AI is already in classrooms, but when students can get answers to everything with one click, what happens to learning? This week's news suggests that the guardrails are missing, and the struggle is disappearing.
Here is this week’s newsletter on artificial intelligence in the education system, where concerns about AI in teaching once again dominate.
On TV Syd, a father warns against the use of SkoleGPT in his son’s 5th-grade class. He believes, that AI can hamper children’s development if it is introduced too early. Danmarks Lærerforening (the Danish Union of Teachers) shares the concern and calls for a national strategy, while the Minister for Children and Education, Mattias Tesfaye, has launched a review together with Digitaliseringsministeriet (the Danish Ministry of Digitalisation).
At upper secondary schools (gymnasiums), senior teacher Mikael Busch finds that students submit polished assignments without genuine learning. He argues that rote learning and repetition are still necessary methods, especially in foreign-language teaching, and warns against treating the use of AI as a competence in its own right.
At universities, a similar picture is emerging. In an op-ed in Information, research assistant Rikke Abraham Callesen describes how students use ChatGPT without understanding the technology. She criticises that Københavns Universitet (University of Copenhagen) has allowed AI in exam assignments without ensuring guidance, strategy or continuing professional development.
The question of clear frameworks is also the focal point of a new pilot. From next school year, over 40 upper secondary schools (gymnasiums) will take part in a trial of a revised Informatics C curriculum, which emphasises algorithmic understanding, machine learning and ethical considerations. At the same time, Datatilsynet (the Danish Data Protection Authority) has released a podcast about children and AI with concrete advice for parents and educators.
In Berlingske, Erik Henz Kjeldsen and Ole Kaag Mølgaard from Lex warn that AI-based answer systems are about to replace traditional search engines. The problem is that they deliver confident answers without sources, creating a new form of misinformation. If pupils do not learn to think critically and the technology simultaneously undermines source criticism, we face a double problem.
This week’s stories point in the same direction. AI is already in classrooms and universities, but the necessary frameworks are still lacking. In this week’s commentary, I write about how we can distinguish between the productive struggle that fosters learning and the friction that simply gets in the way.
Happy reading.
TV Syd reports that several Danish primary and lower secondary schools (folkeskole) have begun using SkoleGPT, but that this has prompted concern among both parents and professionals.
Among the critics is senior teacher and father of a 5th-grade pupil, Thorsten Westphal, who believes that the use of AI such as SkoleGPT can hamper children’s development, especially if introduced too early. He calls for clear frameworks and suggests that older pupils should be the first to use the technology in teaching.
The concern is shared by Regitze Flannov from Danmarks Lærerforening, who points to a lack of a national strategy and insufficient continuing professional development for teachers. She emphasises, however, that pupils must learn to understand and use AI responsibly and critically as part of their digital literacy.
The Minister for Children and Education, Mattias Tesfaye, acknowledges the need for political action and has launched an inquiry in collaboration with Digitaliseringsministeriet.
Kommunernes Landsforening (Local Government Denmark) has declined to comment, as collective bargaining negotiations are ongoing.
Upper secondary school teacher and lecturer in German, Mikael Busch, expresses concern in an opinion piece about how AI affects students’ learning. Drawing on his own experience, he questions whether students are actually acquiring knowledge when assignments are increasingly completed with help from AI tools such as ChatGPT.
He warns against treating the use of AI as a competence in its own right, because it is primarily the most able students who know how to use the technology meaningfully. For most students, acquisition still depends on repetition, rote learning and practice.
In foreign-language teaching, he particularly highlights the need to learn vocabulary and grammar by heart. Here he argues that rote learning should be reintroduced as a method of genuine acquisition that can strengthen both character and the capacity to learn.
In an op-ed in Information, research assistant Rikke Abraham Callesen criticises the fact that Københavns Universitet (University of Copenhagen) has allowed the use of generative AI in written exam assignments without adequate support, guidance or strategy. In her teaching on the anthropology programme, she finds that students make extensive use of AI, especially ChatGPT, to formulate research questions, understand texts and get feedback. But this happens without them being equipped to understand the technology or the consequences of its use.
