I wonder if lack of consequences might be related to lack of proof. Cheating in the past looked like reusing an essay you found online or paying off someone to write it for you -- methods that offer a definitive way to prove it happened.

AI isn't particularly provable. Worse, a lot of professors are lazy and will rely on tools that tell them something was produced with AI; and like any tool, they'll produce false positives. Just imagine being expelled for writing because you don't make spelling mistakes, and do use em-dashes, bullet points and typographic emphasis.