News for hackers. Stuff that matters.

A comment on LLMs can hide arbitrary undetectable information in their responses.

Re:LLMs can hide arbitrary undetectable information in their... by cryptohell on Monday January 22, 2024 @04:51PM (#39095986)
ChatGPT could encode usernames, timestamps and other session context into its responses in a way that would only be retrievable by OpenAI and provably invisible to everyone else
[ Parent | on HN ]

There are no comments on this comment yet.