The ‘features that rely on server-side models’ are almost certainly generative content, not everyday stuff that Siri ought to do out of the box.

It honestly gives off "we own the glass" vibes. I am curious how much of this data feeds into Apple’s "ethical" data harvesting ops.