Commentary: AI Destroys 133 Years of Princeton University Tradition — When "Cheating" Gradually Becomes Common Sense

Princeton University voted in 2025 to restore proctored exams, ending a 133-year-old tradition. When AI makes cheating the norm, does it simultaneously destroy the traditional notion that "knowledge requires assessment"? (Recap: Princeton's 133-year unproctored tradition ended by AI, nearly 30% of students admit to cheating) (Background: Ministry of Education's "AI in Every Library": library card holders get free access to ChatGPT and Claude, rolling out across 47 national universities in Taiwan) - Princeton faculty voted in 2025 to restore proctoring, ending the unproctored Honor Code tradition that had stood since 1893 for 133 years - A school newspaper survey found 30% of graduates admitted to cheating, 45% knew but didn't report it, and take-home exams have dropped by more than two-thirds over the past year - Professors are switching to oral exams, real-time Google Docs tracking, and blue book written tests, but once trust flips to "presumption of suspicion," there's no going back Final exams begin, the professor hands out the papers, and then walks out of the classroom. No proctors, no surveillance cameras, no phone restrictions. All that remains in Princeton University's classrooms are students, blue books, and a single signature line: "I have neither given nor received any assistance." This is what exams have looked like at Princeton University for the past 133 years. Then in 2025, faculty voted to bring teachers back into the classroom to proctor. The tradition lasted only two and a half years after ChatGPT's commercial debut. In 1876, Princeton's student newspaper The Princetonian published an editorial that read: "Assume students are dishonest, and they will become so; treat them as people of honor, and they will learn to behave accordingly." This editorial gave rise to Princeton's Honor Code, formally adopted in 1893. The content was simple: no proctoring of students, no surveillance, and student misconduct would be tried by a jury of peers. F. Scott Fitzgerald, while studying at Princeton, said that violating this code "simply wouldn't occur to you, just as you wouldn't think of rifling through your roommate's wallet." This continued for 133 years, until AI arrived in every student's phone. The numbers are brutal. In the 2024-25 academic year, 82 Princeton students were found to have engaged in academic "dishonesty." Three years ago, that number was 50. The school newspaper's anonymous survey of the graduating class received 501 responses, and the numbers were even uglier: 30% of students admitted to cheating, and 28% had used ChatGPT without permission. But the truly devastating number is the 45% of students who knew classmates were cheating and chose not to report it. In the Honor Code's design, peer reporting is the last line of defense, but nearly half of students no longer feel it's necessary. Morality is like a contagious disease. Princeton students openly discuss on the anonymous community app "Fizz" who used AI and who diligently wrote their entire report. Those abiding by the Honor Code begin to ask themselves, "Why am I not using AI?" When "not cheating" goes from virtue to disadvantage, trust begins to undergo a bank run. Princeton's Economics Department will next year require students to orally defend their research papers. AI can write the text, but at least for now, it can't stand on stage for you and answer or pose questions. History professor David Bell replaced take-home exams with in-class blue book written tests, and moved short essay assignments into Google Docs, so he can watch the entire typing process — including deletions, edits, pauses, and those perfect paragraphs that suddenly materialize out of thin air. Professor David Bell himself has noticed the chain reaction caused by reversed trust: "Students can sense it. I changed how I assign work to block AI, and they know that means I no longer trust them." The original system ran for 133 years with "trusting students" as the default. Now it's been forced to change to "suspecting you of cheating" — in essence, the Honor Code has already been broken. The university restoring proctored exams is only a small part of how AI is changing human values. What we actually want to discuss is: what are exams for in the first place? What does a diploma actually prove? If exams are meant to prove "a human can demonstrate what they've learned independently, without any other assistance," then AI is undoubtedly the most direct contemporary threat. Exam formats must be thoroughly cheat-proof, beginning to use oral exams, in-class pen-and-paper writing, and recording students' thought processes, replacing all assignments that can be completed by an AI model in thirty seconds. Princeton's professors are already doing this. But if exams and diplomas are meant to prove "whether a human can get things done well," the conclusion flips completely. A student who can write an excellent report using AI, and a student who writes the same report without AI — if their output is indistinguishable, what is actually being tested? The ability to resist temptation? Or the ability to deploy tools? No one in this world cares whether you typed every word yourself. What people care about is whether what you produce is usable. (Or whether it's compelling.) If exams are meant to prove "human integrity," then let's just wait and see this world be taken over by cheaters. This is not the first time a technological upgrade has knocked down a trust mechanism. In 1999, Napster came online. Before that, the music industry's view was that "consumers will pay for an entire CD just to hear one song." It evaporated within 18 months. The industry spent ten years moving from iTunes to Spotify before finding a new equilibrium. In 2008, the Bitcoin whitepaper threw out a premise: humans don't need to trust financial intermediaries, cryptography can replace trust. The entire crypto industry is essentially rebuilding a verification mechanism that doesn't rely on human goodwill (trustless). Princeton's Honor Code and the belief systems of past eras share the same fragility: they all assume "the cost of cheating is high enough." Once that cost is compressed by technology to near zero, no matter how lofty the system's design philosophy is, it cannot hold up. Education shifts from "trusting humans to do the right thing" to "assuming humans won't do things the right way." The final moral debate will gradually escalate into whether technological upgrades themselves are a kind of error. Princeton's Honor Code wasn't destroyed by AI — it "graduated." At the same time we can imagine: Princeton students generally come from substantial backgrounds before being admitted. After graduation, they will gradually climb the social hierarchy and gradually acquire the power to control this world. Students at elite institutions cheating collectively is faintly telling people that students around the world are "cheating." This is a major declaration: education will change dramatically because of AI, old values will be adjusted, and perhaps what counts as cheating today won't count as cheating five years from now. "Hey professor, I can't turn off my brain-computer interface for the exam — they're built in."