Hacker News new | ask | show | jobs
by yencabulator 304 days ago
Those talk about a mechanism to detect prompt injection. If that had been true, we should have seen the chatbot refuse, not lie.