The University of Groningen and Hanze University of Applied Sciences are approaching AI as something students need to understand and use critically. Their published policies, projects and plans show a combination of adoption and restrictions. For students, the difficult part is that access to a tool and permission to use it in an assessment are separate decisions.
At RUG, the university-wide rules allow AI as a supporting tool for activities including brainstorming, language correction, translation and practice questions, subject to additional faculty and course rules. If it generates content that replaces part of your own work, disclosure is required: identify the tool and version, explain the purpose and describe how you used it. Additional course conditions should be communicated beforehand through the syllabus or Brightspace.
Disclosure does not automatically make the use acceptable. RUG considers it cheating when outsourcing prevents an examiner from assessing your own knowledge and skills, or when required disclosure is missing. It also explicitly says AI-detector scores cannot serve as evidence of cheating. Suspicions go through the examiner and Board of Examiners; a software percentage cannot settle the case.
Hanze students should check their own programme’s rules. Its 2026–2027 Student Statute distinguishes institution-wide regulations from programme-specific education and examination rules, with programme information available through MijnHanze. A lecturer demonstrating an AI tool in class does not establish permission to use it for every submission. Check the assignment instructions and ask about the particular task you intend to delegate.
RUG’s current FAQ says Vibe Free, part of its Mistral pilot, is available to all RUG students, with a limit of 25 prompts a day. The university says students should enter through its institutional login route. It also warns that the free version is unsuitable for tightly timed assessments because of its usage limits. Availability still leaves individual programmes free to set educational restrictions.
Teachers are using AI too. RUG permits Mistral-assisted feedback on essays and theses under specified data-protection conditions. Students must be informed, and teachers retain responsibility for assessment. The guidance rules out letting the system make assessment decisions on its own. For a student receiving feedback, a useful question is how the lecturer reviewed the suggestions before passing them on.
At Hanze, the experiments are already specific enough to picture. Its Slim leren met AI project reports involving 80 students and four teaching teams, exploring how AI fits into professional assignments and students’ learning.
In Hanze’s account, lecturer Jascha Tuinier introduced an AI coach in Start up Scale up. In search-engine marketing, Carina van Selling asked students to use Copilot to help connect Google Analytics to a website. Barbara Ruppel had Management Accounting students create revision materials, including flashcards. These are different educational uses: coaching, technical assistance and practice.
The results were more complicated than simply giving students access. Lecturer Mieke Nauta described students as less skilled with AI than teachers had assumed. Some questioned its output; others accepted it uncritically. The accounting pass rate rose slightly, but Hanze’s account explicitly says it was difficult to establish whether the AI study support caused that improvement.
A 2026 exploratory study by Hanze researchers Eelco Braad, Mieke Nauta and Jascha Tuinier adds detail about student behaviour. Using interviews, diaries, focus groups and questionnaires, it found students using generative AI both to produce work and to obtain explanations, exercises and learning materials. The repository identifies the available manuscript as a preprint that had not undergone peer review. It offers insight into particular learning practices, rather than a reliable percentage of all Groningen students using AI.
The distinction between assistance and substitution becomes especially consequential when sources are involved. On 12 August 2026, the Raad van State overturned a fraud sanction against a RUG Marketing Management master’s student. His thesis work had been invalidated, and he had been excluded from the thesis course for a year after problems were found in its bibliography. He said AI had scrambled an existing reference list while helping organise it.
The court examined that explanation. Eight of eleven disputed references matched existing sources in an earlier version and a corrected bibliography; it traced the other three to relevant scientific articles. It found that fraud had not been established beyond reasonable doubt and ordered the thesis to be assessed. Carelessness could still affect the grade. The ruling turned on the evidence in this case, including earlier drafts, rather than granting general permission to submit unchecked AI references.
There is another easily missed boundary: being allowed to read material does not necessarily mean being allowed to upload it to a chatbot. RUG’s Mistral guidance says subscription journal articles require permission under the publisher’s licence; open-access material must also be used according to its licence. Its broader AI guidance tells users to protect personal and organisational information. That matters for internship documents, interview transcripts and research data as much as for essays.
The institutions’ longer-term plans point towards further integration. RUG’s 2026–2031 institutional plan names AI in teaching, research and university operations, alongside an ambition to reduce dependence on major technology companies. Hanze’s new strategic plan likewise gives AI a place in education and its own working practices. These are stated ambitions, rather than evidence that every course or administrative process has already changed.
A Hanze research project running throughout 2026 is investigating AI as a learning partner for entrepreneurship students, including whether it can support reflection and students’ ability to manage their own learning. RUG is also training educators: its 2026 AI in Education summer-school programme focused on redesigning teaching activities and assessment while addressing overreliance and the loss of skills. The work involves changing what teachers ask students to do, as well as choosing which software to use.
AI is becoming course content in its own right, too. Hanze’s 2026–2027 catalogue lists a five-credit module on AI in healthcare for Medical Imaging and Radiation Therapy students. It covers applications in diagnosis and treatment, generative AI in patient communication and documentation, and questions about privacy, bias and responsibility. Learning to judge AI’s limitations is part of the professional skill being taught.
A useful personal test is to close the chatbot and explain your argument, calculation or design decision yourself. If that is difficult, bring that exact gap to your next tutorial. It gives your lecturer something specific to help you learn.
