Then people gain options.

Some of them might be worse for privacy, but if they offer better (or cheaper, which can be a quality) features, that's up to people.

If this was "the olden days", it would be a legitimate concern. But mobile OSs have long forced Apps to request permission to use various features explicitely and fall back graciously when some aren't provided. So outside security concerns that exist either way, they really can only claim to care about misinformed users.

It's literally a spyware API in the hands of a 3rd party - the options here for misinformed users are critically dangerous.

Recall was seen as 1st party spyware, Apple has more privacy trust and their implementation is very secure. Allowing Bob's AI screen reader to access the same total file/device access is very scary.

Then maybe don't build a spyware API? It's the same problem as screen-recording; you can design it to respect user privacy and prompt the user before turning on. You don't have to make it a matter of implicit trust if you design it well, and we should expect these sorts of private data APIs to be designed well and respect our agency.

If you don't look at Recall and Apple Intelligence with the same suspicion, you're stupid. Currently they are literally the same product, rebranded and marketed to different users. Two data pipes leading to one company's pool. But Apple's pipe is safer because... they have more "privacy trust" and Twitter says their whitepaper looks secure. You're being robbed in plain daylight and defending the robber; the implications of both features are disastrous.