Introduction

The advent of generative AI fundamentally challenges how universities design learning, support students, assess achievement and safeguard academic integrity. It unsettles two assumptions on which much of higher education still depends: that the work students submit is a reliable indication of what they have learned, and that achievement can be attributed with sufficient confidence to the individual student. When AI can generate fluent text, plausible arguments, code, summaries, reflections, or analyses within seconds, learning processes can be bypassed and evidence of achievement can become difficult to interpret. Educational administrators, program leaders, teachers, educational designers, quality assurance staff and examination boards are therefore confronted with the same underlying question: how can higher education respond when generative AI may invalidate familiar processes of learning, assessment and attribution? This Companion does not treat generative AI as a temporary problem to solve, but as a persistent condition that requires deliberate educational design choices.

At the same time, generative AI also presents opportunities for higher education. It offers the prospect of more personalized learning, in which students receive adaptive feedback and support tailored to their individual pace and needs (Zawacki-Richter et al., 2019). AI tutoring systems, simulation tools and intelligent study assistants can extend the reach of teaching beyond contact hours and may help address accessibility barriers for students with language difficulties or learning differences (Perring, 2025). For researchers and teachers, generative AI can support literature review, writing processes and new methodological possibilities (Kusumegi et al., 2025). These developments invite educators to reconsider how time, attention, and learning activities are distributed in teaching, and how carefully designed learning environments might create more space for critical reasoning, ethical reflection and collaborative problem-solving.

Universities are responding to this situation in different ways. As Pearson (2025) describes, some institutions are moving quickly toward a broad adoption of generative AI, while others emphasize restriction or prohibition, and many are searching for a middle ground. Some are even wondering whether universities themselves risk becoming obsolete in an AI-rich knowledge environment (Clune, 2025). This Teacher Companion takes that institutional variety seriously, but starts from the position that the most useful question is not about banning or embracing generative AI, but rather how programs and teachers can make informed, proportionate and educationally justifiable choices.

Generative AI is also connected to broader societal concerns, including the growing influence of big tech, environmental impact, privacy, security and the public values at stake in educational technology. These issues are addressed more extensively in the student version of the Companion. This Teacher Companion treats them as part of the context in which educational decisions about generative AI must be made. Readers are therefore encouraged to consult the Student Companion alongside this Teacher Companion, especially when discussing responsible use with students.

Universities and staff also have specific rules to follow. These follow from both legal regulations and university or faculty policies. In this Companion, we will discuss these as far as possible based on the VU Amsterdam context. But we emphasize that this book must be mainly be regarded as a textbook, not a policy document.

Finally, there is a big impact on research. Generative AI influences the way data can be collected and processed (Kusumegi et al., 2025). We will not explicitly dive into that in this Companion.

All in all, knowledge about generative AI, the impact on learning, teaching and assessment, and the process, procedures, and culture on how to handle it effectively, must be on the radar of teachers and program leaders. It calls for growing AI-literate students and AI-literate academic staff.

#1.1The VU framework for generative AI in education

The leading document for faculties, program directors and teachers and the further institutional support units is the framework for Generative AI in Education. This document was approved by the Executive Board in April 2025 and is VU policy. It is general by design. It sets the boundaries within which generative AI may be used in teaching and leaves faculties to write the rules that apply to a specific program. The document states that it will be revised regularly, so check that you are reading the current version.

The Framework describes that VU Amsterdam has a duty of care to provide a safe digital learning environment, and the framework is cautious for that reason. It names three risks: data management, the quality of education, and the effect of AI use on students’ academic skills. It accepts that there are opportunities and states that much more research is needed before they can be assumed. What it asks of you is a judgment about whether AI use serves the learning outcomes of your own course.

Five of its provisions apply directly to your teaching:

• You state in the study guide, the syllabus or Canvas whether and how generative AI may be used. The learning outcomes decide, and you are expected to have considered the added value beforehand.
• If AI literacy is not among your learning outcomes, students must be able to complete the course without generative AI.
• You cannot require a tool for which the VU holds no license.
• Nobody enters personal or confidential data into an AI tool. This applies to students, teachers and support staff alike.
• Where students may use generative AI, discuss it with them. Writing the rules down is the minimum, and a conversation is what makes the expectations clear, above all about assessment.

The framework also makes four recommendations to teachers, examiners and program directors. Build your own AI literacy. Review whether your intended learning outcomes still fit a field that AI is changing. Choose an assessment form through constructive alignment and decide there whether AI use is allowed. Follow the student’s process through formative dialogue, with regular feedback moments in large writing assignments. That last recommendation does double duty, because knowing how a student writes is what makes an unexplained change in their work visible.

On fraud the framework is short. Three situations count: using AI where you did not permit it, presenting AI output as one’s own work, and being incomplete about permitted AI use. The sanctions are set out in the Rules and Guidelines of your examination board. One condition holds across every application of AI in education, which the framework calls human-in-the-loop: a person always checks.

