Key points
- An advisory committee found insufficient cooperation and undisclosed AI assistance in published work.
- Logs showed the provost used Claude for editing, code debugging and figure generation.
- Schnell claimed he viewed the technology as an editorial aid that did not require disclosure.
Santiago Schnell, the provost of Dartmouth College, resigned this week following an investigation into his use of artificial intelligence. An advisory committee determined that he failed to disclose significant AI assistance in both an academic paper and a Washington Post op-ed, despite publisher requirements for such transparency.
The evidence
Ars Technica reported that the committee accessed Schnell’s enterprise Claude account via Dartmouth.reported The logs revealed he used the large language model to revise papers, find bugs in code and edit or generate figures. For example, he asked the model to simulate a reviewer for a chemical dynamics paper to provide a referee report.
Evidence: reported
Dartmouth's provost warned us about AI cheating—but his own AI use cost him his jobaifleet-ars
The decision
The committee cited “insufficient cooperation” in its findings because Schnell did not provide logs from his personal large language model accounts. He also declined to meet with the investigators, although he did hand over his university laptop. The full scope of his AI use remains unclear due to these missing records.
The takeaway
Schnell maintained that he used chatbots merely to refine arguments and check for grammatical errors, not to write content from scratch. He told one editor he used AI as an editorial tool to improve clarity and expression. He stated he did not disclose the assistance because he did not understand that this type of editing required it.
Building on this
Developers building enterprise AI tools for academic or professional writing must account for these disclosure ambiguities. The documented capabilities include high-level editorial refinement, code debugging assistance and simulated peer review. Interfaces that allow users to toggle between drafting and editing modes could help institutions track usage for compliance. Limits remain in distinguishing between acceptable editorial aid and prohibited content generation, a boundary Schnell apparently misunderstood. Systems should log specific prompt types, such as “refine sentences” versus “write section,” to provide audit trails. The lack of personal account logs in this case highlights a gap in monitoring; enterprise tools must enforce data retention for compliance purposes. Builders should design workflows that explicitly categorize AI involvement levels, ensuring users can generate the necessary disclosure statements automatically. This reduces the risk of non-compliance by making the extent of AI use transparent by default.
Why it matters
- Enterprise AI deployments in regulated sectors require clear definitions of acceptable use, as editorial assistance may still mandate disclosure.
- Monitoring gaps in personal accounts versus enterprise tools create compliance risks that organisations must address through policy or technology controls.
- The distinction between drafting and editing is technically ambiguous, necessitating better logging and classification features in AI interfaces for auditability.
- Leadership understanding of disclosure obligations is critical, as misunderstandings can lead to severe professional consequences and institutional reputational damage.
Topics: academic integrity, AI compliance, enterprise AI, large language models, disclosure policies
Updated
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
What specific tasks did the provost use AI for?
He used it to revise papers, find bugs in code, edit figures and simulate peer review feedback for academic submissions.
Why did the committee cite insufficient cooperation?
Schnell did not provide logs from his personal AI accounts and refused to meet with the investigating committee members.
Did Schnell admit to using AI to write his papers?
No, he insisted he used it only as an editorial aid to refine arguments and correct grammar, not to generate full content.
Connected stories
Reading receipt
- An advisory committee found insufficient cooperation and undisclosed AI assistance in published work.
- Logs showed the provost used Claude for editing, code debugging and figure generation.
- Schnell claimed he viewed the technology as an editorial aid that did not require disclosure.