Recently, a colleague of mine posted on his Facebook page an article about an AI-supervised entrance exam to the National Autonomous University of Mexico (NAUM) that went so poorly, 58,000 of the 160,000 applicants will have to take it again. This was the first time the entrance exam was given remotely. A “lockdown” browser with an AI-powered webcam, along with proctoring software, was required to take the entrance exam. It was an absolute disaster.
Between 2021 and 2025, 3.5% scored a 100 or more on the 120-question multiple-choice exam. This year (2026), it was 16.3%. At the high end of the scores, also between 2021 and 2025, 0.9% scored a 110 or more. This year (2026), it was 5.5%.
A commission was appointed to find out what had gone wrong, and its recommendation was to give a control exam in person. At least 58,000 people from 2021 to 2026 will have to take the control exam. Places at NAUM will be determined by the results of the new test. The question that the commission did not answer was how the remote exam went so wrong. There were multiple security systems in place besides the Lockdown Browser from Respondus and the AI-powered webcam. UNAM used a proctoring system from Territorium to monitor the students' webcams, watching to see if someone else took the place of the applicant or if the applicant left the camera frame.
There was also one supervisor for every 150 applicants. Supervisors received alerts from the system and used those alerts to notify UNAM. None of the security seemed to work correctly. UNAM cancelled about 2% of the tests for undisclosed reasons, so apparently, it caught some of the cheaters but not anywhere near what was needed to justify the use of all of the systems. It appears that, in the end, the only thing that can defeat cheating is making people sit in front of someone who is watching them. Bonus: no expensive security systems needed.
My colleague posted the article to his Facebook page and said, paraphrasing here, “When this happens, and it will happen, know that our current leaders at <college redacted> were warned but it didn’t seem to bother them as long as enrollment went up, not down.” I have known this person for decades. We worked at the same high school in the same department. We both left out of disgust and frustration. He found a full-time job at a local community college, the one about which he is speaking.
He has a tendency to get carried away when trying to make his point. In this case, he’s not. He is 100% correct. I know the community college he works for; I used to work there as an adjunct (part-time) professor until I got a slightly better-paying job at another community college as an adjunct professor. Today, community colleges, colleges, and universities are more interested in getting as much money as possible from as many students as possible for as long as possible. The more money brought in, the higher the salaries for the administration. Forget the students actually learning something; the administration is after the money.
Apparently, the more expensive the school, the worse the problem is. Inside HigherEd, an online publication reporting on events, resources, jobs, and other things in higher education, reported that a professor at Brown University, a very old, very expensive private university in Providence, R.I., allowed his students to take the midterm exam at home. He had never done that before and probably never will again. When most of his students made perfect or near-perfect scores, he knew they had cheated by using AI or some other method.
The final exam was in person. Eighteen students dropped the course, nine students simply didn’t show up for the final, and three who did take the exam made a zero. More than half of the students who took the final failed. Certainly, a world of difference from the mid-term. In the end, the professor dropped the mid-term exam and made the final exam 80% of the grade. Nineteen students out of a class of 59 failed.
When the professor brought the information he gathered to Brown’s Standing Committee on the Academic Code, crickets. When he went public months later, the Committee, through his department head, told him to submit individual complaints for each student suspected of cheating along with copies of their exams, and they would look at the complaints. “Their response, I must tell you, is seen as appalling and insufficient…,” the professor is quoted as saying at the end of the article. He will meet with his dean to see what else can be done.
I have been teaching since 1998, in both high school and college, when the Internet was still being figured out, and AI was science fiction. None of what I read while researching this article surprised me that much. The professor from Brown and his reaction to the lack of action from the college was a little bit surprising. He, apparently, thought that the Standing Committee on the Academic Code was actually going to do its job. Foolish man. No, like public education, the higher education viewpoint became “It just has to look good” quite some time ago. The colleges and universities do not want to do anything to turn off the spigot of money from mom and dad, the government, or whoever is paying the tuition.
I teach U.S. History and English Composition at the college level, the freshman survey-level courses. I give exams for U.S. History but not for English Composition. For English Composition, the papers I assign are the exams. I write my own questions and, when the students are face-to-face with me, they handwrite their answers in a Blue Book. In my online courses, I tell my students that if they suddenly go from my having to remind them to capitalize proper nouns to writing like they’re in graduate school, I’m going to notice, and it won’t go well for them. I choose to face the problem head-on.
I know my students will be tempted to use AI to cheat, but hopefully, if I tell them I’ll know, they won’t use it like they were intending to. I tell them, “I want to know what you know, not what the computer says you should know.” My approach has been pretty successful so far. I see AI as a tool, just as I saw Wikipedia as a tool. Using AI as a tool and not for cheating will not work in any school setting unless everyone is on the same page. When administrators refuse to enforce the rules, what recourse does a professor have? I think the professor at Brown did what he could do.
As far as I know, I’ve never had a student use AI to cheat in any of my classes. I’ve had students not come to the final exam, then email me later and ask for mercy. I’ve even had the dean of my section at the community college ask me to let a student take the final exam late because the student claimed his Internet connection stopped working. (I said no). With that particular student, his Internet connection only failed on test days. My dean wasn’t happy when I told him that. Because I am an adjunct professor, there’s not a lot the school can do besides not assigning me classes. That’s happened twice.
To combat students cheating by using AI, administrators are going to have to grow a backbone and enforce the policies on cheating. Until then, it will be up to the professors to stop it as best they can. I wish it weren't that way, but when your boss won’t back you up because he doesn't want to lose his nice, cushy job, there are few other options.
Also for our VIPs: The World’s Greatest Soap Opera That No One Is Paying Attention To
