AI and academic integrity: what universities actually say

“Am I allowed to use ChatGPT for this assignment?” This is a question with no universal answer in France: each institution sets its own rules, and they can be directly opposed from one university to another. What is constant, though, are the underlying principles these texts all share. Here’s what official institutional charters, and a framework currently being discussed by the ministry, actually say.

Two opposite starting points, depending on the university

The charter at the Université d’Angers (which also applies to Polytech Angers, its engineering school) starts from the principle that a complementary use is allowed by default: generative AI “can be used as an assistance tool in preparing assignments, reports, and dissertations,” particularly as a language assistant or a help with finding information. But it sets a clear limit: “unless stated otherwise, work produced by learners cannot be produced by such tools.”

The charter of the UFR Sciences Humaines et Sociales (Humanities and Social Sciences department) at Université Paris Cité starts from the opposite principle: “without instruction, no tool or resource is authorized.” It’s up to each instructor to explicitly state, on the assignment itself, whether AI is permitted — in its absence, use is banned by default.

Two institutions, two opposite starting points. The only way to know what’s allowed for a given assignment is therefore to check the rule specific to your own institution and, if in doubt, ask the instructor or thesis advisor directly — never assume that another university’s rule applies at yours.

What comes up in (almost) every text

Despite this difference in starting point, the underlying principles overlap strongly:

Transparency is non-negotiable. The Angers charter notes that, since June 2023, the European Code of Conduct for Research Integrity has recommended transparency: “concealing/hiding the use of generative Artificial Intelligence […] in writing assignments is now considered a breach of research integrity.” A working document produced at the ministry level (see below) repeats exactly the same principle for research: concealing the use of generative AI in a publication is described as a breach of scientific integrity.

You remain solely responsible for the content. Whether or not AI is authorized, both university charters cited here are unequivocal: the student remains fully responsible for the accuracy of information and citations produced by the AI, including its errors. “The AI made a mistake” is not an acceptable excuse.

Sources drawn from AI must be cited like any other source. The Angers charter explicitly points to its institution’s plagiarism page and notes that information obtained via AI must be cited so as not to constitute plagiarism.

Some uses remain banned, even where AI is broadly allowed. For a dissertation or research report, the Angers charter bans using generative AI to “fabricate research data for a survey” — a point that matters in particular if your work relies on fieldwork, interviews, or real data collection.

How to disclose your use of AI

Good practice, repeated in both university charters, is to mention your use of AI within the final document itself — in a dedicated section, an appendix, a footnote, or an introduction — explaining precisely how the tool was used (document research, rewording, language proofreading, generating a first draft to be reworked, etc.). Université d’Angers even provides a disclosure questionnaire for its students.

A working document hosted on the website of the Ministry of Higher Education and Research — a draft charter still marked “DRAFT” at this stage, written by staff from Nîmes Université as part of inter-university discussions — takes the same approach for students: “make sure with the instructor that using AI is allowed as part of a course or an assessment” and “disclose the use […] in the relevant passages, following the institution’s recommendations.” This text isn’t (yet) a charter officially adopted by the ministry, but it gives a sense of where these discussions are heading at the national level.

What remains banned almost everywhere

Three uses stand out as problematic across all the texts reviewed:

• Claiming AI-generated content as your own, without disclosing or verifying it.
• Using AI to paraphrase an existing text in order to get around a plagiarism detector — a use explicitly targeted by the recommendations reviewed.
• Submitting personal or confidential data to a consumer-grade generative AI tool: the Angers charter bans this outright, including for interview transcripts or internal data from a company hosting an internship. Only data belonging to the public domain may be submitted to this kind of tool.

Penalties for non-compliance

The Angers charter specifies the possible consequences of fraudulent use (plagiarism, failure to disclose, data falsification): rejection of the assignment, report, or dissertation; referral to a disciplinary committee; and, in cases of proven fraud, penalties that can go as far as expulsion.

Key takeaways

• There’s no single national rule: some institutions allow AI by default with conditions (Angers), others ban it by default unless explicitly authorized (UFR SHS, Paris Cité). Always check your own institution’s charter.
• Transparency is the most consistent principle: never conceal your use of AI in graded work.
• You remain responsible for the content produced, even if the AI got something wrong — verification is on you.
• Disclosing your use happens in a dedicated section of the document (appendix, footnote, or introduction), not silently.
• Fabricating research or field data with AI remains banned, even where general assistance use is otherwise tolerated.

To help you tell apart citation, honest paraphrasing, and plagiarism — with or without AI in the mix — the citer-sans-plagier skill applies these same principles of transparency and verification.