Manual for AI use in Law: “Although you can’t possibly cover all bases”

Red, amber, green: what students are or are not allowed to ask a chat bot (or maybe a little)

22-10-2025 · News

MAASTRICHT. What can and can’t students use artificial intelligence (AI) for when writing their thesis and legal academic texts? The Faculty of Law at Maastricht recently published a manual that clarifies what and when. It is an ongoing project to determine what is preferable, said associate dean of research Anke Moerland, “but we have to move with the times, AI is a fact and it is being used”.

There is currently no uniform, university-wide policy when it comes to the use of AI and tools such as ChatGPT – although there is a so-called policy framework with general guidelines, there is little practical advice – so the Faculty of Law decided to take the matter into their own hands. The result is a seven-page document to aid all the bachelor, master and PhD students.

“Our focus is mainly on the use of AI in theses, what is acceptable and what isn’t?” said Moerland, who in addition to being the associate dean is also a professor of Intellectual Property, Frontier Technologies and International Trade at the Faculty of Law. The manual, which according to Moerland is “a living document” that must be updated whenever there are new developments, contains examples of what is allowed (green) and what is not allowed (red). And in between those two sits the amber light: be careful.

For example, if someone were to ask ChatGPT for a summary of a law or of court decisions, they would be on thin ice: if correct, the summary may be used to understand the material better, but it must not be included in the thesis. Anyone asking for a completely formulated research question or for analysis to be broken down into easily understood snippets is most definitely in the wrong. On the other hand, students are allowed to look up information and literature, and the definitions of words. Having a thesis checked for grammatical mistakes is also fine, although the student would not be permitted to accept any suggestions from ChatGPT how to rephrase the text.

Experimenting

Moerland: “Maybe not every lecturer agrees with what is in the manual right now, but we have to move with the times: AI is a fact. And if students learn to use it critically, it may also be a good thing.” It means that so long as the thesis is not the main focus yet, there must also be room on the course for the use of artificial intelligence when writing legal texts. The faculty has no problem with students experimenting, but it does want them to develop their own writing abilities. “Lawyers depend on language and being able to formulate texts themselves. Before students start using AI tools, they have to know how to do that in the right way.” That is another job for the faculty (also described in the manual), as in the end, each student must become a responsible and experienced user.

Are those assessing the work then sufficiently able to determine what was produced by a student and what by a machine? “Unfortunately, that isn’t always immediately obvious. Some do have a feel for it, they see things, but you can never be entirely certain. That is why we demand transparency from the students when it comes to writing their thesis,” said Moerland. They have to show which prompts they gave ChatGPT and have to make clear arrangements with their supervisors what versions of their thesis they have to preserve. “If there are any doubts about potential fraud, then the supervisor has to speak to the student and record the necessary details for the Board of Examiners. And even then, you can’t possibly cover all bases, it isn’t perfectly black and white. Not even with this manual.”

Faculties make up their own rules

The Faculty of Law is not the only one with its own guidelines. Other faculties such as the Faculty of Psychology and Neurosciences (FPN), Arts and Social Sciences (FASoS) and the School of Business and Economics (SBE) have also made it a priority. “Faculties working on their own guidelines is not that remarkable,” said Moerland. The Executive Board would have liked to develop a uniform AI policy for UM as a whole, but that had proved impossible, due to the different wants and needs and the implementation. So instead, it is up to the faculties to work out the exact details. Moerland: “There are certain aspects where you can make the same decisions, but in the end, a legal expert will need something slightly different to an economist or a psychologist.”

For example, FPN is considering how students could use AI to test their own knowledge. In an effort to look beyond just ChatGPT, the faculty plans to compile a list of fifty AI tools – how do they work and what are the pros and cons. They have asked all the lecturers for input.

FASoS has created a document of nearly twenty pages which states, among other things, that lecturers can indicate in their own teaching material what is and is not allowed – for example, translating or rephrasing work, but also requesting analyses or interpreting research data. Students are always required to justify their use of AI tools.