Hacker News new | ask | show | jobs
by nicce 392 days ago
We should handle LLMs as insider threat instead of typical input parsing problem and we get much better.
1 comments

All text input is privileged code basically. There is no delimiting possible.