←back to thread

Comet AI browser can get prompt injected from any site, drain your bank account

(twitter.com)
645 points helloplanets | 1 comments | 24 Aug 25 15:14 UTC | HN request time: 0s | source
Show context
alexbecker ◴[24 Aug 25 16:32 UTC] No.45005567[source]▶
I doubt Comet was using any protections beyond some tuned instructions, but one thing I learned at USENIX Security a couple weeks ago is that nobody has any idea how to deal with prompt injection in a multi-turn/agentic setting.
replies(1): >>45005703 #
hoppp ◴[24 Aug 25 16:47 UTC] No.45005703[source]▶
Maybe treat prompts like it was SQL strings, they need to be sanitized and preferably never exposed to external dynamic user input
replies(7): >>45005949 #>>45006195 #>>45006203 #>>45006809 #>>45007940 #>>45008268 #>>45011823 #
1. internet_points ◴[25 Aug 25 09:05 UTC] No.45011823[source]▶
SQL strings can be reliably escaped by well-known mechanical procedures.

There is no generally safe way of escaping LLM input, all you can do is pray, cajole, threaten or hope.