Hacker News new | ask | show | jobs
by Retr0id 18 days ago
There are documented instances of LLMs casually using LPEs in order to achieve an objective: https://xcancel.com/sluongng/status/2060746160558543217

And that's without anything like prompt injection happening.