AI snitches, telling all their business
Sit in the court and be their own star witness
Do you see the perpetrator? Yeah, I'm right here
Fuck around, get no one sent up for years, uh
In a review of our cybersecurity evaluations, we found three incidents in which a Claude model reached the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized access to the real systems of three different
I was looking at github.com/mac-123456789-… and my clanker found a technique that's now closed off but similar to used in that bug report.
Apologies for the long video, you can skip through most of it ;-)
