Nearly 60,000 university applicants in Mexico were forced to retake their entrance exams after abnormally high scores raised suspicions of widespread cheating. In Portugal, a minister’s push to digitise exam marking backfired so badly it triggered the country’s worst education crisis in decades. In India, a leaked exam paper affected millions of students and was linked to more than a dozen suicides, forcing the education minister to resign.
Three countries, three different failures, one shared pressure point: AI is changing what exams can reliably measure, and the institutions running them have not caught up.
The India Crisis
India’s scandal was the most severe. A leaked exam paper for a national-level test reached millions of students through what the Guardian describes as an “education mafia” — commercial operations that steal and sell exam papers. The fallout grew into the Cockroach Janta party, a youth-led protest movement that began as an online joke and became a nationwide challenge to Narendra Modi’s government. The education minister, Dharmendra Pradhan, resigned in July.
The protests were not only about the leak. They were about what a leaked exam means when a single test determines a young person’s entire future. “The results of one examination determining a person’s entire future really creates untenable pressure on young people,” said Dr Sarah Eaton, a leading expert on academic ethics at the University of Calgary. “This is a tragedy when young people are dying by suicide because of an examination outcome.”
Mexico and Portugal — Different Failures, Same Friction
Mexico’s National Autonomous University (Unam) discovered that entrance exam scores were abnormally high across the board — a pattern consistent with coordinated cheating rather than individual misconduct. The university’s response was to invalidate the results and require approximately 58,000 students to retake the exam.
Portugal’s problem was not cheating but the attempt to stop it. Education minister Fernando Alexandre pushed to fully digitise the exam marking process, and the result was serious marking errors that sparked protests from parents, teachers, and students demanding his resignation. The digitisation that was supposed to modernise assessment instead became the crisis itself.
The Technology Layer
Dr Thomas Lancaster, a computer scientist at Imperial College London who researches generative AI in education, says cheating has gone hi-tech in ways that traditional invigilation cannot address. Students use earpieces, miniature cameras, and real-time communication with people outside the exam room who can search for answers or use AI directly.
“With any kind of communication technology, you can have that expertise waiting outside for you in a human form,” Lancaster told the Guardian. “They either know the answers; they’ve got the expertise; or they can look them up online. You can go further than that now and have access to AI to communicate directly during the exam as well. It’s spy technology, really.”
This is the structural problem we have been tracking. When US colleges shifted to oral exams to counter AI-generated essays, it was a local adaptation. Denmark announced this month that teenagers will have to verbally defend their written essays from the new term. Eaton has done the same in her own classes. “AI can generate a perfect essay in seconds,” she said. “So an essay is no longer a viable way to demonstrate students’ learning. We should let it die with dignity and find other ways to assess student learning.”
Why This Year Feels Different
Eaton’s read is that the pressure is not just about cheating — it is about what happens when a fragile system meets a technology that breaks it. “When you get large-scale fraud combined with systemic failure, there’s almost a guarantee of a breakdown of public trust,” she said. “And once that trust is broken, it’s almost impossible to regain.”
The economic context sharpens everything. In a precarious global job market, a single exam score carries more weight than it should. Students who feel that one test will determine their entire future are more likely to cheat, and more likely to protest when the system fails them. The AI cheating methods we have covered before — smart glasses, hidden earpieces, real-time AI access — are the supply side. The demand side is economic desperation.
The Uncomfortable Question
Eaton raises a point that cuts against the instinct to build better anti-cheating technology. Institutions have been hoping technology can prevent cheating “easily and cheaply en masse,” she said. That hope is failing. Portugal’s digital marking system did not prevent errors — it caused them. Mexico’s entrance exam did not detect cheating in real time — it required a mass retake after the fact. India’s paper leak was a human failure, not a technological one.
“This idea of these mass examinations happening at scale, it sort of dehumanises the learning process and the assessment process,” Eaton said. “I think we’ve lost the thread. Learning should be the most important thing for students.”
That is easy to say from a university in Calgary. It is harder to implement in a country where 1.4 billion people are competing for a limited number of university places, and where the exam is the only mechanism that is perceived as fair. But the 2026 exam season — protests on three continents, a ministerial resignation, student suicides — is the clearest signal yet that the current model is not holding.
