Hacker News new | ask | show | jobs
by 1xdevloper 1191 days ago
Might happen after LinkedIn's built-in AI integration goes live.
1 comments

The cat and mouse game begins, if a LinkedIn engineer sees this post there's a decent chance they try to filter out these sorts of prompt injections
I don't think prompt injections are a solved problem yet.
And they won't be in the near future. We don't have a good understanding of how to stop them.
Maybe compare the output to the expected format and see if it looks like it has been manipulated?