She points out that the university has introduced the option without ensuring that teachers and examiners understand the technology or are able to assess its use. The result is that many students feel uncertain and are left to experiment with AI as a new tool on their own, which weakens their chances of achieving academic understanding and independence.
Rikke Abraham Callesen calls for both continuing professional development for teachers and mandatory teaching for students in the critical and constructive use of AI. She argues that the university owes it to both staff and students to take responsibility rather than allowing developments to unfold without direction.
Rikke Abraham Callesen
From the 2026/27 school year, over 40 gymnasiums will take part in a pilot with a revised curriculum for Informatics C. The trial focuses on strengthening students’ understanding of core digital technologies and the associated ethical considerations. The curriculum covers topics such as algorithmic understanding, design processes, debugging, machine learning, and the use of big data and AI in a societal perspective.
The examination format in Informatics C will remain unchanged during the trial period. Participating schools must attend briefing days and contribute to the evaluation of the trial, which runs through the 2028/29 school year.
The experiences from the trial will form the basis for a possible adjustment of the subject in the longer term.
A new podcast from Datatilsynet (the Danish Data Protection Agency) spotlights how children and young people encounter AI in their digital everyday lives. Lawyer Johan Daugaard and communications consultant Hisar Sindi shed light on how AI features in social media, games and chatbots, and what it means for the protection of personal data.
The podcast also offers practical advice for parents and educators on how to build good digital habits and navigate safely in an AI-saturated online world.
In an opinion piece in Berlingske, Erik Henz Kjeldsen and Ole Kaag Mølgaard from Lex, Danmarks Nationalleksikon (Denmark’s National Encyclopaedia), warn that we are losing access to reliable sources as AI becomes the dominant conduit of information. They point out that large language models such as ChatGPT, Gemini and Copilot often deliver incorrect answers without citing sources, and that the traditional search engine is being replaced by what they call “answer machines”.
This development constitutes a new type of misinformation. AI generates content without intention or source, yet it can still be misleading because it sounds plausible. Unlike classic fake news with a deliberate author, AI errors are a side effect of the systems’ design and their lack of understanding of content.
At the same time, the volume has exploded, and misinformation can now be delivered personally and tailored to each user’s preconceptions. This weakens both our critical evaluation of sources and the shared factual basis for public debate.
The authors describe how AI’s illusion of authority — that is, the combination of confident, polite and grammatically correct answers — risks undermining trust in knowledge, experts and shared truths.
If we can no longer rely on the internet as a guide to knowledge, the foundation for democratic debate and public enlightenment disappears.
Erik Henz Kjeldsen
Reading this week’s news, a familiar picture emerges. AI meets strong resistance across all parts of the education system. From the father who fears that SkoleGPT will make his son less able, to Mikael Busch, who sees students submitting polished assignments without having learned anything.
In general, the commercial side of AI is not designed for use in teaching: it is built to deliver results quickly, not to sustain the resistance that learning requires. All these concerns are entirely real. When students can get an answer to everything with a single click, it is natural to ask what actually happens to learning.
Last week I gave a presentation at ATV’s conference on AI in education (the Danish Academy of Technical Sciences), where I focused on the difference between productive and unproductive friction. I believe we can use this distinction to design teaching that lets AI remove the drudgery but preserves the thinking.
AI has the potential to remove both, and this is where it becomes difficult in teaching.
Take foreign-language teaching, where the toil of learning vocabulary is not a barrier to learning. It is the process itself, and that kind of productive friction is what we must hold on to. But that requires us to explain to students why the effort is valuable. Not just to ban the shortcut, but to justify the detour.
And in all of this, AI can remove all resistance, but then we risk the material being explored superficially and losing its value. The student gets answers without engaging, and this is precisely where motivation disappears.
Perhaps that is why the question of why one should learn something looms so large right now. We need a better answer than ‘because you have to’.
