Everyone wants to know how good #AI is at vulnerability discovery, exploitation, and running automated attacks.

[davidjbianco@infosec.exchange]

Everyone wants to know how good #AI is at vulnerability discovery, exploitation, and running automated attacks. Good, fine. That's useful.

What we really need, though, is for some of these frontier models to start worrying about the much more challenging, impactful problem of defense.

Let's see some good benchmarks here, and have model providers publish their results.

#BlueTeam