The serious issue of the recent UNAM (National Autonomous University of Mexico) fraud in online admission exams snapped me back, quick as lightning, to an anecdote from my adolescence and high school. Professor Canseco, a scholar who taught us Mexican history, had his sights set on one of my classmates, whom he increasingly accused of using a cheat sheet during his exams. However, he never caught him red-handed and, without evidence, could not formally charge him. His only argument was that my friend’s behavior in class and personality—admittedly, unconcerned about academics—did not align with his high grades in the subject. Indeed, we all knew Juan was not a “very dedicated” type of student; instead, he was a brilliant cheat-sheet user.
Determined to confirm his suspicion, Canseco prepared to catch him. Legend has it that during the next exam, the teacher did not leave his desk for a moment, watching the student who had already become an obsession for him. Fixing his eyes on my friend, never taking them off my friend for even one moment, he spent the hour by his side, sure that he was hindering any attempt at fraud. At the end, satisfied, he collected the students’ exams. But he did not know who he was dealing with. In the next class, the exams already graded, he handed dear Juan his paper, remembering, with resigned irony, the old magician motto: “The hand is quicker than the eye.” My friend had scored 10 out of 10.
Returning to the UNAM admission exams: Clearly, like Professor Canseco, the university authorities were sure that, with the numerous control resources they had implemented, they could monitor the first online admission exam for the bachelor’s degree, convinced that they were scrutinizing each applicant in a way that was impossible to circumvent. However, the fraud was enormous and, once again, quite dramatic – it had to be recognized that the hand is still faster than all available technology.
Perhaps it is worth briefly recounting the surveillance measures by the company UNAM hired to administer and monitor the process, which I imagine were expected to be almost infallible.
Thus, as I mentioned, the first online admission exam for UNAM bachelor’s degrees was chosen to reduce expenses and facilitate access for the thousands of students who previously had to travel from various Mexican states to Mexico City to take the exam in person. Now, registrants no longer had to stand in long lines at mass-application venues (such as Azteca Stadium) and remain at least 2 meters from the nearest applicant to avoid copying. During the online exam administration, the students sat in front of a computer with their cameras and microphones on. Access to the exam platform began with recognizing the student’s identity (previously duly registered), which immediately prevented the student from opening any web page or program on the computer to perform any unnecessary function to answer the exam questions (for example, no copy, no paste).
The clear and strict instructions and reasons for exam cancelation were made known to the candidates: no visual or auditory support material could be used (no book or paper, or electronic devices, including watches; for obvious reasons, glasses were not excluded because of already proliferating intelligent versions), nor were students permitted to exit the system to open other tabs; nor could anyone accompany them (the system could capture out-of-camera range presences, voices, noises and movements) and much less communicate, for any reason, with a third person. It was also forbidden to “remain outside the camera’s field of view; deactivate or cover it or the microphone with any object; cover one’s face with hair, hands, or arms; read aloud the test questions or avert one’s gaze repetitively.”
The camera and microphone used an artificial intelligence system trained to detect any of these violations. But the AI did not have authority to cancel the exam; instead, it immediately alerted human supervisors (hired by the company, as a kind of qualified call center, with one supervisor for every 150 examinees), who assessed the relevance of sending a case to the decision-making panels, comprised of people from UNAM. Only these were responsible for the final resolution.
This process annulled 2% of the exams, that is, approximately 3,200 of the nearly 159,000 completed. The disqualified applicants received evidence of their irregularities.
But those were not all the anomalies detected. In the final tally, the exam scores showed results never seen before, with a fivefold increase in the number of students with more than 100 correct answers on the exam (which had 120 questions), compared to previous years (jumping from 3.5% to 16.3%). The number of exams with perfect scores (120 out of 120) was also scandalously abnormal; incidentally, most were from applicants to the Faculty of Medicine, which in previous years had required a minimum of 111 correct answers for admission, which was striking.
A change in the results was perhaps foreseeable if responding at home, alone, and online allowed for better performance. Still, the starkly contrasting statistics suggested a good number of illicit cases. (We should be cautious about jumping to conclusions if we want to avoid further normalizing the culture of fraud, and saying too soon “Oh, they must have cheated,” as if this were the only thing that could be expected of our young people; all possible variables must be discarded before claiming—as in the end, I believe, correctly—that many of these cases were dishonest).
There could have been many ways to carry out this type of fraud. First, some examinees might have circumvented the contracted company’s surveillance measures. Here, above all, deception is associated with the use of artificial intelligence, discrediting it (even more than it already is). But there is room for confusion in this. (It gives the impression, pardon me, that, once again, we are inclined to assign blame for the sake of venting without understanding what we are talking about). I will try to explain:
Fantasy, always open to imagining monsters and seeing AI as one of them, seems to assume that an artificial superintelligence, in the hands of some applicants, would have been able to circumvent the platform’s security undetected and answer the questions “from the inside,” which sounds more like a conspiracy than reality. To begin with, it is doubtful that such an expensive hack would have been accessible to many students eager to get into university. What happened is much simpler. Some of the young people said they asked the AI for advice on circumventing the exam’s security measures. Their methods were along the lines of “Locate the blind spots of your camera and place a second device there, which you can consult for answers.” This does not imply a complex tactic to outwit the platform, much less hack it; above all, it is nothing any good contemporary cheater could not devise.
