One kind of AI-enabled plagiarism that can be very difficult to detect is done by students who are writing in a second language. They copy-and-paste text from Wikipedia (or whatever) in their first language into DeepL or Google Translate and submit the output to their teachers as if it were their original work.

Teachers sometimes become suspicious because the students’ submitted texts contain many fewer grammatical errors and nonidiomatic expressions than the students’ speech or in-class writing. But it is hard to prove plagiarism if the teachers cannot do web searches in the students’ first language.

I did hear once, though, of a cut-and-paste MT translation being caught by a teacher because it included some large unique numbers, which did appear in a Google search.

A related issue that writing teachers are struggling to come to terms with is whether second-language writers are cheating when they use MT to translate texts that they originally wrote in their first language. Some teachers, especially in language-focused classes, believe that the use of MT hurts the students’ language acquisition and thus defeats one purpose of the writing assignments; they also think it is unfair to other students who struggle to write in the second language without the help of MT. Other teachers believe that writers should be allowed to use whatever tools are available to them as long as their content is original.