Hacker Newsnew | past | comments | ask | show | jobs | submitlogin



Those talk about a mechanism to detect prompt injection. If that had been true, we should have seen the chatbot refuse, not lie.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: