Starting July 1, 2026, Princeton will proctor all in-person exams, ending a 133-year tradition. And it wasn't just faculty pushing for the change. Students said the old system had become too hard to enforce. In a 2025 survey of graduating seniors, 30 percent admitted to cheating at least once, and 28 percent reported using ChatGPT on an assignment that specifically prohibited it, more than double the percentage from the year before.

Stanford is doing the same thing. Its student-written Honor Code, dating back to 1921, established that students were expected to hold themselves and each other accountable, and that instructors would not take unusual precautions to prevent cheating, which included proctoring exams. After a pilot program, Stanford allows exam proctoring in all classes.

So what's going on? Colleges are being forced to ask a question they never really had to ask before: How do we know what the student actually knows?

Most people don't think about this part: honor codes weren't just about students doing their own work. They also relied on students reporting each other when they saw cheating. AI made that nearly impossible.

Cheating on a small personal device doesn't look like anything from across the room. And even when students notice something, many choose not to report it. Fear of social media callouts, doxxing, and peer shaming has pushed most Princeton Honor Committee reports to anonymous submissions. In the 2025 graduating senior survey, 45 percent said they knew a peer had violated the honor code but chose not to report it. Only 0.4 percent did.

Expecting students to resist the temptation of widely available AI tools during an exam, while also monitoring their classmates' behavior, while also trying to concentrate on their own work, is a lot to ask. Princeton and Stanford both concluded it was too much.

This one might surprise you: AI is making handwritten exams popular again. The California State University system paid OpenAI $17 million to give all 22 campuses access to a high-powered educational version of ChatGPT, with the goal of helping students learn to use it for school and future careers. But the announcement caught faculty and students off guard, leaving everyone to figure out the ethics on their own.

The result was confusion across the system. A recent Cal State survey of over 94,000 students and employees found that 52% of faculty said AI had a negative effect on their teaching, and 67% of students felt their professors don't teach them how to use it effectively. As of April 2026, only 0.7% of students and 16% of faculty had completed the voluntary AI training the university offered.

Some faculty embraced ChatGPT and built it into their courses. Others returned to blue books and scantrons. Some professors required AI-powered tools in one class while banning any use of AI in another. Students caught between those policies had no consistent guidance on what was actually allowed.

That is the tension colleges (and students!) are living in right now. AI is useful enough that universities want students to learn it, and disruptive enough that individual professors are redesigning everything around it, often in opposite directions.

You might wonder why colleges don't just use software to catch AI-written work. Some tried. Vanderbilt shut down Turnitin's AI detection tool because false positives were too risky. [A false positive means a real student gets flagged for cheating when they didn't cheat.] When thousands of papers run through the same tool, even a small error rate creates serious problems for real students.

Most colleges have concluded the better response is to change how learning gets demonstrated, not just police what gets submitted.

At San Jose State University, the writing center created an AI Writer Toolbox after tutors kept seeing students penalized by professors for using the very AI tools the university wanted them to learn. The toolbox helps students use AI responsibly, including how to cite it and how to avoid crossing into academic dishonesty. It also includes a disclosure tool where students fill out a form explaining how they used AI on an assignment. The form generates a certificate they can submit with their work.

That points to where things are heading. Submitting the assignment may not be enough. You may also need to document how you completed it.

Not every response to AI is about enforcement. Some schools are going the other direction entirely. Purdue approved an AI working competency as a graduation requirement starting Fall 2026. Ohio State launched an AI Fluency initiative with a goal that every student in the Class of 2029 graduates AI fluent in their field. 

The message from these schools isn't never use AI but make sure to:

• Know when it's allowed and when it isn't

• Use it honestly when it is

• Disclose it when required

• Never submit something you can't explain or defend

Admissions essays are part of this conversation too. Duke no longer assigns separate numerical ratings to essays. A polished essay is no longer clean proof of independent writing ability. What matters more now is specificity, personal detail, and authenticity. Generic polish has never been easier to produce. Your actual story hasn't gotten any easier.

The same principle that applies to your future coursework applies to your applications now: don't submit anything you can't speak to.

There is something worth acknowledging here. Honor codes built on mutual trust are a beautiful idea. The notion that a community of students could hold itself accountable, without surveillance or supervision, says something genuinely optimistic about human nature. But cheating wasn't invented by AI. It was happening long before ChatGPT existed, and probably long before the Princeton Honor Code was written in 1893. What AI changed wasn't student integrity. It changed the visibility of the problem and the difficulty of addressing it the old way.

The bigger lesson isn't really about AI at all. It's about being willing to ask whether the systems we rely on still work for the world we are actually living in. Princeton and Stanford asked that question. The answers weren't comfortable, but they were honest. That same standard applies to all of us. Whatever you submit, in your college applications or in your future coursework, make sure it reflects what you actually know and can actually defend.