And apparently it is so entrenched now that even the review sites/researchers/etc. don’t bother doing research into what data Apple harvests anymore. They just believe them implicitly. There are services on all their platforms like mediaanalysisd that scan files and categorize into categories like “human, pet, good photo, bad photo, drivers license, badge, social security number” and they even have the capability to analyze video and “auto-caption” the video, they claim it’s an accessibility feature. It is also a great metadata generation and harvesting tool for all media on the device, and trivial for the contents of that metadata to be spirited away to Apple or elsewhere.

Classification excerpt from that process in particular:

meme_document_check_or_checkbook
meme_curation_meme
meme_curation_screenshot
meme_document_boarding_pass
meme_document_currency_or_bill
meme_document_driving_license
meme_document_office_badge
meme_document_passport
meme_document_receipt
meme_document_social_security_number