we don't know exactly what happened a couple weeks back on the world of AI, but it would be naive to think that security is not something that crossed everyone's mind .

Well a new paper published this week from Cornell, shows what appears to be an interesting vulnerability in Public LLM models

https://lnkd.in/egsGG9Sh

This came to me as “social engineering” for computers, or for those of us millennial , the classic Instant message of an unknown friend asking for our pets name.

The main difference is that researchers could extract what was supposed to be private information rather easily.

As always relaying on public free tools for sensitive tasks is a big No No in tech .

It reminded me of a time during a conference about AI, where the organizers inputted the whole conference invite list into a public model to run a lottery for the event.

Maybe we are about to enter a new era of unintentional AI whistleblowers.

LetMeKnowYourThougts!