If “hate speech” has a computable signature, then not-“hate speech” also has a computable signature.

Therefore, we can expect all kinds of bad actors to take advantage of that.

Sooner or later, Internet outrage mobs will form around vulnerable people whose speech wasn't not-“hate speech” enough.

I'm sure that plenty of people, around here, can dream up some more nightmare scenarios.