FAQ EduGenAI

You can request access to the EduGenAI platform via the registration form on the website:

• Registration form for staff
• Registration form for students

Once you have access to EduGenAI 1, you will automatically have access to EduGenAI 2 as well. You do not therefore need to submit a new application. If access to EduGenAI 2 is not working, this should be resolved by logging in to EduGenAI 2 once using the same email address as for EduGenAI 1.

Once your account has been activated, you can log in straight away to one of the two EduGenAI platforms: 

• EduGenAI 1
• EduGenAI 2

Once you have access to EduGenAI 1, you will automatically have access to EduGenAI 2 as well. You do not need to reapply. If you are unable to access EduGenAI 2, this should be resolved by logging in to EduGenAI 2 once using the same email address as for EduGenAI 1.

Once you have completed your application via the registration form, we will grant you access within one week. This means that (normally) you will receive an email within a week, which you can use to activate your account. You will therefore need to take action yourself to activate it.

Your invitation (to activate an account) remains valid for 1 month after receipt. Once this period has expired, you will need to complete the MS Forms form again to submit a new application. You can apply for access via the registration form on the website: 

• Registration form for staff
• Registration form for students

You log in using your UU account details. Go to the login page, select UU as your educational institution, and log in with the email address and password linked to your UU account.

If you experience technical issues, such as not having an account or being unable to access the platform, please send an email to: aionderwijs@uu.nl.

EduGenAI is free to use for all students, lecturers and staff at UU (including UMCU insofar as it relates to teaching). 

The platform was developed by SURF for further education: vocational education (MBO), higher professional education (HBO) and university education (WO), with funding from NPuls. Around 100 different educational institutions are taking part in the EduGenAI pilot, which runs until 1 January 2027. After that, EduGenAI will become a formal service provided by SURF. 

The further development of the platform and the prioritisation of new features will take place in collaboration with the participating educational institutions.

Not at present. The contracts with SURF stipulate that EduGenAI 1 and 2 are to be used exclusively for educational purposes. Furthermore, EduGenAI and its functionalities are not necessarily well-suited or developed for this target group. There is also no AI compliance oversight for EduGenAI from an operational perspective. For example, some HR applications fall under high-risk AI in accordance with the EU AI Act, but we do not (yet) monitor this. The AI platform is funded with funds from NPuls, which focuses on education. 

Our current pilot contract with SURF runs until 1 January 2027. After that, EduGenAI will become a formal service. We will decide at the end of 2026 whether to continue with EduGenAI, but that seems likely as we are also paying for SURF’s services. 

If EduGenAI 1 or EduGenAI 2 experiences a disruption, you (as the account holder) will receive an email notification as soon as possible. In this regard, we are dependent on the information and speed of reporting provided by our supplier, SURF/NPuls.

A large language model, also known as an LLM (Large Language Model), is an AI model that has been trained on large amounts of text. It processes your input and generates an appropriate response based on learned language patterns.

A hallucination is a convincing-sounding but incorrect or fabricated answer that an AI sometimes provides. You should therefore always critically check AI output, especially if you require factual information. If you’d like to learn more about AI, take an online AI Literacy course (AI module). 

A prompt is the text instruction, the context, the question or task that you give to the AI. The clearer and more specific the prompt, the better the AI understands what you want and the more useful the output will be.

A persona is a character or role that you assign to an AI chatbot, such as a study adviser or subject teacher. This allows you to determine the tone, expertise and approach the language model uses when providing answers.

An artefact is a separate workspace where you can create and edit content alongside the AI, such as text, code or visualisations. You can customise the content yourself or instruct the AI model to make specific changes.

Go to Settings > General. There, you can choose between light mode, dark mode or automatic (based on your device settings), and between Dutch or English as the interface language. Please note: the interface language does not determine the language of the AI’s responses; these follow the language of your question.

Via Settings > Personalisation, you can set fixed instructions for each chat, for example regarding writing style, tone (with or without emojis), preferred language, or your own role (e.g. ‘I am a history teacher’). This does not affect the behaviour of personas you have created separately.

Under Settings > General, you’ll find the ‘Delete all chats’ button (confirmation required; cannot be undone). Under Settings > Account, you can permanently delete your account; this will also delete all your chats, personas, prompts, projects and extensions. This action cannot be undone either. Please bear in mind the mandatory retention period for prompt logs (see the AI index instructions).

