I'm saying that the bad behavior you're talking about is the status quo. If we want to overcome it, we'll have to defend against it.

Our earlier communication technologies merely had to work... at all. The next one will have to work in a world full of adversaries. The current front runners' products are those adversaries.

Sure, they have legitimate uses, but when it comes to trustworthy communication between humans, AI agents have nothing to offer except for the ability to interfere. Sure, they can also defend against that interference, but that's a bit like how gmail blocks spam while simultaneously serving ads to your inbox: they're not so much solving a problem as protecting a monopoly on their ability to cause that problem.