The policy also includes that experimentation with generative AI in education should take place in a structured and responsible manner. For teachers, it is expected that they do not experiment with using and building apps in an uncontrolled manner. Experiments and pilots are not ad hoc activities, but should be registered and conducted via the VU Education Lab, where expertise, support, and knowledge sharing are coordinated. This ensures that experiences, risks, and best practices are systematically collected and inform broader institutional learning and decision-making.

#1.2VU-UvA Taskforce on AI Literacy

The joint VU–UvA Taskforce on AI in education positions generative AI not as a problem to be solved, but as a structural development that requires deliberate educational choices. Its recommendations emphasize three parallel responsibilities for universities: to provide clear institutional guidance, to safeguard values such as privacy, security and fairness, and to create space for experimentation and informed use in teaching and learning.

The Taskforce also developed an advice document about AI Literacy. First, we note that there are a multitude of frameworks currently developed in various places. We will not dive into them here. For the sake of simplicity we will use this in this Companion, the UvA-VU Taskforce advice as a starting point as the set of learning objectives they formulated are quite straightforward. Second, later on this Companion, we will expand and elaborate more on that set of learning objectives.

According to that document: “Teachers who help to develop AI literacy among students need, at the very least, a basic understanding of AI, regardless of whether they use AI themselves. Teachers need support to move in that direction and to develop their own AI literacy. VU CTL and UvA TLC are already working on programs for teacher professional development focused on AI in education.”

The Taskforce divides the specific learning objectives for students, but also staff, into three clear, categories (based on the Dublin Descriptors). These objectives are designed not only to equip individuals with technical skills, but above all to educate them as critical and responsible academic users:

What – Knowledge and understanding (The theoretical basis)

This learning objective focuses on understanding the underlying technology, so that learners can assess its capabilities and limitations. This entails that learners must:

• Understand how generative AI and LLMs (Large Language Models) work. This means they know that these systems make statistical predictions based on patterns, and do not actually ‘think’ or understand facts.
• Have an understanding of algorithms and data structures, including how the input of training data directly influences the output and can lead to built-in biases.

How – Understanding and Application (Practical Skills)

This concerns the functional ability to deploy AI effectively and methodically during one’s study or work:

• Targeted prompting: The ability to formulate effective, detailed instructions (prompting) to elicit high-quality and usable results from an AI system.
• Information evaluation: The ability to critically check and validate AI-generated output for factual accuracy, reliability and any ‘hallucinations’.
• Information selection: Knowing which data sources are suitable and safe for analysis using AI systems.

Why – Ethics and judgment (Critical reflection)

This is the academic core, in which learners learn to weigh up the broader impact of the technology:

• Ethical awareness: Understanding the risks posed by AI in terms of privacy, data security, intellectual property and copyright.
• Social and environmental impact: Being able to reflect critically on the broader consequences of AI, such as social inequality, filter bubbles, and the significant ecological footprint (the electricity and water consumption of data centers) of large-scale AI use.

The combination of the AI Literacy Companion for Students and this AI Literacy Companion for Teachers tries to fill in these learning objectives with substantial content and exercises for deepening understanding.

For the teacher Companion, additional learning objectives are of importance, namely how the use of AI impacts education, learning and educational design. But also what can and cannot be done from a formal and legal perspective (GDPR, AI-Act). Teachers need specialized technological, pedagogical, and ethical knowledge (TPACK) to design curricula that either mitigate risks induced by AI or reap the opportunities provided by AI. This includes the ability to adapt curricula as AI evolves You'll read more about this in Chapter 3.

#1.3From the VU framework to faculty policy

The framework does not regulate what happens in a classroom. What it does is distribute responsibility, and the faculty is the level at which that responsibility turns into a rule you have to follow. The framework sets out the chain as follows.

The Executive Board is responsible, from an institutional perspective, for an organization that handles AI responsibly. It commissioned the VU-wide Artificial Intelligence Competence Network for that reason.

The faculty board is responsible for translating VU AI education policy into faculty policy.

The examination board draws up the Rules and Guidelines, and records in them when AI use counts as fraud or plagiarism. It can advise the examiner on whether the text in the study guide is in line with those rules.

The teacher or examiner states in the study guide, the syllabus or Canvas whether and how (gen)AI use is permitted, and can draw on the technical, legal and didactic expertise of the Competence Network.

The student is responsible for the content of the work submitted.

Four faculties have filled this in so far, and three of them are described here. The Faculty of Science was first, in April 2025, with a four-part guideline that it calls a living document. SBE followed in October 2025 with a policy that is, in its own words, heavily based on the one from Science. The Faculty of Medicine went its own way in November 2025, building on the AI-GO framework for AI literacy and on the AI Act; its vision has been approved and its policy was still in draft at the time of writing.

That faculties copy from each other is intended rather than accidental. The framework names the SBE approach explicitly as an example that other faculties can adopt, so that teachers do not all have to invent the wheel themselves.

What this means in practice: before you write your own rules for a course, find out whether your faculty has already decided the question for you. Much of what follows in this Companion is a choice at course level only where the faculty has left it open.