Students are letting AI take their entire courses for them. The question is what your workplace can learn from that

A teacher in the US, with twenty years in the profession, sat reading the posts in her own course discussion and finally asked herself out loud: am I only talking to bots here?

She was, to a worrying extent. The New York Times recently looked into what’s happening in American online courses, and the answer is hard to exaggerate. Students aren’t just asking a chatbot for help with an assignment anymore. They hand over their login details to an AI agent and let it take the whole course: watch the lecture videos, do the exams, write the essays, and even join the discussions, pretending to be the student and chatting with classmates and teachers about the course literature.

While the agent studies, the student does something else. The paper mentions doomscrolling and video games.

What an AI agent is, in one minute

A regular chatbot answers when you ask. An AI agent takes a goal and carries it out on its own: it logs in, clicks, reads, writes, and submits, step by step, without anyone sitting next to it. It’s the same technology that can prepare your morning by reading your calendar and pulling up the right documents. The technology isn’t the problem. The problem is what happens when no one sees what it’s doing, and whose name it’s doing it in.

No one wants to be the one to stop it

The most thought-provoking part of the investigation isn’t the students. It’s the adults.

According to the paper, AI companies are doing very little. None of the three services students use most refused to write an entire essay, none prevent their models from logging into learning platforms, and teachers’ requests that the AI at least identify itself as AI when joining a course discussion have been ignored. One company responded that such a requirement would jeopardise students’ privacy. Read that sentence twice.

Universities, meanwhile, are sitting on the fence. Many have million-euro contracts with AI companies and hand out AI tools to both teachers and students, while every individual teacher is left to decide whether to encourage, tolerate, or penalise the use. A teacher at a university that just signed a big AI deal summed it up: I get no guidance on how to discourage students, I get courses on how to use it.

Encouraged and forbidden at the same time, and the rule depends on who you happen to ask. Does that sound like a workplace you recognise?

The retreat to the physical world

Some institutions have drawn a conclusion, and it’s telling: the only thing that reliably works is to leave the screen behind. A law faculty in Chicago has banned computers, tablets, and phones in first-year lecture halls. Princeton has scrapped its over hundred-year-old honour code, which let students take exams without supervision, after an AI scandal.

An honour code that held through world wars and pandemics didn’t hold up against AI agents. Not because students became worse people, but because the system relied on an assumption that no longer holds: that the person submitting something is the one who did it.

The same story is playing out at your workplace

Swap the students for employees, the learning platform for your systems, and the essay for a quote, and it’s the same story. The ingredients are familiar:

AI use that no one sees, in private accounts and tools management doesn’t know about. Mixed messages from the top, where AI is both encouraged and viewed with suspicion. And rules that, in practice, are decided by each manager individually, just like by each teacher.

The school world’s answer, to ban and retreat to pen and paper, isn’t an option for a workplace. No one bans computers in an office. What’s left is the path that works: make the use visible instead of invisible.

That’s why we built EuroWork so you never have to wonder who did what. AI answers are marked as AI answers, which has been required by the EU AI Act since August. Our AI colleague presents a plan and waits for approval before each step, and every run leaves a receipt of what was done and what it cost. Nothing is sent in someone else’s name, and what an agent writes becomes a draft that a human approves, with their own name and their own responsibility.

The teacher in the investigation really just wanted one thing: to know when she was talking to a machine. That’s not asking much. It should be the bare minimum, in school and at work.

Three questions for your next management meeting

Do you know which AI tools are being used in your systems today, and in whose name?

Are you sending mixed messages, with an enthusiastic central deal and silence on the local rules?

And if a customer asked who actually wrote your latest bid, how sure are you of the answer?

Anyone who can’t answer has the same problem as the universities. The difference is, you still have time to choose the other way.

Source

The investigation was published in the New York Times on 10 August 2026 and is behind a paywall: nytimes.com/2026/08/10/us/ai-cheating-online-degrees.html