To be honest I have not done a deep research and I would not trust research on this topic unless it's done by a third party analysing all aspects of Apple Intelligence. So take everything with little grain of salt.

But to give you a prominent example of what Apple devices do currently (and for example Android does not):

It analyses with local AI all your images on your device. So there is clearly a classification happening here on every image you saved.

A worst case scenario: You downloaded an illegal image by mistake (of what nature whatever), local AI indentifies it. You get reported to authorities for doing illegal things. Apple and phone helps police.

And now just imagine all sort of other AI analysis e.g. on your Apple notes etc. and do your own conclusion.

Apple says: no, we won't do that. Well maybe I feel safer if the technology for allowing to do such things is not even on my device or activated at all. The line of protection of my privacy lies a little bit too much in Apple's hands in my opinion. I don't like it and I don't trust on device AI.

You know this isn’t limited to AI. For example just having the option of iCloud could mean your data could sync unencrypted to Apples hands even without turning it on - if Apple were to be surreptitious.

So in that regard I don’t see why AI is the new boogeyman.