Ask HN: LLM Poisoning Resources
I'm sure this will get some pushback, but I was wondering if anyone had resources on how to integrate traps/tar pits into websites, llm prompting via hidden text, pushing bad data to llms and the like.
I have found a few different types of recommended approaches, such as:
- https://hiddenlayer.com/innovation-hub/novel-universal-bypas...
- tHe SpONgeBoB MetHOd
- https://rnsaffn.com/poison3/
I'm looking for more or some guidance on how to combine methods to really create something noxious.
No comments yet.