Bad actors tend to be some of the first to use new technology (because new and nobody knows what they're looking for yet) so that's not surprising, and this problem has been going on in science long before AI became a buzzword.
I'm more interested when the gap starts to close and we get to use AI to start reproducing studies to actually verify results.
Erm, honestly. We already have reproducility issues without AI. I mean, scientific retractions are at an all time high and thats publications not necessarily just from the time when AI got more accessible.
Retractions are good actually because the scientific record is being corrected. It's stuff that doesn't get retracted or slips under radar which is the issue. Journals resist retractions cos it makes them look bad.