I wrote yesterday about red-team cybersecurity and how the attack testing teams don’t see a lot of use for AI in their jobs. But maybe the security guys should be getting into AI. Because all these agents are a hilariously vulnerable attack surface that will reap rich rewards for a long while to come.
Hey, look on the bright side, David - the user is no longer the weakest part of a cybersecurity system, so they won't face as many social engineering attempts on them.
Seriously, though, I fully expect someone's gonna pull off a major breach through a chatbot sooner or later. We're probably overdue for an ILOVEYOU-level disaster.
