I'm interested in how Google does this on a technical level. Anyone have insight? Do they have a database of "known hateful files" and compare the md5s or som

I'm interested in how Google does this on a technical level. Anyone have insight? Do they have a database of "known hateful files" and compare the md5s or something like that? do they scan text and videos for word combinations and sentences known to be hate speech?