Dartmouth provost Santiago Schnell has resigned and is leaving the university after a faculty-led review found he used AI to alter his writing without the disclosure that a scientific journal and The Washington Post required. President Sian Leah Beilock announced the resignation in a letter on 28 October, and published the committee’s report alongside it.
The case turns less on whether a senior academic used AI tools, which Dartmouth itself provides through an enterprise Claude account, than on whether he told the outlets that published his work. It is one of the clearest public tests so far of how disclosure rules for AI apply to people at the top of an organisation.
In brief
- A five-member committee found the provost used AI to change published work without the disclosure two outlets required, and was “not forthcoming” about his use.
- It cleared him on a third paper, did not rely on AI detectors and made no disciplinary recommendation.
- Schnell resigned; history professor Matthew Delmont becomes interim provost until June 2028.
What the review found about the Dartmouth provost’s AI use
Beilock set up the committee on 25 September, asking in a letter that day for “a rigorous, objective review of how Santiago has used AI, measured against the criteria of the publications to which he submitted his work.” It was chaired by David Kotz, a computer science professor and former Dartmouth provost.
Its report, dated 27 October, concludes that the provost “used AI to meaningfully alter his writing in violation of at least one publication’s AI use and/or disclosure requirement, failed to disclose his use of AI in clear violation of a national newspaper’s stated policy, was not forthcoming about his AI use, and fell short of the standards set forth by Dartmouth’s Code of Ethical Conduct, the Academic Honor Principle, and Geisel’s Policy on Expectations for Professionalism”.
The first case was a chemistry paper in ChemSystemsChem, published in September. According to the report, logs of Dartmouth’s enterprise Claude account show he used it “to substantially edit his research manuscript, troubleshoot errors in computer code, and generate figures” after peer review. One prompt asked Claude to imagine being a reviewer and write a referee report. The paper carried no disclosure. Schnell told Beilock he had relied on older author instructions and said: “This wasn’t the case, but a disclosure would have been the best approach.” The committee found the journal’s owner, Wiley, had required AI disclosure “since at least 2023”. It also notes that when Schnell later wrote to the journal, he described using ChatGPT but did not mention Claude.
The second case was an August op-ed in The Washington Post. Schnell had told Semafor he used ChatGPT “to refine a few arguments, check for grammatical errors, and submit it to the Washington Post.” The committee says the Post’s terms required contributors to agree their content “was not created or manipulated with artificial intelligence or editing software”, and calls his conduct a “clear violation”.
On a third paper, in the Bulletin of Mathematical Biology, he used Claude and ChatGPT to review code. The committee “could not conclude there was a violation” and “found no evidence that the Provost used AI to edit or manipulate the text of the paper itself.”
On honesty, the report says Schnell asked Claude “Are there many signatures of AI writing in this article?” about a forthcoming article, then told its editor he “did not understand that this type of AI-assisted editing required disclosure”. The committee says “this statement understates the nature of the Provost’s actual AI use according to the Claude logs.”
The limits of the review and Schnell’s side
The committee stresses that its findings are limited. It reviewed a targeted set of his writings rather than all 30 or more published since July 2025, and it did not use the AI detector Pangram “or other AI tools” to judge his use. It says it was “not proceeding under the procedures for assessing complaints under these policies nor recommending any related disciplinary action.”
It also says it had “insufficient cooperation from the Provost”. He gave the committee his Dartmouth laptop, but according to the report he did not supply logs from personal AI accounts or meet at the time requested.
The former Dartmouth provost has acknowledged using AI as an editing aid. He told Semafor in August that he wrote the op-ed “based on my own thinking” and that AI detectors are “biased against non-native English speakers, such as myself.” He declined to comment to The Dartmouth, and did not respond to NHPR.
Beilock wrote that he resigned “out of respect for his colleagues and the institution” and that “This was his decision. We remain grateful for his leadership and wish him well.” Matthew Delmont, a history professor and associate dean, “will serve as interim provost through June 2028”.
Why AI disclosure rules at work are still a patchwork
The report shows how uneven AI disclosure rules are, even for one person. Wiley required disclosure. The Post asked contributors to confirm their work was not created or manipulated with AI. Springer’s journal changed its policy between submission and publication. The journal for the forthcoming article had no specific AI policy at all. One author, working in one year, faced four different standards, and the committee had to check archived versions of the rules to decide which applied.
That is the problem Beilock pointed to in September, when she wrote that “we need role-appropriate rules and policies for this technology.” The finding against the Dartmouth provost is less about using a tool than about the gap between what he did and what he told people. Using Claude through a university’s own enterprise account left logs, and those logs, not detector scores, carried the case.
For any organisation rolling out AI assistants, that is the practical lesson. Disclosure duties already exist in contracts, terms of submission and codes of conduct, often written before generative AI and rarely read again. Lawmakers are adding more, from Pennsylvania’s AI advert labels to New York’s ban on chatbots posing as human with children. Questions over whether a president’s Grok prompts count as records show the same issue in government. Schools are testing AI tutors too. The Dartmouth case suggests the people setting AI policy will be held to it first.
Dartmouth, which is marking 70 years since a summer workshop on its campus launched the field of AI, now has to settle those shared AI norms without the provost who was meant to lead them.
Sources
Primary:
- Dartmouth, Office of the President, “An Update on the Provost Review”, Sian Leah Beilock, 28 October 2026
- Dartmouth Advisory Committee, “Review of Concerns Regarding AI Use by Provost Santiago Schnell”, report to the President, 27 October 2026 (PDF)
- Dartmouth, Office of the President, “A Review and a Community Conversation on AI”, Sian Leah Beilock, 25 September 2026
Press:
- The Dartmouth, “Santiago Schnell resigns as provost of the College”, 28 October 2026
- Semafor, “AI writing has already begun to appear on the opinion pages”, 26 August 2026
- NHPR, “Dartmouth provost resigns after review finds undisclosed AI use”, 29 October 2026
- Inside Higher Ed, “Dartmouth Provost Resigns Over AI Use”, 28 October 2026
Related stories
- Pennsylvania Signs AI Ad Disclosure Law With Fines Up to $3,000
- New York Law Bars AI Chatbots From Posing as Human With Children
- House Democrats Set 11 November Deadline Over Trump’s Grok Prompts
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X
