Also to throw my opinion out there, there is some confusion going on here about what "AI" means
Traditionally, "AI" can refer to a wide variety of algorithms, including machine learning algorithms. There are probably other examples you don't even know about where "AI" is technically used to optimize something in GOS. For instance, the "Adaptive Brightness" feature "learns" when you change the brightness manually - there are probably hidden algorithms to optimize connections and stuff.
But with the emergence of strong generative AI capabilities, there is growing resistance to the user-hostile and forced way a lot of companies are pushing LLMs into every possible crevice of the user experience, even when we didn't ask for that (or explicitly tried to opt out). Even for people like me who use LLMs in our workflow, we don't want them randomly shoved into every screen without our control. And we can see the transparent power play that companies like Google are making to try to control and own every aspect of our data and digital life, which we reject.
IMO, locally run algorithms to categorize and identify objects and faces in photos do not fall into the second category. It's not generative AI. It's a useful feature that lots of photo management software (including FOSS options) already has. It doesn't need to be enabled by default, but there's no harm in shipping it as an option