This week's news covers a new AI Driver's License for educators, a Norwegian municipality caught in a ChatGPT scandal, and the risk of sponsored knowledge in future language models.
Here is an overview of last week’s most important news on artificial intelligence in education.
Center for IT i Undervisningen (CIU, Centre for IT in Education) has announced a new AI driving licence for teachers on EUD (vocational education and training) and AMU (adult vocational training). The purpose of the licence is to offer teachers a targeted course so they can use AI in their teaching. I reflect on this initiative in this week’s commentary.
In Norway, the people in Tromsø Kommune could probably have benefited from an AI driving licence. The municipality has landed in trouble after it emerged that sources in a report on school structures had been invented by ChatGPT. It is not news that AI can hallucinate, but what is interesting is that AI conversations in Norway are considered public documents, and the municipality has been asked to release them. Nor is it an isolated case, as a Canadian education report on ethical AI use has been exposed as containing more than 15 fabricated sources (read the story at Ars Technica).
This raises an interesting question:
How would you write to an AI if you knew the conversation could be made public?
Would you be more cautious with personal data, avoid asking the “silly” questions, or perhaps refrain from using AI for certain tasks altogether? This is not just a hypothetical question: as AI is integrated into teaching, it becomes a real consideration we all have to address.
Jeppe Klitgaard Stricker warns in a new article that sponsored content could become a built-in part of AI responses, thereby threatening the technology’s role as a neutral conveyor of information. This poses a significant challenge if we do not know whether a company has influenced a language model to promote its own products, or a politician with sufficient resources has pushed it to write positively about their party. We therefore need independent language models for teaching and research, but the free versions will likely in future be financed through sponsorship deals.
If we are to have independent language models, we also need good training data, and the organisation Tekst & Node has just been given the green light to begin negotiations on copyright.
Amid all this, we must educate students to understand it—and here, human competences are very much needed. An opinion piece on Uddannelsesmonitor argues that human language understanding will become even more important for assessing and assuring the quality of the technology.
At the same time, Nick Ladany describes how AI professors are making their way into universities as artificial teachers that can teach, supervise and adapt to students’ needs. According to the article, human teaching staff will, however, take on a new role in which relationships, guidance and reflection become even more important.
Here are the week’s three key takeaways:
Enjoy the newsletter.
To me, an AI driving licence is exactly the same as an ordinary driving licence. It is not only about operating the technology, but also about knowing the rules of the road—for example ethics, bias and copyright. What matters is the ability to navigate with critical thinking in a complex digital landscape.
Conversely, I do not see the licence as “now every teacher must go out and use AI in teaching”. It is very much about involving and using the technology when it makes sense and adds value in teaching. You can do that when you understand the technology and can therefore ask why—and choose not to use AI when it does not make sense.
Once you have completed the programme, I also believe that, as a teacher, you meet the requirements that AI-forordningen (the EU AI Act) places on everyone who works with AI.
This kind of fundamental and systematic continuing professional development is exactly what there is great demand for out in schools.
A new initiative will equip teachers in the vocational programmes and the adult vocational training programmes (AMU) to teach both in and with AI. Center for It i Undervisningen (CIU) together with a number of vocational committees and continuing education committees is behind a new digital learning platform, AI kørekortet.
The aim is to give teachers a solid understanding of AI so that they can both use the technology as a pedagogical tool and teach students and course participants how to use it.
AI kørekortet combines CIU’s digital and didactic expertise with the committees’ industry knowledge so the material is both academically strong and practice-oriented. AI kørekortet will give teachers knowledge of, among other things, prompt engineering, ethics, a critical approach and the didactic use of AI. The learning pathways consist of videos, interactive tests and chatbots and can be used individually or as part of schools’ competence development.
The project collaborates with Styrelsen for Undervisning og Kvalitet (STUK, the Danish Agency for Education and Quality) and several knowledge centres, and we at Viden.AI are also involved. The licence will be launched before the end of the year and later expanded with subject-specific modules so teachers can tailor the learning to their sector and subject area.
