Dartmouth wrote an AI rule for students. Now it applies to the provost

Three universities have caught faculty or senior leaders using AI in ways their own student rules would punish. Northeastern defended its AI posture and kept the tuition. The University of Hong Kong took away a leadership title and disciplined the student. Dartmouth, whose provost is now under review, has not decided, and its decision will set the evidence standard for every Dartmouth student accused next.

I ran the Dartmouth situation against the CrisisCommand® Knowledge Base. Northeastern and Hong Kong came back as the closest precedents. They made opposite choices, and Dartmouth's committee has both on the record.

Three schools, three decisions

Northeastern decided to defend the institution. In February 2025, a business student found ChatGPT artifacts in her course materials: leftover prompt text and distorted images. The syllabus barred students from unauthorized AI use. She filed a formal complaint and asked for more than $8,000 back. The professor told The New York Times he had run class files through ChatGPT "to refine them." The university said it "embraces the use of artificial intelligence to enhance all aspects of its teaching, research, and operations" and denied the refund the day after she graduated. The cost was a national news cycle about a rule that bound students and not their professor.

The University of Hong Kong decided to act at both levels. A fertility study co-authored by an associate dean and published in China Population and Development Studies contained fictitious citations that a doctoral student generated with AI and never declared. On December 17, 2025, HKU confirmed the fabrications. The senior author stepped down as associate dean, left his research committee roles, and applied to retract the paper. The student faces discipline, and HKU announced mandatory AI training and assessment for all researchers. The cost was a leadership title and a published paper.

Dartmouth has not decided. On September 21, the student newspaper The Dartmouth reported that nine of the provost's 2026 papers and op-eds scored a median 96% "AI-written" on the Pangram detector. His five papers from before late 2022 scored 100% human. One flagged piece was an August Washington Post op-ed about AI cheating. His emailed reply to the student reporters scored 100% AI. The provost has written that he should always disclose AI use in public writing, and he has argued that detectors are biased against non-native English speakers like him. On September 25 the president announced an independent committee review with no timeline. The campus AAUP chapter is circulating a petition for his removal.

The committee is writing student policy

Dartmouth tells students that failing to acknowledge AI use is grounds for discipline. The fall 2026 faculty handbook does not mention AI. With no faculty rule to apply, the committee will reach for the only written AI standard the college has, and whatever evidence it accepts becomes the evidence every student defense cites.

If a detector score is enough to find against the provost, it is enough to find against a sophomore. If "I used it for language and clarity" clears him, every accused student will offer the same sentence, and the college will already have accepted it. Any campus with a student disclosure rule and a silent faculty handbook is one student newspaper investigation away from the same position.

How the gap opens

The sequence is repeatable. A student reporter runs a senior leader's published writing through a detector. The score publishes before the institution has a statement. The leader explains, and the explanation gets tested too. Then the institution discovers that its only written AI rule governs students and its only process for AI disputes is student conduct.

Detector evidence is already contested inside that process. In February 2026 a court ordered Adelphi University to strike an AI-plagiarism finding that rested on a Turnitin flag, calling the finding "without valid basis and devoid of reason." Whatever standard Dartmouth sets for its provost will be tested against that ruling the first time a student appeals.

What works

One disclosure rule for everyone who publishes under the institution's name. Students, faculty, and administrators follow the same text.

Evidence rules written before the case arrives. A detector score opens a review. A finding requires more: drafts, version history, or an interview.

Disclosure at the moment of publication. A one-line note on the op-ed or paper settles the question before a reporter raises it.

Reviewers outside the reporting line, named publicly. The review of a senior leader carries weight only when the campus sees who is conducting it and why they are independent.

What does not work

Defending the institution's AI strategy when the question is one person's conduct. Northeastern answered a complaint about one course with a statement about the university's AI ambitions, and the double-standard story ran anyway.

Disclosing after the scores publish. Critics at Dartmouth describe the provost's disclosures as piecemeal and after the fact, and that timing drives the hypocrisy charge.

Letting the detector decide alone. The Adelphi ruling shows what a detector-only finding costs when a student takes it to court.

Answering an AI authorship question with AI-drafted text. Dartmouth's reply to the student reporters became the next headline.

The decision in front of Dartmouth

Northeastern defended its posture and got a national story about a double standard. Hong Kong gave up a deanship and a paper, and wrote a rule that covers every researcher. Dartmouth still has the choice. Write the faculty rule, decide what evidence counts, and apply both to the provost and the sophomore in the same terms.

---

Built from public reporting as of September 28, 2026. CrisisCommand is not affiliated with, and does not represent, any institution named here. No one at these institutions requested or reviewed this analysis. This is decision support, not legal advice; engage counsel on specifics.