From Tunis to Rabat, through European capitals, the fraud in national exams is gradually moving away from paper crib sheets toward the connected earphone and, now, artificial intelligence. A fundamental question arises: can we still monitor an exam in an era when every smartphone houses an assistant capable of drafting an essay in a few seconds?
It was barely 8 a.m., Wednesday, June 3, 2026, when the main session of the baccalaureate tipped into controversy. A teacher filmed herself presenting an essay produced by artificial intelligence and claiming that the topic had circulated a few minutes after the start of the exam. The scene, viral within a few hours, sums up the challenge faced by national exams: fraud no longer merely bypasses surveillance, it speeds ahead of the institution itself.
There were 162,435 candidates this year taking the main Tunisian baccalaureate session, organized from June 3 to 10, 2026, compared with 151,808 in 2025, an increase of 10,627 candidates. And from the first days, mini earphones, connected glasses, specialized Facebook groups, and increasing recourse to AI confirmed that the fight against fraud has become a true technological race.
Cheating is not increasing everywhere — but it is getting tougher
First observation, counter-intuitive: in volume, cheating in the bac does not “explode.” In Tunisia, it has even declined. In 2024, 821 cases were recorded during the main session, of which 83 % involved electronic devices. In 2025, the total fell back to 647 — distributed among 433 cases in public institutions, 143 in private ones, and 71 among independent candidates. Fraud has thus decreased in volume, but its methods have become more technological. The handwritten crib sheet has not disappeared, but it has become marginal.
The most concerning phenomenon for authorities lies elsewhere: professionalization. Specialized services now describe real structured organizations, active on social networks, offering “packs” for candidates. Among the equipment seized in recent months are mini earphones, Bluetooth pens, SIM cards, connected glasses, and even a drone. Cheating has become a market.
In response to this shift, the legislator is trying to catch up. In May 2026, lawmakers filed a bill providing, for organizers, financiers, or facilitators of fraud networks, penalties ranging from one to five years in prison and fines between 20,000 and 100,000 dinars. The text must still go through the parliamentary route before potential entry into force.
Where AI really comes into play
We must dispel a frequent misconception. The bac is a supervised and handwritten exam: you do not “paste” a text produced by an AI as in a homework assignment. The AI thus enters indirectly, in two ways.
The first is real time: an outside accomplice interrogates a model and dictates the answer to the candidate equipped with an earbud. The AI does not replace the connected device — it amplifies its effectiveness. The second path is the exploitation of leaks: when a subject circulates before the scheduled time, the AI can produce a model answer in a few seconds, as illustrated by this week’s Tunisian incident.
Morocco was one of the first in the region to organize its countermeasures. For the 2026 session, which concerns about 520,000 candidates, the kingdom generalized individual identifiers with QR codes and distributed 2,000 electronic detectors designed to spot active phones in exam centers. A device designed to respond to the rise of earphones, real-time assistance services, and, indirectly, AI-generated answers.
A shared challenge all the way to Europe
The trend has nothing Maghreb-specific about it. In the United Kingdom, where statistics are the most rigorous, phones and other connected devices accounted for 44.3% of the fraud sanctioned during the summer 2025 exams, i.e., 2,225 cases. They have constituted the leading category of fraud every year since 2018. In France, commission-handled frauds rose by 14% between 2023 and 2024, more than half linked to new technologies.
AI also raises the inverse risk there, just as grave: unfounded accusation. In France, a candidate allowed to take the exam on a computer due to a disability had their baccalaureate canceled in 2025 on the basis of a suspicion of AI use, notably due to an atypical writing style. The case reveals the other danger of this new battle: lacking a reliable method, the fight against fraud can also lead to fragile accusations.
The real answer cannot be limited to better monitoring
This is perhaps the most important point, and it goes beyond mere technique. Internet jamming, network outages, and seizures of equipment are defensive responses that, even according to observers, struggle to keep up: fraud networks professionalize faster than the devices meant to counter them.
The most lucid reflection may come from Rabat, where the Minister of Education asked the question differently: at a time when AI is accessible to everyone from a phone, the answer does not lie solely in surveillance, but also in a redesign of the exams, to evaluate personal reasoning, the ability to argue, to verify information, and to orally defend a process — all of which are skills harder to delegate discreetly to a machine.
The bac, the main symbol of social ascent across the region, thus plays a part that goes beyond it. It is no longer only the integrity of an exam that is at stake, but the school’s ability to measure what a student truly understands, builds, and defends, in a world where a machine can already produce an answer on their behalf.
