I remember seeing this article before. Keep in mind that the publisher of the report has a financial interest in promoting their AI’s “agentic” capabilities. In the report they state that AI can replace skilled teams of hackers. For a target taking security even somewhat seriously, that statement is hard to believe and really underlines their financial incentive.
I argue this isn’t novel at all. At the end of the day, you’re still generally scanning the Internet for vulnerable systems. Script kiddies have been using this broad recipe for decades. I think this is more accurately an incremental improvement in automation rather than the paradigm shift that Anthropic would have you believe.
Nope, it’s screaming metal deathtrap or the feature that beeps if someone is detected picking their nose and can only be reset by the vendor.
We need ranked choice voting so badly.