By Chuck Gallagher — Business Ethics Keynote Speaker and Trainer
TL;DR: Chuck Gallagher, AI ethics speaker and author, argues that the fight over whether students are cheating with AI or simply learning a new way misses the real failure — institutions never drew the line, disciplined 2.6 percent of the students who crossed it, and taught a generation exactly what the rules are worth.
A professor at the University of Chicago says that walking across his own campus now feels like being trapped in the middle of a crime spree. He is not describing a burglary ring. He is describing his students.
That professor is John Paul Rollert, who teaches leadership, ethics, and politics at Harvard and the University of Chicago. On August 30, 2026, The Atlantic ran his essay under the title “A Society of Cheats.” His argument comes down to a single word. Earn. Rollert wrote that every paper a student does not write and every serious reading they hand to a chatbot to summarize is an exercise in cutting corners — and that at graduation they collect a degree they did not earn.
He did not stop there. Rollert predicted that these graduates will walk into the working world carrying the qualities of a criminal syndicate: a knack for cutting corners, a disdain for community rules, and an assumption that integrity is the reserve of suckers and wimps.
Strong words. I understand the anger underneath them. But before you sign the whole indictment, look at what the students say.
Are Students Cheating, or Using a Tool Nobody Defined?
The Harvard Crimson’s 2026 senior survey found that 33 percent of graduating seniors admitted using AI on an assignment against their instructor’s permission, a share that has held steady in recent years. Here is the part that ought to stop you cold. Of those students, 93 percent said they were never discovered at all. Another 4.5 percent were caught and faced no discipline. Only 2.6 percent were disciplined.
Now set that beside one more finding from the same survey. Forty-three percent of those seniors said AI had harmed their education. Thirty-three percent said it had improved it.
Read that again. A third broke the rule. Almost none paid for it. And more of them believe the technology hurt them than helped them.
That is not a portrait of a criminal syndicate. That is a portrait of young people doing something they know is wrong, getting away with it clean, and not even feeling good about it afterward. As an AI ethics speaker and author, I have sat across from executives who described their own first compromise in almost exactly those terms. Nobody caught it. Nobody enjoyed it. They did it again.
What Does a 2.6 Percent Discipline Rate Actually Teach?
The lesson a 2.6 percent rate teaches has nothing to do with artificial intelligence.
It teaches odds.
I know something about how that lesson takes hold. I made choices that cost me my freedom and put me in a federal prison, and the slide did not begin with a dramatic decision. It began small, with something nobody noticed. When a compromise goes unnoticed, the mind does not file it under “I got away with that.” It files it under “that was fine.” The next one comes easier. Every choice has a consequence — but a consequence that never arrives is a consequence nobody believes in.
Watch how the slide works on a campus. You are behind on the reading, so you feed it to a chatbot for a summary. Then you ask for an outline, because the summary was so good. Then you ask it to draft the paragraph you are stuck on. By spring you are editing something you did not write and signing your name to it. Trust me on this. Nobody wakes up one morning and decides to become a cheat. They decide something much smaller, forty times in a row.
Has the System Given Up on Drawing the Line?
Now the other half of the question, the half the outrage tends to skip. Universities are not merely failing to catch students. Many have stopped trying.
Inside Higher Ed reported in August 2026 that Yale, Vanderbilt, Johns Hopkins, and Indiana University have adopted policies barring or discouraging faculty from relying on AI-detection programs as sole evidence of cheating. At least a dozen institutions, among them Northwestern, Georgetown, and New York University, have disabled Turnitin’s AI detection entirely after it produced far more false positives than the company suggested. Indiana’s Kelley School of Business now tells faculty to stop trying to catch AI use and to design assignments that require visible reasoning instead.
Some schools are moving the opposite direction. This fall the University of Chicago Law School is piloting a general prohibition, with narrow exceptions, on electronic devices in core first-year classes. Writing in The Washington Post in August 2026, Santiago Schnell argued that a degree should distinguish what a student can do alone from what they can produce with a machine.
