Correct. It's abusive to apply probabilistic AI models to real human individuals and penalize them for things they haven't even done yet with no recourse.

I hope you will remember this the next time your employer asks you to build an AI moderation, credit evaluation, or anti-fraud system that will harm much larger numbers of innocent people far than one mean website.