Another crime imputed (again, very lightly) to AI in this fraud is that many, once they circumvented the surveillance, would have used it to consult the answers. Well, yes, of course; although the truth is that being able to consult a document would not have mattered whether it was a book, notebook, study guide with answers, stolen exam, virtual encyclopedia, or artificial intelligence platform. Perhaps the easiest option for cheating would have been the last one, given AI’s speed and scope, but that should not put it in anyone’s crosshairs as “the” tool for fraud. It would be like blaming Wikipedia because some students used it to answer.
So far, some of the committed fraud could have been achieved by circumventing surveillance, as any expert high school student does in class, and by consulting all kinds of cheat sheets and documents, many of them quickly generated by artificial intelligence.
Let us now move on to something more serious: the possibility of systemic fraud organized by people outside the student population to benefit some “well-connected ” students and/or to do business. This could have taken various forms.
A curious inference detected the simplest: it turns out that, in addition to the unusual improvement in grades, there was another strange downward irregularity: the bad grades, those with less than thirty correct answers, including zeros, increased drastically. This led experts to suggest that some “aspirants” took advantage of the fact that the exam would be administered on different dates to different groups and registered only to get the questions, without even bothering to simulate some answers.
What would they have done with that information? The question escalates the problem. Give it to their sister or friend? Sell it personally to an acquaintance or many? Give it to an organization — small or large — that would negotiate with it? All three options are possible. So is the fact that, at the culmination of the fraud, the distribution of the questions — by business or by “gift” — came from within the university institution itself or from the contracted company. (Notably, in offering her support to UNAM in the investigation of this matter, the President of Mexico offered that the Attorney General’s Office, the FGR, could, in effect, investigate said company).
I will only mention the enormous range of possibilities we encounter. None of them is extraordinarily different from what we have experienced before, in a thousand ways. We know that this type of fraud has always been possible (and by “always” I mean since prehistory, when the first young sapiens were put to the test) and that, again, technology is only a secondary resource. Whether someone sells, buys, or obtains the answers through a cheat sheet, a book, by phone, online, or by telepathy, matters little.
When UNAM became aware of the fraud and announced it, it had already published the names of the applicants who had been accepted and could register. So, anticipating any reaction from the institution, both some of those already accepted and those rejected began to demand, the former, that the grade be upheld, and the latter, that the exam be repeated. The latter, without a doubt, was the most logical, and the rector announced a new face-to-face proctored exam.
This was carried out recently. It was open, not, of course, to those whose exams had been canceled, but only to those who appeared on the accepted list, also allowing those who were not on it to compete again, if they had attained, at least, the same number of correct answers as the students admitted in past years. The two groups totaled about 58,000 students.
The face-to-face exam was held in four venues across the country on computers without an internet connection, under the surveillance of live proctors who were as strict as possible (remember: the hand is faster than the eye).
Absenteeism was enormous: 37%. This could have been due to many causes. Face-to-face implies more effort, and many applicants were presumed to have already been accepted to other universities, but suspicion that the new difficulty scared away several fraudsters continues to grow.
At the time of this writing, there are two important pending issues: the list of the newly accepted has not yet been published, and there is still no progress in a possible investigation into what would have been a systemic fraud, coming from within the institution itself, from the company that administered the exam, or from individuals and companies associated with them.
I imagine – perhaps naively – that as for the new list of accepted students and the admission to classes, everything will proceed normally. (Most likely, whoever obtained a good grade honestly in the first exam will also do so in the second; however, some injustices cannot be ruled out, and, therefore, certain protests and legal appeals will arise to be addressed. Likewise, some failed defrauders will want to clear their name, staging protests that I don’t think have much significance (of course, unless some “influential” surname appears among them).
As for investigations into institutions and companies, we should expect anything. The scandal may thrive or come to an end, expose one or more officials, or come to nothing; dethrone the contracted company or vindicate it; find a scapegoat among individuals who would have acted in isolation or were hired by some fraudulent company, or allege the fraud only to the students whose hands were quicker than the eye, whether bionic or biological. I dare to think, however, that things will soften and be forgotten, at least publicly, without casting shadows of doubt over the second selection process, already quite chaotic.
As for the inquiries about the use of artificial intelligence in this fraud, I think that either I am not understanding something, or it is a somewhat rhetorical and exaggerated distraction. What, in this case, can be said about AI that is not like a movie plot? I don’t know. I don’t think such fiction is necessary, given the readily available mechanisms to cheat; we don’t have to assume sinister technologies. However, we will have to be attentive and alert, lest monsters do exist.
Finally, as for the future, I dare guess that in the coming years, UNAM will continue to hold its online admission exams, but only as a first-round filter before a more limited face-to-face exam. That is, the same as this year, 2026, but officially formalized. I think it would be the best way, one that would take the most advantage of this experience learned the hard way.
We’ll see.
