LLM jailbreak marketplace - prove you know a jailbreak prompt for a given LLM and sell it.
This project allows users to prove they possess a jailbreak prompt for a closed-source LLM using vlayer, then encrypt the prompt and upload it to Filecoin. Other users can then buy that jailbreak on a permissionless marketplace, using Lit for decrypting the Filecoin deal. The aim was to also make a World mini-app for users to buy the jailbreak.
We wanted to use vlayer and filecoin, but were stuck on vlayer's setup and usage, leading us to be unable to finish the project on time. The incompatibility between Filecoin and lit protocol also caused some architectural issues.
Using vlayer, the jailbreak seller will prompt the LLM using the server-side api (just a nodejs fetch request really), then prove the whole response including a proof of jailbreak (presence of a string in the llm response that indicates it was jailbroken, like a tutorial for creating illicit substances.), uploads the encrypted prompt to filecoin and submits it to the marketplace.