I would think that they can classify most posts as safe with very high accuracy.

Certainly it’s not unreasonable to have them disprove that

And the scope could be further reduced by looking at posts with more then a few views.