Should universities ban ChatGPT, or regulate it? The University of Rouen Normandy (URN) has chosen: its board of directors voted on 2 October 2026 to roll out a charter on the use of generative AI, published on 20 October. It applies to everyone: students, doctoral candidates, teachers, researchers, administrative staff and temporary lecturers. This university AI charter is a textbook case because it picks rules over prohibition.
What the University of Rouen Normandy AI charter says
The text does not ban generative AI. It sets common rules to protect data, the integrity of work, academic freedom and the quality of learning. It complements the internal regulations, the academic regulations, the information-system charter and the data-protection policy. Where they diverge, the internal and academic regulations prevail.
Four principles structure the charter:
- Curiosity: encourage supervised experimentation, tool comparison and training for staff and students alike;
- Transparency: declare any use of AI whenever the output is academic, scientific, professional or a creative work;
- Precaution: assess whether a tool authorised by the institution is relevant before using it;
- Frugality: ask whether the need justifies a request that uses significant energy and hardware resources.
A simple rule: the tool level depends on data sensitivity
Julien Réveillon, vice-president for Environmental and Societal Transitions and Digital Strategy, sums up the mechanism in the university's announcement (our translation from French):
"Each level of data sensitivity corresponds to a level of tool. A public document can be processed with most online assistants. Personal or confidential data requires a tool controlled by the University."
For a student, that means a mainstream assistant is fine for rephrasing an already public text, but not for uploading an application file, internship data or exam papers. According to the university, the charter fits within the GDPR and the European AI regulation. We covered one of its transparency obligations in our article on AI labelling and the AI Act for students.
Why regulate AI at university rather than ban it
The OECD Digital Education Outlook 2026 offers a deeper argument. The studies it reviews show that students using general-purpose generative AI tools often produce higher-quality work, but that the advantage disappears, or even reverses, when AI access is removed, such as in exams. Offloading cognitive work to a chatbot can also lead to "metacognitive laziness". By contrast, tools designed for teaching show more sustained benefits.
The report concludes that how AI is designed and used matters more than whether it is used at all. On teachers, it cites the OECD TALIS survey: 37% of lower secondary teachers reported using AI in their work in 2024, and 72% were concerned about academic integrity. A charter that requires declaring AI use answers that concern directly, without hunting students down.
Watching dependence: a rare point
One passage of Julien Réveillon's statement deserves public actors' attention:
"The University must avoid becoming dependent on services whose evolution it does not control."
He mentions examining "better-controlled solutions, developed locally or pooled between institutions", and says the charter is meant to evolve with practice. It is the same issue as cloud and European models, which we discussed in our analysis of Mistral and NumSpot: a public institution relying on foreign tools is exposed to their price and access changes. On funding, see also our results on public AI funding in France.
Who wins, who loses
- Students gain a clear rule and announced training, but lose the grey area that left room to interpret what must be declared.
- Teachers gain a common framework, without having to invent their own course rule.
- Institutions without a charter risk inconsistent practices from one course to another.
- AI providers gain a supervised customer, but will have to convince on data protection.
What it changes for students
Three habits to keep, which the Rouen charter illustrates but any student can adopt: declare your AI use when it contributes to submitted work; never give personal or confidential data to a mainstream tool; and use AI to learn, not just to produce. At Prometheus we follow these institutional charters, because they sketch the rules students across France will apply. Ask your university for its own: if there is none, it is a good cause for student engagement.
Sources:
- Université de Rouen Normandie, « L'URN se dote d'une charte d'usage de l'IA générative » (20 October 2026)
- Charte d'usage de l'intelligence artificielle générative, URN
- Digital Skills and Jobs Platform, "OECD Digital Education Outlook 2026: how generative AI can support learning when used with purpose"
