Dartmouth provost resigns after review finds undisclosed AI use in his writing

Dartmouth Provost Santiago Schnell resigned on Wednesday, October 28, after a faculty-led committee concluded he used AI to meaningfully alter his writing in violation of at least one publication's disclosure requirement. President Sian Leah Beilock says leaving was his decision. Inside Higher Ed reported the resignation came less than two weeks before a faculty no-confidence vote. Matthew Delmont will serve as interim provost through June 2028.

The trail began in August, when Semafor reported that the detector Pangram flagged Schnell's Washington Post column, "Universities are fighting AI cheating. But there’s a deeper problem," as at least 80 percent AI generated. Semafor counted it among 50 guest columns with AI-generated text in the Times, the Journal and the Post over one month. The student paper The Dartmouth then ran Pangram on nine of his articles this year and reported 96 percent "AI-written."

Now the part worth reading twice: the committee's eight-page report did not use detectors, which experts say can be unreliable. It used publisher policies, correspondence and records from his Dartmouth Claude account. Those logs showed him using Claude to edit a ChemSystemsChem manuscript, debug code and generate figures after peer review, with no disclosure. They also show him asking the model whether another paper had "many signatures of AI writing," then submitting a revision with the kinds of changes Claude suggested. The committee says he did not hand over logs from his personal LLM accounts.

My take: the detector lit the fuse, but the logs made the case.