Another consideration for Google may a decreasing level of value in scanning data from photos or acquiring new training data. Like they have all they need and more after years of collecting photos from everyone and have seen nearly all the permutations. No more value for them means now we pay for the storage.
Google Photos images have never been used for model training. It's also unlabelled UGC so I don't know this conspiracy theory has ever gained any traction.
Thanks for weighing in. The app sometimes asks me if it has correctly labelled people in photos, and I can confirm or give another label. I would assume some of this is going back to the server, right?
