These are not the kinds of hashes used for CSAM detection, though, because that would only work for the exact pixel-by-pixel copy - any resizing, compression etc would drastically change the hash.
Instead, systems like these use perceptual hashing, in which similar inputs produce similar hashes, so that one can test for likeness. Those have much higher collision rates, and are also much easier to deliberately generate collisions for.
Instead, systems like these use perceptual hashing, in which similar inputs produce similar hashes, so that one can test for likeness. Those have much higher collision rates, and are also much easier to deliberately generate collisions for.