They used to trust everyone.
Then the algorithms arrived.
Princeton University is reversing a policy older than the internet, older than most modern nations. Starting July 1, eyes will be watching. Humans, not cameras. Instructors sitting in rooms with students who take the honor code personally—or pretend to.
This marks the first time since the 1880s that exams will be supervised on campus.
A massive cultural pivot.
One triggered by the sheer ease of AI cheating.
The faculty wanted it.
The students, surprisingly, did too.
Why? Because the old system—reliant on anonymous peers to police each other—is breaking under the weight of generative AI. Smartphones hide everything now. A student can prompt a bot, glance at a screen, and write nothing from their own brain. Detection is nearly impossible if no one is looking over their shoulder.
And who wants to be the one reporting it? Social media retaliation looms large. Doxxing. Bullying. It makes honesty dangerous.
“If students alone are present in the exam room and students are unwilling to to report, then there is is no check against misconduct.”
Michael Gordin, Dean of the College
The numbers don’t lie. A 2025 survey found 30% of students admitted to cheating.
Thirty percent.
Yet, few appeared before the Honor Committee.
The silence is deafening.
Princeton’s administration voted unanimously in April to restore proctoring. It is a return to a practice abolished in 1893 specifically to build a culture of trust.
Now that trust needs a bodyguard.
Students must still sign attests saying they followed the rules.
But someone will be there to watch them do it.
A Pattern Across Higher Ed
Princeton isn’t alone.
Everyone is scrambling.
Duke University stopped using numerical ratings for college application essays in 2024. The logic was cold but practical. You can’t trust the words anymore. If AI wrote the passion, is it still passion? Christoph Guttentag, dean of undergraduate admissions, said they could no longer assume essays reflected the true applicant. They still grade grades and activities, obviously.
But the soul of the essay? GONE. Or at least, unverifiable.
It’s a broader anxiety.
Researchers at Foundry10 see students paralyzed by uncertainty. They don’t know where the line is. Can they use AI to brainstorm? Yes. To fix grammar? Sure. To write the actual argument?
Plagiarism.
That line is blurry, so schools are stepping back and putting fences around the pasture.
Jennifer Rubin, a senior researcher there, notes that increased oversight is the default move when norms fail.
Proctors relieve immediate pressure. They stop the easy wins.
But AI is ubiquitous. It lives in our pockets. It waits.
Putting a human in the room buys time. It buys clarity.
Does it solve anything?
Probably not.
It just makes the game harder.
The tech evolves faster than the rules anyway.
Will students adapt?
Probably.
Will schools add another layer?
Inevitably.
The trust experiment is on pause.
For now, there are eyes in the room.
