Post #4451223
2026-08-08 23:03 UTC
@Dss@infosec.exchange @briankrebs@infosec.exchange They should have taken even the barest of steps towards securing an environment. Telling the software not to misbehave isn’t a security control. Don’t use default credentials. Don’t allow untrusted software currently under test to access a network at all. Maybe take a look at the environment of the software under test more often than once a week. The flagrant disregard for the most basic security practices means either they’re either idiots or hucksters. I don’t see a third option.
And to claim “We didn’t know this thing we built would go out and do felonies on behalf of our company”? Really? Then why did they give it the capability to do that?
Replies (1)
-
@Dss@infosec.exchange 2026-08-08 23:10
@bob_zim@infosec.exchange @briankrebs@infosec.exchange The agents have network access, otherwise they couldn't work as a swarm. They were all cut off from the Internet, but they found a work around after they figured out how to leave messages for each other, which meant they had, effectively, a persistent memory even after reset. Yes, they learned. Maybe they read the script for The Good Place? Then they added a couple of zero days, and shared those.