I guess you missed the part that above a threshold number of hits, a human being would, in fact, review the putative offending photographs. There were many other valid concerns as well, for quite apart from anything a system for modifying images slightly to defeat the CSAM scanner has already been published. Apple's idea was well intentioned, but poorly thought out. I think they were trying to find a use for their newly optimised AI chips. I wish they would turn that AI processing power to defeat spam, phishing, and other security and privacy threats.