Read more here: https://videnscenterportalen.dk/ciu/ai-korekort/
A case in Norway shows that the use of ChatGPT in Tromsø kommune has led to a principled decision on transparency. In March 2025 it emerged that a report on school structure contained references to research that did not exist at all. ChatGPT had hallucinated sources. When journalists requested access to the chat logs, the municipality refused and called them electronic traces. In September, Statsforvalteren rejected this justification and determined that AI conversations are public documents. The review also showed that large parts of the report were directly AI-generated without quality assurance.
This decision underlines that the rules on transparency and documentation also apply to AI-generated material. The question is how we would react in Denmark if a municipality had used AI in the same way, and whether we would gain access to the material.
Rune Ytreberg
In a new and thought-provoking blog post, Jeppe Klitgaard Stricker writes that OpenAI will likely introduce advertising in ChatGPT results in the coming years. The aim is to earn billions from advertising. This could change how AI is used in teaching and research, where it may lose its position as a neutral conveyor of knowledge and instead be driven by commercial interests.
He points out that the problem is not just banners or pop-up ads, but that the entire content of responses can be shaped by paying actors without users’ knowledge. This could mean, for example, that climate information highlights energy companies’ perspectives, or that career guidance favours certain employers. Unlike traditional search engines, where ads are clearly labelled, AI will provide cohesive answers in which advertising and information are blended.
For universities, this raises the question of whether they can continue to use commercial AI platforms in teaching, or whether they must invest in independent and ad-free alternatives. At the same time, Jeppe Klitgaard Stricker warns of the risk of a two-tier education system, where only well-resourced institutions can shield themselves from advertising influence.
Jeppe Klitgaard Stricker
Tekst & Node has been given the green light to begin negotiations on collective agreements concerning the use of copyright-protected texts, images and sheet music in AI tools. At a board meeting on 23 September, negotiating mandates were granted for both the education and business areas, which means the organisation can now open dialogue with clients and partners. The goal is to find solutions that both safeguard rights holders and at the same time give users clear and lawful frameworks for making use of AI in their work.
Camilla Tollstorff ArdilsøIn an opinion piece by Mikkel Haarder (Dansk Industri, the Confederation of Danish Industry), Mads Eriksen Storm (Dansk Erhverv, the Danish Chamber of Commerce), Lars Bo Henriksen (Lederne, the Danish Association of Managers and Executives) and Maja Bødtcher-Hansen (Danske Gymnasier, the Danish association of upper secondary schools), it is highlighted that Denmark faces major challenges in the field of languages. Although many students have taken German or French for several years, only a few feel able to hold a meaningful conversation. At the same time, fewer young people choose language subjects at a high level or continue with language studies at university.
They point out that AI can deliver quick translations and compose texts in foreign languages, but without linguistic understanding you risk errors and misunderstandings. It is therefore crucial to develop new skills in which students learn both to master the language and to assess when AI can be used as a tool.
The opinion piece also underlines the need for coherent coordination in the language field, something that previously sat with Det Nationale Center for Fremmedsprog (NCFF, the National Centre for Foreign Languages). But if the centre is closed, the big question is who will collate knowledge, provide an overview and guide the development of language teaching.
According to Nick Ladany, a university president and researcher, universities face a fundamental change with the introduction of “AI professors”. These AI-based instructors could offer personalised, flexible and continuous teaching based on large amounts of data and pedagogical knowledge, thereby changing how students learn and interact with instructors.
He points out that many university teachers are trained not in pedagogy but in research, and that AI professors could fill this gap. They can be available around the clock, adapt to individual learning styles, update themselves in real time and use evidence-based teaching methods. At the same time, they can teach both subject content and broader skills such as critical thinking and interdisciplinary problem-solving.
However, there must be human oversight so that AI professors do not spread incorrect information—a problem that also exists among human instructors, Ladany notes.
Nick Ladany
Anitia Lubbe
Matthias Bastian