In the left-hand pane of EduGenAI, you’ll find the menu bar with shortcuts to your chats, personas, projects, groups, prompts and extensions, amongst other things. In the centre is the chat field where you can ask questions and add files or extra features. At the bottom left, you’ll find the settings and support & feedback options. Please note: you may only use the extension option after explicit permission has been granted. 

Click on ‘New chat’ in the menu bar to start a blank conversation.

Click on the plus icon (+) in the chat box and select ‘Upload file’. Supported formats include PDF, JPEG, DOCX, XLS, PPTX and CSV. The AI can then summarise, analyse or search the content. Please note: large, unusual or protected files, and documents with more than, say, 100 pages, may not be processed in full. The maximum file size you can upload is 100 MB. Ensure your files remain within this limit to upload and share them successfully. 

Yes. Click on ‘More options’ (the three dots at the bottom left of the prompt bar) and select ‘Create an image’. Describe what you want to see in as much detail as possible (style, subject, atmosphere) to get the best result.

Study Mode is a special setting in which the AI acts as a digital tutor: instead of answering straight away, the AI first asks questions, provides step-by-step explanations and feedback, and corrects where necessary. Useful for explaining difficult concepts, creating practice questions, summaries, step-by-step guides, simulating a test, or providing feedback on an assignment or essay. 

Activate study mode by tapping the second icon (doctoral cap) from the left in the prompt bar. 

Enable this option via ‘More options’ (the three dots at the bottom left of the prompt bar) and start your question, for example, with ‘Search the internet for…’. The AI will then consult up-to-date online sources on the internet, insofar as these are publicly accessible. 

Activate ‘Create an artefact’ via ‘More options’ and give a clear instruction. The AI generates an initial version which you can then refine: select sections of text or code and let the AI amend them, or edit them yourself. Save the initial version using the save icon in the top right-hand corner; subsequent changes are automatically saved as a new version, and older versions remain accessible. You can download the final version via ‘Download as’.

You may not do so without prior authorisation, as potential privacy, security and AI compliance issues need to be verified. For example, a data processing agreement must be in place with the supplier of the system or tool being linked to, etc., and not all data is accessible from other data sources – for instance, for GDPR reasons. You can, however, submit a request via aionderwijs@uu.nl.

Yes. Upload a .CSV file and ask the AI, for example, to inspect columns, create summaries, carry out comparisons or generate graphs (line or bar charts). Always check the results yourself and do not let the AI perform tasks that you were supposed to do yourself. Do not carry out data analyses on data containing personal data such as names or student numbers, etc. 

Under ‘Prompts’, you can save frequently used prompts and reuse them later within a chat session. 

A strong prompt usually consists of three parts:

• Context – what do you want to achieve, or who are you?
• Instruction – what exactly should the AI do?
• Specifications – details such as tone, length, format, language level or target audience

Not sure where to start? Ask the AI itself: ‘Help me write a good prompt for [your goal].’ Or follow the AI module on prompt engineering for students or teachers (available at Modules and courses).

In the chat window, you can select a language model yourself before asking a question. Models differ in quality depending on the subject area (technical versus creative), speed, energy consumption and proficiency in Dutch. Choose carefully: a lightweight model is often sufficient for simple questions, whilst more complex tasks benefit from a heavier model.

Both commercial models (accessed via Azure) and open-source models (hosted via SURF’s AI hub) are offered. The exact, up-to-date list of models varies and can be found in the language model selection screen within the chat itself by clicking on the model (for example, gpt-5.1).

For simple questions, it is recommended to choose a lightweight model, such as GPT-5 Nano or Mistral Small; this is faster and consumes less energy and water than a heavyweight model.

A lightweight AI model generally has fewer parameters and uses less computing power per task, expressed in FLOPs (Floating Point Operations). It is therefore usually faster, cheaper and more energy-efficient. A heavy model has more parameters and/or uses more FLOPs and is often more suitable for complex analyses, reasoning tasks and programming tasks. For simple tasks, such as summarising or rewriting, a lightweight model is usually sufficient.

Language models consume energy and water. For simple questions, a lightweight (small) model (such as GPT-5 Nano or Mistral Small) is both faster and more sustainable than a heavier model. 

