Dartmouth Provost Resigns Over Undisclosed AI Use Caught by Claude Logs
Dartmouth's provost resigned after enterprise Claude logs proved he secretly used AI to write and edit academic papers.
Drama · Source: Inside Higher Ed
What happened
Dartmouth Provost Santiago Schnell resigned on October 28. A faculty committee concluded he used artificial intelligence to meaningfully alter his writing without telling anyone. He stepped down less than two weeks before faculty leaders planned a no-confidence vote. President Sian Leah Beilock stated the resignation was his decision. History professor Matthew Delmont will serve as interim provost until June 2028.
The scandal began in August. Semafor reported that an AI detector called Pangram flagged his Washington Post editorial as at least eighty percent AI-generated. A student newspaper found similar results across nine professional articles. Students and faculty demanded his firing. Beilock convened an independent advisory committee to investigate the claims.
The committee made a crucial choice. They completely ignored the AI detector scores. Instead, they pulled his emails and logs from Dartmouth's enterprise Claude account. The logs caught him. He used Claude to substantially edit research, fix computer code, and generate figures. He asked Claude to review a peer response he had already written using ChatGPT. He even uploaded a manuscript and asked Claude if it contained signatures of AI writing. He failed to disclose any of this to publishers. When confronted, Schnell claimed he relied on older guidelines. The committee proved the publisher required AI disclosure since at least 2023. He also refused to hand over logs from his personal AI accounts.
Key facts
- 80 percent — AI-generated text flagged in his Washington Post column by Semafor
- 30 — Articles published by Schnell since joining Dartmouth in July 2025
- June 2028 — End date for Matthew Delmont's term as interim provost
Why it matters
This changes how organizations enforce AI policies. AI detectors are officially too unreliable for high-stakes human resources decisions. Enterprise logs are the new standard of proof. When you use a corporate AI account, your employer sees every prompt. Schnell lost his job because his chat history proved his intent to hide the AI assistance. Institutions will stop trying to ban AI and start auditing it instead.
The second-order effect is a massive shift toward mandatory disclosure. Dartmouth actually encourages AI use. The president explicitly called for more AI on campus. The crime was the cover-up. Every platform, publisher, and university will now write strict disclosure rules. We are moving from a culture of avoidance to a culture of strict accounting. The burden of proof shifts from the reader guessing if text is AI, to the writer proving how they built it. If you build with AI, you must declare it. Hiding your AI use is a fireable offense.
For builders
Enterprise logs are the ultimate audit trail
If you build B2B AI tools, robust admin logging is a mandatory feature. Companies will pay for oversight. If you are an employee, assume your boss reads every prompt you type.
AI detectors are dead for enforcement
The committee explicitly refused to use AI detection software. Stop building compliance tools based on guessing text origins. Build tools for provenance and verifiable document history instead.
Disclosure is a massive compliance market
Publishers and universities require AI transparency. Startups that automate AI disclosure tracking will win big. Institutions will pay heavy premiums to avoid this exact public relations disaster.
My take
I always say using AI to write is just smart leverage. But you cannot preach academic integrity while secretly asking Claude to hide your AI footprints. Lying about it is professional suicide. Disclose your tools, own your output, and stop treating AI like a dirty secret.
Original reporting: Inside Higher Ed. This is my rewrite and opinion.