We only get to see the system's failures, and only the scandalous or funny at that. The system probably flags tens of thousands of uploads every day. Note how,

We only get to see the system's failures, and only the scandalous or funny at that. The system probably flags tens of thousands of uploads every day. Note how, in this case, it probably did correctly identify segments of a certain length and entropy. At that point, it needs to loop in a human to decide if the audio can be copyrighted. It does so, with the uploader being the first human it asks. Yes, that last part may be annoying, and slightly cheap. But the uploader is also in a better position to judge the matter than some content moderator, so I don't know why it's supposedly scandalous.