What the faculty review actually found
A Dartmouth faculty led committee determined that Provost Santiago Schnell used Claude and ChatGPT to substantially alter his own academic writing, including a journal manuscript and a Washington Post op-ed, without disclosing that use, a violation of at least one publisher's explicit AI use policy. President Sian Leah Beilock announced Schnell's resignation on October 28, characterizing it publicly as his own decision, though the timing followed student and faculty petitions to the Board of Trustees calling for his removal. History professor Matthew Delmont has stepped in as interim provost through June 2028, an unusually long interim appointment that suggests the university anticipates a deliberate, extended search rather than a quick replacement.
The specific detail that elevates this beyond a routine academic integrity story is how the investigation was actually conducted. The committee reviewed logs from Dartmouth's enterprise Claude account and Schnell's university email, the kind of audit trail that enterprise AI deployments are specifically built to provide. Schnell reportedly did not supply logs from personal large language model accounts when the committee requested them, a gap that means the committee's findings rest on what the enterprise account logs could show, not necessarily a complete picture of his AI use across every tool available to him.
Why this is a governance case study, not just a scandal
Every enterprise technology leader who has argued internally for mandating enterprise AI accounts over allowing employees to use personal ChatGPT or Claude subscriptions for work now has a concrete, high profile example of exactly why that policy matters. The enterprise account's audit trail is precisely what made this investigation possible at all, and the gap created by Schnell's personal account usage, which the committee could not fully audit, illustrates exactly the blind spot that persists whenever an organization tolerates AI use outside its own governed, logged infrastructure.
For CIOs building or refining AI usage policy, this case is a useful, concrete argument to bring to leadership conversations about enforcement: a policy requiring all AI assisted work to run through enterprise accounts earns its keep well beyond licensing costs or abstract data security, specifically by preserving the ability to investigate and substantiate a usage policy violation when one is alleged. A policy that exists on paper but is not enforced at the account level produces exactly the evidentiary gap Dartmouth's committee ran into here.
The disclosure standard this case actually tests
What makes this case more complicated than a simple plagiarism scandal is that AI assisted writing disclosure norms are still genuinely unsettled across academic publishing, with different journals and publications maintaining different, sometimes contradictory, policies on what level of AI assistance requires disclosure and in what form. Schnell's violation was specifically a failure to disclose AI use where at least one publisher's policy required it, which is a narrower and more specific finding than a general claim that using AI to alter academic writing is itself improper.
This distinction matters for any organization developing its own AI disclosure policy, whether in an academic, enterprise, or regulatory context: the policy needs to specify clearly what level and type of AI assistance requires disclosure, to whom, and in what form, rather than relying on a general expectation that people will disclose AI use whenever it feels substantial. Vague disclosure expectations produce exactly the kind of after the fact dispute that unfolded at Dartmouth, where the underlying question of whether disclosure was actually required became entangled with the more damaging question of why disclosure did not happen regardless.
What a C-suite level violation signals about policy enforcement
A provost is not a junior employee experimenting with a new tool without fully understanding institutional policy. Schnell held one of the most senior academic leadership positions at Dartmouth, with presumably full visibility into whatever AI use policy the university had established. That a violation at this level occurred, and specifically involved a failure to produce complete records when asked, suggests that AI use policy enforcement at the most senior levels of an organization deserves at least as much scrutiny as enforcement at the staff level, an assumption many organizations' current compliance programs do not actually reflect in practice.
Most enterprise AI governance programs focus audit and monitoring attention disproportionately on frontline and mid level employee usage, under an implicit assumption that senior leadership either does not use AI tools directly or can be trusted to self report accurately when they do. This case is a direct challenge to that assumption, and a reasonable prompt for any CIO to ask whether their own organization's AI usage monitoring and account enforcement actually applies evenly across the full organizational hierarchy, rather than concentrating scrutiny where it is administratively easiest to apply.
The incomplete investigation problem
Because the committee could not obtain logs from Schnell's personal AI accounts, the full scope of his AI use across his tenure as provost remains genuinely uncertain, regardless of what the enterprise account logs that were reviewed actually showed. This is an uncomfortable but important limitation to sit with: an investigation can produce a confident finding about the specific instances it could document while still leaving the broader question of total AI use scope unresolved, simply because personal account usage falls outside what any institutional audit can reach without the individual's voluntary cooperation.
For any organization building AI governance policy, this gap argues for building personal device and personal account usage restrictions into policy from the outset, rather than relying entirely on enterprise account monitoring and trusting that employees will not supplement that usage with personal tools for the specific interactions they would prefer not to have logged. A policy that only governs the enterprise account, without any mechanism addressing personal account usage for work related tasks, leaves exactly the gap this investigation ran into.
The decision this reader actually owns
If your organization's AI usage policy currently governs enterprise accounts only, without addressing personal account usage for work tasks, this case is a concrete, citable example to bring into your next policy review, specifically on the question of what happens when an investigation needs records your policy has no mechanism to compel.
More broadly, audit your own organization's AI usage monitoring for whether it applies as rigorously to senior leadership as it does to frontline staff, since Dartmouth's case suggests that assumption of even enforcement across hierarchy levels is worth testing directly rather than taking for granted in your own institution.
