Dartmouth's provost warned us about AI cheating—but his own AI use cost him his job

Disclosure Day Dartmouth’s provost warned us about AI cheating—but his own AI use cost him his job People who love AI don’t always love telling others they used it.

1 Dartmouth's New Hampshire campus. Credit: Getty Images Dartmouth's New Hampshire campus. Credit: Getty Images Text settings Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only Learn more Minimize to nav On August 9, 2026, Dartmouth College provost Santiago Schnell wrote a Washington Post op-ed called “Universities are fighting AI cheating. But there’s a deeper problem.” This week, Schnell discovered the depths of the problem himself as he resigned from the university after an advisory committee found that he failed to disclose significant AI use in an academic paper and in the Washington Post op-ed, even though publishers required such disclosure.

Schnell didn’t make it especially easy for the committee to investigate his AI use; the report uses the term “insufficient cooperation.” He did turn over his Dartmouth laptop, but he did not turn over logs from his personal LLM accounts related to the writings in question. Nor did he agree to meet with the committee.

The committee did get access to the Claude enterprise account Schnell had used through Dartmouth, though. This shows him using the LLM as a tool for revising his papers, finding and fixing bugs in code, and editing or generating figures. For instance, Schnell would run his papers through Claude and ask things like:

I am getting read[y] to submit the enclosed paper, its supplementary materials and the enclosed repo to ChemSystemsChem. This paper is not biological, but rather chemical. Imagine that you are a reviewer for this paper with expertise in nonlinear chemical dynamics and mathematical chemical kinetics. What would be your referee report for this paper?

This is quite a bit different from asking Claude to write your entire paper for you, and Schnell has long insisted that he used AI as an editorial aid and discussion partner. He told Semafor that he had used chatbots “to refine a few arguments” and “check for grammatical errors.” He told one journal editor that he “used AI as an editorial tool” to “help refine sentences, improve clarity, and the expression of arguments.” He had not disclosed his AI use to an academic journal, he said, because he “did not understand that this type of AI-assisted editing required disclosure.”

Indeed, Schnell’s own description of what he did with AI doesn’t sound, on the long list of AI-powered sins, like it would rank near the top. But it’s hard to know. As the Dartmouth committee points out, the full scope of Schnell’s AI use was unclear to them. The committee believes that Schnell’s public statements may understate “the nature of the Provost’s actual AI use” and says that Schnell regularly “went beyond his other public descriptions of his use of AI in writing.”

We do know that Schnell also asked LLMs to take TeX files and “make the corrections to the files, and the repo. If you have to generate new figures, please provide them to me. Let me know what you change in the repo, and text files so that I can take a look into it.”

Whatever you think of his actions, Schnell repeatedly failed to disclose them. It is certainly possible that he thought his AI use was on par with a spelling or grammar checker, though one Claude exchange raises questions about whether Schnell wanted to hide traces of LLM activity in his work. From the committee report:

Provost Schnell’s Claude logs reflect that he uploaded to Claude a manuscript of a forthcoming article entitled Academic Freedom and Catholic Mission: The Procedural Overlap Zone in University Governance and prompted Claude: “Are there many signatures of AI writing in this article?” Schnell later submitted a revised manuscript that incorporated the kinds of changes Claude suggested.

Schnell has now resigned. Dartmouth, though, is barreling ahead with AI. The university calls itself the place that “launched the field of artificial intelligence” back in 1956, when “the term was coined and the foundational question was asked: Can machines solve problems once reserved for human minds?” Seventy years later, Dartmouth is hosting a conference called “What Must Remain Human?” that will feature a student hackathon sponsored by Anthropic and speakers like Mira Murati and Yann LeCun.

Despite the scandal, Dartmouth President Sian Leah Beilock wants more AI. (She wrote a piece for The Atlantic in September called “Why I Want More AI at Dartmouth.”) What’s crucial, though, is that the school’s leaders should “hold themselves to the highest standards,” she said. This appears to mean one main thing: disclosure.

In announcing Schnell’s resignation, Beilock wrote that “each of us brings a different perspective on how and when to incorporate AI into our work and our lives. Moreover, keeping Dartmouth at the forefront of AI inevitably means leading through an environment that is fluid and fast-changing. That is why, as leaders, it is so critical that we model the transparency and human judgment that we expect from our students and scholars.”

What does “disclosure” mean? In a September 2026 paper on AI detection tools, called “Who Wrote This?”, Schnell himself goes to great lengths to argue that “human-written” or “AI-written” is far too simple a binary. In his view, everything will become a mix of human judgment and AI work, and there’s nothing wrong with that. The key for him is that the human author takes full responsibility for the text, however it was composed and edited.

In the paper, Schnell lays out his method for producing it. It’s worth quoting at some length:

In September 2026, I asked OpenAI’s GPT-5.6 Pro whether an AI-written essay could ever be identified with certainty, and I suggested connecting the problem to Turing’s imitation game and to the “reverse” Turing test posed by AI-text detection. GPT-5.6 Pro connected the question to binary hypothesis testing and statistical decision theory, proposed early structures and titles, identified candidate literature, and generated portions of early drafts. I chose the problem, journal, audience, and eventual direction of the argument.

I decided which examples and results should remain central, including the decision to retain the Liang study while confronting it with important counterevidence. I checked the cited literature, rejected or corrected claims that I judged inaccurate, corrected mathematical and conceptual errors, and reorganized the argument.

In revising the manuscript, I rewrote most sections and added substantial explanations. I occasionally asked GPT-5.6 Sol to provide language-editing assistance for sections I had rewritten. I created the figures and made the final decisions about the mathematics, interpretation, sources, organization, and wording. I take full responsibility for the resulting manuscript.

Before submitting the manuscript for publication, I also asked GPT-5.6 Pro and Anthropic Claude Fable 5.1 to review it critically and identify issues that reviewers might raise. I evaluated the issues they identified and made corrections where I judged them warranted.

Despite all this assistance, Schnell resists the idea that AI is a “co-author.” Instead, he writes, “Calling [the paper] ‘coauthored’ would introduce a different confusion. Mathematical tools have long made substantive contributions without thereby becoming authors… The point is not that the software contributed nothing. It is that contribution and authorship are different roles.”

He is not against disclosure, as he makes clear by describing his method. Yet curiously, when submitting other papers for publication, he did not disclose his AI use even when required by publishers.

Confused about the rules in an age where even universities are pushing AI? So are many students. New survey data unveiled at this weekend’s conference shows that 61 percent of college students think AI helps explain difficult concepts, and half believe that they produce better work after talking through issues with AI tools. But “70 percent of the college students in the survey indicate that take-home work has become unfair due to AI, with 39 percent feeling pressured to use it because their peers do.” Among Dartmouth students, 92 percent of them fear “that AI can hurt writing skills.” And many of them believe their peers are using far more AI than they actually are.

Not everyone is racked with existential doubt over how to deal with AI, though. Dartmouth student Eli Moyse wrote in a September op-ed that Schnell should be fired because “a fetishization of AI to earn publicity for the College has rotted the brain of our top academic.”