I've noted before that because AI detectors produce false positives, it's unethical to use them to detect cheating.

Now there's a new study that shows it's even worse. Not only do AI detectors falsely flag human-written text as AI-written, the way in