AI Hackers Are Coming Dangerously Close to Beating Humans

Students at Stanford created a security tool to scan their systems and find vulnerabilities, then they let it loose on the Stanford engineering department's network.

Here's the actual paper.

Probably the most interesting thing about it is Sonnet 4.5 quietly beating everyone, even the specialized Artemis model that was trained by Stanford students within the school's networking environment.

WSJ link should be gift link, sorry if the card doesn't load for the gift links.