They basically do copyright enforcement in two cases:
1. Automatically, based almost entirely on sound, using ContentID - this is heavily weaponized, gamed, and generally over the top.
2. "Manually" (but really automatically) in response to requests from "verified" (sure) rights owners - this one is also heavily gamed, but seems to make up a much smaller fraction of takedowns.
1. Automatically, based almost entirely on sound, using ContentID - this is heavily weaponized, gamed, and generally over the top.
2. "Manually" (but really automatically) in response to requests from "verified" (sure) rights owners - this one is also heavily gamed, but seems to make up a much smaller fraction of takedowns.