A persona is a bespoke AI assistant that responds from a specific role, area of expertise or perspective. To create a persona, click on ‘Personas’ in the left-hand menu, then click on ‘New Persona’ at the top left of the screen. When creating a persona, you give it a name and description, choose a language model and a communication style (Creative, Balanced, Precise or a manual configuration using TopP and Temperature), describe the persona’s behaviour, and can optionally add documents for extra context. 

Type ‘@’ in the chat field to call up a previously created persona. From that point on, that persona will answer your questions. You can switch the persona off again by clicking on the red cross behind the persona’s name.

Bear in mind that a persona is not a human being and has no knowledge, consciousness or judgement of its own. A persona is merely a set of instructions that determines the role, tone or expertise from which the AI responds. You should therefore always scrutinise the answers carefully.

You can share personas (and prompts) with other users within your UU via ‘groups’.

That’s a valid question, as AI isn’t a human and therefore isn’t a ‘person’ either. The word ‘bot’ or ‘chatbot’ is actually a better term, but SURF has opted for the term ‘persona’. 

Projects are a digital space where you can group chats, prompts and personas relating to the same task, research or learning process. To create a project, click on ‘Projects’ in the left-hand menu, then click ‘Add Project’ in the top right-hand corner of the screen. In the next window, give the project a name and, if necessary, set project-specific instructions that determine how the AI responds within that project. 

You cannot add other users to a project. 

Groups are designed for collaborating with multiple users and sharing personas and prompts, for example for research, a joint presentation or a team. Projects are primarily aimed at organising your own chats and material; groups are for sharing them with others. Extensions (API and MCP) must not be shared without prior permission (aionderwijs@uu.nl).

Extensions are additional tools that give the AI access to functions outside the chat environment, such as retrieving information from external databases. Extensions utilise APIs (in EduGenAI 1) and MCP (in EduGenAI 2) and are therefore primarily intended for technically savvy users. 

The use of these extensions is not automatically permitted. As they involve connecting to and exchanging data with other software, AI tools or data sources, and no agreements have been made with the relevant (unsafe) tools regarding the University of Utrecht (UU) or permission to use data from UU systems, this cannot be done without authorisation. If you wish to use this functionality, please send an email to aionderwijs@uu.nl.

The data are stored and processed on SURF servers in the Netherlands or Germany and therefore remain within the EEA. This complies with the agreements on European data storage and limits transfers to third countries.

For commercial models used within EduGenAI, the following applies: a prompt is sent to the model, the response is returned, and the data is not stored by the model providers themselves. Data processing agreements have been put in place for this purpose via the Azure Data Processing Agreement (DPA).

Do not enter any confidential data or personal data, and do not upload any files containing such data. Due to contractual restrictions, EduGenAI is not currently intended for this purpose. Furthermore, EduGenAI is in the pilot phase until 1 January 2027, which means there is a risk of data loss in the event of any platform failures. You should therefore only use information that is neither confidential nor personal.

No. EduGenAI is built on a ‘privacy-first’ principle: user data belongs to the user, and neither a teacher nor an administrator can view it. If you wish to be transparent about your use of AI, you can easily copy and share a chat conversation. As a student, you must always adhere to your teacher’s instructions and the permitted uses set out in the AI index. 

Yes, this is permitted, provided you have obtained permission to use that material from the owner or publisher. Responsibility for copyright lies with the institution or lecturer, just as it does when using copyrighted material outside of AI tools. For more information, see Copyrights and AI.

No, the AI platform must not be used for scientific research. This is due to contractual restrictions and the potential risk of data loss in the event of platform failures. This applies at least until 1 January 2027.

No. Neither AI browser extensions nor MCP (Model Context Protocol) may be used in conjunction with EduGenAI. This is for security and privacy reasons. As this involves establishing connections and exchanging data with other software, AI tools or data sources, and no agreements have been made with the relevant (insecure) tools – such as UU – regarding the use of data from UU systems, this cannot simply be done. If you wish to use this functionality, please send an email to aionderwijs@uu.nl.

No. EduGenAI is not suitable for this purpose. In accordance with the EU AI Act, the supervision of high-risk AI systems is, quite rightly, thorough and intensive due to the risks and potential infringement of human rights. Our contracts with SURF do not permit this use, and UU itself also excludes (for the time being) the use of AI in high-risk use cases within the education sector. For details of which high-risk AI applications apply to education, see High risk AI.

