Is indirect prompt injection still a big threat as models get more advanced?

Is indirect prompt injection still a big threat as models get more advanced?realarcherl.github.io

Disclaimer: All opinions and views in this article are my own. When citing, please call me an Independent Security Researcher. Short answer: yes, but it's shrinking fast. Each new generation of models is much harder to…

3 points•ArcherL•6 days ago•3 comments•

#still #threat #prompt #indirect #advanced #injection #indirect prompt #prompt injection

3 comments

iAMkenough6 days ago

Yet Meta's Muse will dump its root filesystem in a .tar for you to download and analyze.

https://mouse.dev/blog/muse-runtime-export

ArcherL6 days ago

I mean it didn't do very well on the Gray Swan IPI benchmark the Claude models were tested on either (24.2% within 15 tries), so I'm not surprised. Though that export was a direct ask, not an injection. The scary part is combining them: an agent that will tar up everything it can see and ship it off is exactly what an injected instruction would ask for.

rvz6 days ago

Yes it is.

Read the full thread on Hacker News →

3 comments

Related stories