Quoting: LoudTechieMy proposals were.
Publish cryptographic hashes of restricted material, so small player can collaborate to implement filters for illegal information like CSAM and copyright protected material allowing them to more cheaply preform moderation responsibilities.

Perceptual hashing [External Link], used for copyrighted content and CSAM detection is very different from cryptographic hashing. The goal is to catch images and videos that are "sufficiently similar". It's as vague as it sounds, and unless configured with a very low sensitivity is guaranteed to cause false positives. When configured with low sensitivity, it's possible to bypass. When you take into account the quantity of CSAM perceptual hashes out there, false positives happen regularly. So you can either block everything, or block random things. Far from a solved problem

Using perceptual hashing for copyright enforcement is even worse, because the algorithm has no way to account for exceptions to copyright [External Link]. That should always be a human's judgement call, but with copyright trolls in the picture you get a lot of pressure towards false positives.