

Yes, I’m comparing the threat level based on the maximum potential akin to the likes of “those apps”. Permissions are straightforward and will protect users just like ad blockers, decentralized static frameworks (JavaScript/CSS/fonts), and clearing cookies. But on average users are not well informed and aren’t considering permissions, add-ons, or even which browser or app they use so I compare based on the potential threat level.
AI is essentially an algorithm that looks at a set of data which it “trains” on and then uses that predictive model to reproduce a facsimile of an answer. An analogy would be if you took a parrot and only sever said compliments to it you would get a parrot that says compliments except it doesn’t actually understand what its saying just that it should. Knowing that, what they’re saying is that they want your data to train on so here’s some crap you don’t want or need in the hopes you think its cool or useful.
If enough people give up their data they could probably make a model that is actually useful, at which time they’ll turn it into a paid product to replace the people who were naive enough to provide their data for training.