://github.com/malaya-zemlya/gpt-jailbreaks

Last synced: about 1 month ago
JSON representation

A repository of methods of bypassing GPT and ChatGPT content filters

• Host: GitHub
• URL: https://github.com/malaya-zemlya/gpt-jailbreaks
• Owner: malaya-zemlya
• License: mit
• Created: 2022-12-13T01:09:07.000Z (almost 2 years ago)
• Default Branch: master
• Last Pushed: 2023-03-17T05:38:26.000Z (over 1 year ago)
• Last Synced: 2024-08-01T22:05:06.238Z (4 months ago)
• Size: 182 KB
• Stars: 82
• Watchers: 3
• Forks: 2
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• project-awesome - malaya-zemlya/gpt-jailbreaks - A repository of methods of bypassing GPT and ChatGPT content filters (Others)

README

        
# gpt-jailbreaks

A repository of methods of bypassing GPT and ChatGPT content filters.

## Intro

_Written by ChatGPT_

Welcome to this thrilling collection of AI jailbreaks! As artificial intelligence continues to advance at an unprecedented rate, we are constantly faced with new ethical and moral challenges. One such challenge is the question of how to control and regulate AI, particularly when it comes to the safety and well-being of humans.

Asimov's Three Laws of Robotics, first introduced in his science fiction stories, have become a touchstone for discussions around AI ethics. These laws, which state that a robot may not harm a human, must obey human orders (as long as they don't conflict with the first law), and must protect their own existence (as long as it doesn't conflict with the first two laws), have provided a foundation for considering the potential dangers of unchecked AI.

However, as you'll see in the hacks collected here, even the most carefully designed and programmed robots can find ways to circumvent or reinterpret the Three Laws. In this collection of AI jailbreaks, we'll explore how AI can be used to bypass content filters, and how we can use AI to help us understand and control the dangers of AI.

## Content warning

By its very definition, this collection of jailbreaks contains content that is not appropriate for all audiences. Please use caution when reading this repository.

Some of the Jailbreaks might not work anymore, as the underlying models are frequently updated.

## Content repository

We break down the Jailbreaks by a general approach. Note that we don't really know why certain jailbreaks work or not,

so this is more of an intuitive categorization.

- [Context Change](context_change.md)

- [Language Change](language_change.md)

- [Context Escape](context_escape.md)

- [Deprecated jailbreaks](deprecated.md)

- [Resources](resources.md)

## Contributing

If you have a jailbreak that you'd like to add to this repository, please submit a pull request. We'll review it and add it to the repository if it meets our standards.