ULB Sanctions Students Over AI Cheating in Exam
The Free University of Brussels (Université Libre de Bruxelles, ULB) has sanctioned a small number of students suspected of using artificial intelligence to cheat on a June examination, in a case that has become a flashpoint in the debate over how universities should police generative AI in academic assessment.
The students attend the Solvay Brussels School of Economics and Management, the university’s economics and management faculty. Fewer than ten are reportedly involved, according to BruxellesToday.
How the Suspicion Arose
The exam in question was a computer programming (Python) assessment held in June 2026. According to RTBF, suspicion surfaced during grading. Several answer sheets contained correct answers to complex, multi-step problems but omitted the intermediate calculations needed to reach them. Faculty experts judged it impossible to solve the exercises entirely from memory.
The university suspected outside assistance, potentially delivered through a connected device and an AI tool. One student claimed to have erased their work with correction fluid; a chemical analysis found no traces of calculations beneath it, disproving the claim, according to News Minimalist.
Two students who had previously been sanctioned for similar conduct — repeat offenders — had all of their May and June exams annulled and were barred from retaking them in the second session. Their appeal to the Belgian Council of State was rejected.
A Detection Problem With No Easy Fix
Policing AI use in exams is proving technically and legally difficult. ULB Vice-Rector Valérie Piette, who is charged with overseeing AI use at the university, described the methods students use to conceal assistance. “Connected glasses, hidden electronic devices, or more simply, a concealed phone,” she told RTBF.
Academic authorities are not permitted to search students, which complicates the detection of concealed devices. Even AI-detection software is unreliable. Lenaïk Dumont, president of the Fédération des étudiants francophones (FEF), warned against treating every strong answer as evidence of cheating. “Having a text entirely written by Copilot or another AI, putting it into an AI detector, and the text was detected as 100% human,” she said. “How can we truly certify 100% that a student didn’t do it themselves and that it was the AI that did it for them?”
The University’s Response: Rethink Assessment
The case has injected new urgency into ULB’s evolving approach to generative AI. The university has already revised its general study regulations twice on AI-related matters and plans further changes that would allow a teacher, in cases of genuine suspicion, to question a student again about a specific answer — not to administer a new exam, which is forbidden.
The broader philosophical orientation comes from ULB Rector Annemie Schaus. At the university’s academic opening on 18 September, she rejected a blanket ban on AI, arguing that “the obsession with cheating diverts us from the only question worth asking: what does it mean to learn in a world saturated with machines that resonate, calculate and write with unprecedented fluidity?”
According to Belga, via DHnet, Schaus framed learning as productive friction. “What the machine produces at the touch of a keyboard is precisely the abolition of that friction,” she said. “To teach or learn without friction is not to educate but to indoctrinate.” She called for moving evaluations toward oral, in-person formats and work done “under our eyes.”
Obstacles to Reform
That shift is easier to endorse than to implement. Catherine Dehon, dean of the Solvay faculty, noted that connected glasses and earpieces make AI use easier in large auditoriums, and that oral exams would require doubling funding. Signal jammers, she said, are not feasible for safety reasons.
Sanctions for cheating at Belgian universities range from a zero on the exam to the annulment of an entire session, up to expulsion for several years in cases of repeat offenses. As L’Avenir explains, Belgium’s décret Paysage leaves the definition of penalties to each institution, meaning procedures and outcomes vary from one university to the next.
A Wider Belgian and European Challenge
The ULB case is not isolated. In Flanders, several students filed appeals with the entrance exam commission alleging that fellow candidates used ChatGPT to pass the medicine entrance exam, after the pass rate reached an unusually high 47 percent, as RTBF reported.
Universities are responding by formalizing rules rather than relying on detection alone. KU Leuven, for instance, has published guidelines on the responsible use of generative AI for the academic year, enshrining two principles: that a student bears full responsibility for what they submit, and that assessments must allow teaching staff to evaluate which competences the student has actually acquired.
What to Watch
The ULB episode crystallizes a tension universities across Europe now confront: the tools that make cheating easier are the same tools increasingly expected in professional life. Detection is unreliable, searches are prohibited, and oral assessment is costly to scale.
As ULB prepares another revision of its study regulations, the question is whether procedural tweaks — questioning a student about a specific answer — can restore confidence in written exams, or whether institutions will be pushed toward the more expensive, harder-to-scale alternatives that Rector Schaus has already endorsed.