Update. Of course teachers sometimes make false accusations of #plagiarism even without relying on imperfect tools. Now that they're on the lookout for #AI-generated submissions, the rate might increase.
New study: "When given a mixture of original and general abstracts, blinded human reviewers correctly identified 68% of generated abstracts as…generated by ChatGPT, but incorrectly identified 14% of original abstracts as being generated." https://www.nature.com/articles/s41746-023-00819-6