Santiago Schnell Resigns as Dartmouth Provost Over Undisclosed Claude Use

What changed

Dartmouth College provost Santiago Schnell resigned after an advisory committee found that he failed to disclose significant use of Claude in an academic paper and a Washington Post op-ed published August 9, 2026. The committee found evidence of manuscript revision, code fixes, and figure editing or generation, while also reporting that Schnell did not provide personal LLM logs or agree to an interview.

Why it matters

This case turns “AI assistance” from a vague label into a disclosure problem. Schnell described the tools as an editorial aid for refining arguments, improving clarity, and correcting grammar. But the committee’s access to Dartmouth’s enterprise account showed requests to revise papers, modify repositories, generate figures, and assess whether an article showed signs of AI writing.

That distinction matters wherever people build, publish, or approve technical work. Asking a model for a referee-style critique is different from asking it to rewrite sentences. Asking it to change code and figures moves further still. The boundary is not whether a chatbot wrote every word. It is whether the tool materially shaped work that readers, editors, or institutions were entitled to evaluate transparently.

The most damaging fact here is the repeated failure to disclose the use. The committee said Schnell’s public descriptions may understate what he actually did, although the full scope remains unclear. His resignation makes the practical lesson hard to miss: AI use can look routine inside a workflow and still become a serious professional issue when the surrounding rules require disclosure.