On July 3, 2025, the day before the baccalaureate results were announced, Nina Viriot, an 18-year-old high school student from Paris, received an email that hit like a bombshell: a report of suspected fraud had been filed against her for her philosophy exam.
The reason? The examiner grading her paper suspected the use of artificial intelligence.
This accusation, which could have compromised her future, ultimately turned out to be a mistake. After a thorough review of her case, the Paris education authority cleared her of suspicion, allowing Nina to earn her diploma with honors and an impressive score of 18/20 in philosophy. However, this case raises burning questions about the use of AI detection tools in exams and the future of technology in education.
An error that highlights the flaws in AI detection
Nina Viriot’s story is not an isolated case, but it illustrates an emerging problem: the difficulty of distinguishing human work from text potentially generated by an AI. According to available information, the examiner judged Nina’s paper to be “too perfect,” perhaps due to it being typed—an accommodation granted to the student because of her attention disorder, which allowed her to use a computer not connected to the internet.
Yet, no official AI detection tools were used, and the Ministry of National Education confirmed that examiners do not have specific software or training to spot texts generated by AIs like ChatGPT 4.
This lack of reliable tools poses a crucial question: can we trust current AI detectors? These programs, often based on text analysis algorithms, look for patterns such as an overly uniform structure or a style deemed “artificial.” But their reliability is far from infallible.
A striking anecdote: some detectors have already flagged historical texts, such as the American Declaration of Independence, as AI-generated content!
Suspicion fueled by vague criteria
The initial accusation against Nina was based on the examiner’s “gut feeling,” without tangible proof. This lack of clear criteria for identifying AI use in exam papers reveals a gap in the Ministry of Education’s protocols. In 2024, out of 560 cases of exam fraud in France, 5% were linked to suspected AI use. But how many of these cases were based on subjective judgments?
Nina’s case shows that without reliable tools or adequate training, examiners risk making mistakes, with serious consequences for students, such as blocking their results or their university applications.
The Paris education authority finally admitted to a “clerical error” in the summons sent to Nina for a disciplinary hearing.
But this vague explanation leaves room for doubt: was there an analysis by an AI detector, or was the suspicion based solely on the examiner’s intuition? Without transparency, it is hard to say. What is certain is that media pressure and the public testimony of Nina and her mother, Emmanuelle Viriot, likely accelerated the resolution of the case.
Read also on this topic: Yiaho: The AI that fits the French framework for education
Students facing the temptation of AI
This story raises another question: will students increasingly use AI tools to bypass detection systems? With the rise of platforms like ChatGPT or other AI tools to “humanize” texts and make them undetectable, the temptation is real.
Online services like those on our Yiaho platform already offer to rephrase texts generated by AI to make them more natural. These tools use algorithms to mimic the imperfections of human language, making detection even more complex.
In a context where new technologies accounted for 52% of detected exam fraud in 2024, according to the Ministry of National Education, schools must be increasingly vigilant. Measures such as the use of signal jammers or metal detectors, as seen in Marseille, show a desire to control digital tools. But banning access to AI is one thing; distinguishing honest work from a suspicious text is another.
AI: Toward a new era for exams?
The Nina Viriot case is likely just the tip of the iceberg. As AI becomes more sophisticated, cases of suspected cheating are likely to multiply, as are errors in judgment. To avoid further scandals, the Ministry of Education will need to invest in reliable detection tools and train examiners in their use. But above all, clear and transparent rules must be established to ensure fairness in exams.
As Nina’s mother points out, the current procedure is “deeply inhumane,” leaving students in uncertainty and anxiety. In the meantime, Nina Viriot can enjoy her victory: not only did she get her diploma, but she also proved her integrity. Her story reminds us that AI, while revolutionizing our daily lives, poses ethical and practical challenges in the world of education.
Between the temptation to cheat and the risk of unfair accusations, the 2025 baccalaureate could well mark the beginning of a broader reflection on the place of AI in our schools!
Source: etudiant.lefigaro.fr
