Dartmouth College provost Santiago Schnell resigned after an advisory committee found that he failed to disclose significant AI use in an academic paper and a Washington Post op-ed. The episode raises a practical question for researchers and university leaders: when AI helps shape written work, what should be disclosed, and to whom?
AI assistance met a disclosure requirement
On August 9, 2026, Schnell published a Washington Post op-ed called “Universities are fighting AI cheating. But there’s a deeper problem.” The advisory committee later found that he had not disclosed significant AI use in that piece or in an academic paper, despite publishers requiring disclosure.
Schnell described his use as editorial support. He told Semafor that chatbots helped him “to refine a few arguments” and “check for grammatical errors.” He also told a journal editor he had used AI “to help refine sentences, improve clarity, and the expression of arguments.” He said he did not understand that this kind of editing required disclosure.
That explanation matters because editing assistance is different from asking an AI system to write an entire paper. But the committee’s findings suggest the distinction did not settle the question. Its report said the full scope of Schnell’s AI use was unclear and that he regularly “went beyond his other public descriptions of his use of AI in writing.”
Records showed a wider range of uses
The committee reviewed a Claude enterprise account Schnell had used through Dartmouth. Its records showed him asking the system to review papers, find and fix code bugs, and edit or generate figures. In one exchange, he asked Claude to act as a reviewer with expertise in nonlinear chemical dynamics and mathematical chemical kinetics and provide a referee report on a paper he planned to submit to ChemSystemsChem.
Other requests went beyond sentence-level edits. Schnell asked LLMs to make corrections to TeX files and a repository, and to provide new figures if needed. He asked the system to tell him what it changed so he could review the files.
The committee also cited an exchange about a forthcoming article, “Academic Freedom and Catholic Mission: The Procedural Overlap Zone in University Governance.” Schnell uploaded the manuscript and asked Claude: “Are there many signatures of AI writing in this article?” He later submitted a revised version that incorporated the kinds of changes Claude suggested.
The committee described Schnell’s cooperation with its inquiry as “insufficient cooperation.” He turned over his Dartmouth laptop, but not logs from personal LLM accounts related to the writing under review, and he did not agree to meet with the committee. The report said those limits left the full extent of his AI use uncertain.
Disclosure is about the process, too
The case is not simply a dispute over whether AI can help with writing. Schnell’s later account of his own method shows how many different roles the tools can play: suggesting structures, identifying literature, drafting portions, editing language, and critiquing a manuscript before submission.
In a September 2026 paper on AI detection tools called “Who Wrote This?”, Schnell argued that “human-written” and “AI-written” are too simple as opposing categories. He described using OpenAI’s GPT-5.6 Pro to explore the question, propose structures and titles, identify candidate literature, and generate portions of early drafts. He said he selected the topic and argument, checked cited literature, corrected errors, reorganized the text, and took responsibility for the final manuscript.
He also described using GPT-5.6 Sol for language editing and GPT-5.6 Pro and Anthropic Claude Fable 5.1 for critical review before publication. Schnell rejected calling AI a co-author, writing: “The point is not that the software contributed nothing. It is that contribution and authorship are different roles.” That distinction does not remove the need to follow a publisher’s rules about disclosure.
Dartmouth’s AI ambitions continue
Schnell’s resignation comes as Dartmouth continues to promote its place in AI. The university says it “launched the field of artificial intelligence” in 1956. It is hosting a conference called “What Must Remain Human?” with a student hackathon sponsored by Anthropic and speakers including Mira Murati and Yann LeCun.
Dartmouth President Sian Leah Beilock has advocated for more AI at the university, including in a September piece for The Atlantic called “Why I Want More AI at Dartmouth.” In announcing Schnell’s resignation, she said leaders should model “the transparency and human judgment that we expect from our students and scholars.”
For researchers, the lesson is concrete: AI assistance can take many forms, and describing it as editing may not capture the whole process. When publishers require disclosure, authors need to make their use clear, while retaining responsibility for the work they submit.