And some are teaching the thing outright. Since the fall of 2024, the honors section of Penn State’s introductory chemistry course has had students use AI as a sanctioned exam-prep tool, with instructors tracking how their sense of appropriate use shifted over the term. Same technology. Completely different moral status. The difference is not the tool. The difference is whether somebody bothered to say out loud what honest use looks like.
So Is It Cheating, or Is It Learning?
Both are happening, and the machine does not decide which. The student decides, and the institution sets the terms.
Let me be clear about where I land. A student who uses AI to understand material faster is learning. A student who uses AI to manufacture work they then present as their own is committing fraud, and calling it a new way of learning does not change what it is. But an institution that never defines the line, retires its detection tools, disciplines 2.6 percent of known violations, and then acts wounded when students draw the obvious conclusion is not a victim of anything. It is a participant.
As an AI ethics speaker and author, the question I put to schools and companies alike is the same. What does honest work look like here, specifically, and what happens when someone does not do it? If you cannot answer that in a sentence, the people you lead already have.
Frequently Asked Questions
Is using AI for schoolwork always cheating? No. The behavior, not the tool, determines whether it is cheating. Using AI to summarize a dense reading you then study, to generate practice questions, or to explain a concept you missed in lecture is study assistance. Submitting AI-generated work as your own original effort is academic fraud, regardless of how common it has become.
How many college students use AI against their instructor’s rules? The Harvard Crimson’s 2026 senior survey found that 33 percent of graduating seniors admitted to using AI on an assignment without instructor permission, a figure that has stayed roughly flat across recent years. Of those students, 93 percent said they were never discovered, 4.5 percent were detected without consequence, and 2.6 percent faced discipline.
Why are universities turning off AI detectors? Detection tools have proven unreliable, particularly for non-native English speakers, and several have generated far more false positives than vendors initially indicated. Yale, Vanderbilt, Johns Hopkins, and Indiana University have limited or barred reliance on them as sole evidence, and schools including Northwestern, Georgetown, and New York University have disabled Turnitin’s AI detection outright. The practical effect is that enforcement now depends almost entirely on assignment design rather than surveillance.
What is the difference between using AI to learn and using AI to cheat? The test is authorship and disclosure. If AI helps you understand something and the finished work is genuinely yours, you are learning. If AI produced the thinking and you attached your name to it, you are misrepresenting your own capability. As an AI ethics speaker and author, I tell students and employees the same thing: if you would not describe exactly how you produced the work to the person receiving it, you already know which side of the line you are on.
How should employers respond to AI-assisted work from new graduates? Define acceptable use in specific, task-level terms rather than in a general policy statement, and say plainly what happens when the line is crossed. Ambiguity plus low enforcement is the combination that produced the campus problem in the first place. Assume incoming hires learned their AI habits in an environment that rarely corrected them, and treat the first ninety days as a teaching window rather than a testing one.
Ready to Draw the Line Before Someone Else Does?
Every organization hiring out of this graduating class is about to inherit the lesson those campuses taught. The habits do not stay in the classroom. They walk into onboarding, into expense reports, into client deliverables, and into the first quiet moment when someone realizes nobody is checking. The fix is not a longer policy document. It is a clear line, said out loud, backed by consequences people actually believe in — and a leadership team willing to talk honestly about how small compromises grow. I speak to companies, universities, and associations about exactly that: how ethical failure happens incrementally, why AI has made the first step frictionless, and what it takes to build a culture where honest work is the obvious path rather than the expensive one. If you are wrestling with where to draw the line on AI in your own organization, let us have that conversation at ChuckGallagher.com.
Five Questions for Reflection
1. Where in your organization is the line between acceptable AI assistance and misrepresented work still undefined — and who is currently deciding it for you?
2. If someone in your organization cut a corner today, what are the honest odds they would be caught? What is that number teaching them?
3. Think about the last shortcut you took that nobody noticed. Did it feel like getting away with something, or did it simply feel fine? What does your answer suggest?
4. Are you measuring output you can see, or capability you have actually verified? What would change if you had to tell the difference?
5. If a new hire asked you today, in one sentence, what honest work looks like in your organization, could you answer without reaching for the handbook?