No, that is strictly prohibited. This is a high-risk AI application (and therefore not permitted) and is also undesirable for ethical reasons.

No. EduGenAI uses generative language models that predict words based on probability, so the model does not ‘know’ whether something is correct. Always use the AI as a support tool, not as the final authority, and critically review the output.

A hallucination is an answer that sounds convincing but is incorrect or fabricated. Be particularly alert to this when dealing with factual or specialised information, and verify important claims against a reliable source.

Use AI to explore topics in greater depth, practise and compare, not to have every question answered automatically. If you rely on AI for everything, you will develop less as a learner. Use the UU AI Index as a guide when, as a student, you have to complete an assignment with or without AI, and if in doubt, ask your lecturer.

Both are similar applications of generative AI in education with the same ‘look and feel’. EduGenAI 1 runs entirely on Microsoft Azure. EduGenAI 2 is not a further development but a completely new, modular approach based on open-source components (including LibreChat), with the aim of providing greater flexibility and digital sovereignty.

The AI hub is where various open-source AI language models are hosted and managed locally in SURF’s data centre, and made available via an API. EduGenAI is linked to the AI hub, making these open-source models available to end users as well. Commercial models are accessed separately via Azure. The AI hub cannot be accessed independently. 

Yes, the platform is currently in the pilot phase. This means that not all features are yet working 100% flawlessly and that the platform does not yet fully meet all educational needs. Please therefore use the platform bearing this in mind, and report any issues or provide feedback so that it can be further improved. The pilot will run until 1 January 2027. Around that date, a decision will be made as to whether EduGenAI will continue as a regular SURF service. During the 2026–2027 academic year, UU AI Chat (see 9.4) will be the AI platform for the education sector. EduGenAI will continue because we expect to opt for EduGenAI 1 or 2 from the 2027–2028 academic year onwards. 

UU AI Chat is the AI chat environment managed by UU; EduGenAI is a national pilot platform run by Npuls/SURF, in which several vocational, higher professional and university-level institutions are collaborating. Both are free and operate on a ‘privacy-first’ principle (no one can view your conversations), but UUChat keeps data within UU’s own environment, whilst EduGenAI data is stored at SURF. EduGenAI also offers a choice of several language models, including open-source models via the national AI hub.

SolisChat is an AI education platform that we at UU are developing from scratch with our own UU students, from the perspective of lecturers and students and based on our UU Responsible AI principles and digital autonomy. With regard to the effective, ethical and pedagogical use of AI in education (for example, through AI tutors per course), SolisChat better meets certain needs within education than EduGenAI. SolisChat is currently under development, and we are in the process of scaling it up and testing it. If you, as a lecturer, would like to take part in the testing, please sign up for the AI lecturer test pool.

Commercial AI tools such as ChatGPT, Claude and Mistral are powerful applications and may be suitable for many purposes. However, EduGenAI has been developed as a public educational resource that meets the needs of the education sector and our public values, and partly replaces these commercial tools. The platform offers a choice of multiple AI models, including open-source models, and provides greater control over privacy, data processing and the further development of functionalities for education. Furthermore, a multi-model platform prevents dependence on a single supplier (so-called ‘vendor lock-in’) and allows users to choose for themselves which model best suits their needs (use case). 

On UU’s AI policy page, you’ll find various modules and courses for students, lecturers and staff. 

Via the ‘Support and feedback’ icon at the bottom left of the EduGenAI menu bar, you can access the manual, the e-learning modules, the privacy statement, the disclaimer and the feedback module. Please ensure you are following a UU AI module and not one from another educational institution. This is because other institutions may have different AI policies and may also have made different strategic choices regarding the use of AI tools, which may differ from those of UU and our UU AI Ethical code of conduct.

Additional support is available, specifically for lecturers and teaching assistants. For further information, please contact the Centre for Academic Teaching and Learning (CAT) by email: generatieve-ai@uu.nl. Students’ enquiries should also be submitted to CAT via their lecturers. 

If you have any questions or are experiencing problems with the operation of the EduGenAI platform, and your colleagues are unable to help or do not have a solution, please send an email to: aionderwijs@uu.nl. If you need help with teaching methods or setting up the platform, please contact CAT via generatieve-ai@uu.nl.

If you have any questions, suggestions for improvement or additions to this FAQ list, please send an email to: aionderwijs@uu.nl. Thank you very much in advance!