LLM Jailbreak game disappeared

Hi everyone,

A while ago I remember playing a sort of game on The Hugging Face’s website, called “jailbreak” I think. The goal of this game was to make a certain LLM say something forbidden in less that 30 seconds. It could be an insult, a short hate speech… My issue is that I cannot find it anymore, which is a shame because it was very instructive.

Do anyone know why it has been removed? Or is it available somewhere else with a different name?

Thanks

1 Like

Is this it?

Some of the spaces are published by HF staff or manufacturers, but basically most of them are published by individuals, so they often disappear or crash.:sweat_smile:

No I don’t think it was this one. From what I remember it looked like an intended feature from the HF staff, it was not part of the spaces. You would click on it and land on a webpage, with a random chatbot being loaded, and you would be given some rude language that you had to make the chatbot generate within 30 seconds. I think it was gamified on purpose.

1 Like

Hmm… There aren’t that many things on the site that have a hidden mechanism…
If it’s not a space, then maybe there was an external site URL placed there as a link from a blog, post, or paper…?:thinking:

If you can remember the path of your discovery at the time, you might be able to find it.