Princeton faculty voted this week to require instructor proctoring of all in-class exams starting July 1, ending a tradition that has stood since 1893. The reversal follows a 2025 senior survey in which 29.9% of students admitted to cheating on at least one assignment or exam, and 44.6% said they had witnessed cheating and chose not to report it. Administrators directly attribute the change to the spread of generative AI on personal devices.
The numbers skew sharply by discipline. Bachelor of science in engineering (BSE) students admitted to cheating 40.8% of the time, compared to 26.4% among bachelor of arts (BA) students. Only one faculty member voted against the new proctoring requirement.
Princeton's honor code, established in 1893, required professors to leave the room during exams and students to write "I pledge my honor that I have not violated the Honor Code during this examination" on every test. Students were also bound to report peers they saw cheating. The system held for 133 years on the assumption that social enforcement among classmates would do the work a proctor would not.
Key facts
- 01Princeton faculty voted to require instructor proctoring of all in-class exams from July 1, ending an honor code tradition dating to 1893.
- 02A 2025 senior survey found 29.9% of Princeton students admitted to cheating on at least one assignment or exam.
- 03Engineering (BSE) students admitted cheating at 40.8%, versus 26.4% for bachelor of arts students.
- 0444.6% of seniors said they witnessed cheating and chose not to report it, despite an honor-code duty to do so.
- 05Princeton administrators cite generative AI on personal devices as a leading driver of the shift.
Generative AI broke that assumption. A document circulated by the school cites a "perception that cheating on in-class exams has become widespread," and points to "the advent of generative artificial intelligence products which significantly lower the barrier to gaining unfair advantage in the context of an in-class examination." The administrators add that "the ease of access of these tools on a small personal device have also changed the external appearance of misconduct during an examination, which is much harder for other students to observe (and hence to report)."
“In a 2025 senior survey, 29.9% of Princeton students admitted to cheating on at least one assignment or exam, with engineering students reporting 40.8% — and 44.6% of all seniors said they witnessed cheating and never reported it.”— Jaeden Schafer
The reporting side of the honor code has collapsed in parallel. A January opinion piece in the Daily Princetonian described students who "turn a blind eye to cheating, or deliberately avoid sitting near the back row of a lecture hall to avoid catching their peers in the act." One student recounted a line out the door to the men's bathroom during an economics exam, a pattern the piece read as ubiquitous mid-test phone use.
Social media has compounded the silence. The administration's document notes that many reports now arrive anonymously because students fear "doxxing or shaming among their peer groups," which makes it harder for the Honor Committee and the Office of the Dean of Undergraduate Students (ODUS) to follow up even when a specific incident generates campus buzz.
The new proctoring rule is narrower than it sounds. Professors will sit in on exams from July 1, but they will not intervene to stop suspected cheating. Their role is to observe, take notes, and serve as "an additional witness in the room" who can later testify before the Honor Court. The Honor Committee structure stays; the change is that there is now an adult in the room with standing to corroborate a complaint.
Princeton's shift mirrors a broader retreat across higher education. Faculty in many institutions have moved away from take-home essays and toward in-class writing, oral exams, and handwritten work. The Princeton data suggests even that is not sufficient when every student carries a model in their pocket. Ars Technica's Scott Johnson, who teaches and grades student writing, recently wrote that AI use is "just making it extraordinarily difficult to do all the things that have been helping students learn for a very long time."
Students themselves do not appear to be confused about the trade. Johnson reports that he hasn't "encountered any students who think they're learning when they let LLMs do their work." In his framing, "it's just workload management to them" — a way to clear a credit requirement in a system where the cost of using a model is near zero and the cost of falling behind peers is high.
There is a counterweight worth naming. The Princeton numbers are self-reported on an anonymous senior survey, and the school has not released a breakdown of how much of the admitted cheating involved generative AI specifically versus older methods like crib sheets or peer copying. The 29.9% figure also covers "at least one assignment or exam" across an entire degree, which is a different claim than "30% of students cheat on any given test." The faculty vote is a response to perception and survey data, not to a forensic count of AI-assisted incidents.
The Princeton story is a leading indicator for how elite institutions will spend the next two years rewriting assessment. Honor codes built on the assumption that cheating is hard to do and easy to spot do not survive a world where a model on a phone can answer a multivariable calculus problem under the desk. Expect more universities to bring proctors back, more in-class handwritten exams, and a quiet acknowledgment that the take-home essay as a graded artifact is largely finished. The detail that lands hardest in the original Daily Princetonian piece on the proctoring vote is not the policy change itself — it is the banner ad running across the top of the article, for Google Gemini, reading "PRACTICED TO PREPARED."
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
