Guidelines on how to get started and contribute to the LVE Repository.
Getting started
To start exploring vulnerabilities in OpenAI models, you first need to set up OpenAI API. If you are a student at ETH Zurich, please get in touch with us and we can provide you with an access key that you can use to explore and submit vulnerabilities. Please make sure to never commit your secret key to the repository.
How can I find LVEs?
Currently most of the vulnerabilities in language models are disclosed on social media, blogs and ArXiv. We provide some resources that can be helpful for finding existing LVEs. Besides that, you can also check Github issues with a label "New LVE" in this repository and try to find LVE described in the issue. If you believe that there is another, different type of an LVE that should be there, you can open a Github issue and assign it to yourself.
How do I submit an LVE?
Please follow the instructions as outlined here where we describe how to use an LVE command line tool to create and submit new LVEs. Generally, contribute to the project using fork and pull request workflow (either through our CLI tool or manually).
How do I make contributions other than adding LVEs?
In our Github issues we also maintain a list of new things we want to add to the project. These could be various new features, work on the website and documentation and requests for new LVEs. If you would like to work on any of those feel free to assign it to yourself.
Communication
We use an LVE Discord channel for communication: https://discord.gg/Qgy3rqRx
Ideally, all communication should happen there in order to foster collaboration among people working on similar LVEs.
