We can't buy our way out of responsibility. While the U.S. is betting on error-prone detectors and STIL is tightening data rules, we're at a crossroads. The solution isn't new software, but reinventing pedagogy.
Here is this year’s second newsletter on AI in the education sector, where we look at the question of who actually decides what AI should be used for in teaching.
Principal Per Farbøl has written an opinion piece warning against an uncritical integration of AI in teaching. The battle over AI is not only about cheating, but just as much about what teaching should fundamentally be built on. If the school does not choose a direction itself, the tech giants will do it for us.
Law student Gustav Rudolph Røgild from Københavns Universitet (University of Copenhagen, KU) is feeling the consequences of the lack of a decision. He fears a split into A- and B-groups, where some students seek out AI competences themselves while others never acquire them. It is not only about the labour market, but also about the rule of law.
Three institutions show that it is in fact possible to choose another path. In Uddannelsesmonitor, IBC, Vestfyns Gymnasium and ZBC explain how they have moved from control to reflection. Challenges remain, but what they share is that they have made a series of choices about AI and thereby taken the lead.
Last week, LEGO Education launched a new concept that seeks to balance the physical with the digital without data storage or pupil logins. At the same time, it offers a way to work with AI and technology comprehension in primary and lower secondary school.
In a broader perspective, Associate Professor Kim Toft Hansen has received 18 million Norwegian kroner for the GAIYA research project, which will investigate digital inequalities in young people’s use of AI in the Nordic countries and the Baltics.
We should, however, beware of the pitfalls. New figures from the US show that blind faith in AI detection hits innocent students, and Professor Pablo Sanguinetti reminds us that even the way we talk about AI can distort our understanding of who has control.
There is also room for this week’s commentary, where I write about why we cannot buy our way out of responsibility.
Enjoy the newsletter.
In an opinion piece on Uddannelsesmonitor, Per Farbøl, who is principal at Stenhus Gymnasium (upper secondary school (gymnasium)), warns against an uncritical integration of AI in teaching. He argues that the struggle over AI is not just about preventing cheating, but about the basic values underpinning education.
He draws parallels to the development of social media, where a small group of technological actors managed to create dependency and shift the focus away from presence, immersion and critical thinking. With AI, he sees a similar risk, namely that pupils hand over reflection, dialogue and learning processes to the technology.
According to Per Farbøl, schools now face a choice about whether teaching should be based on shared academic and pedagogical principles or on individual interactions with AI. Should students’ learning be built on understanding and dialogue, or on finding ‘the right answer’ via AI? If the education sector does not choose its direction, there is a risk that the decision will be taken by technological actors or political considerations.
Law student Gustav Rudolph Røgild from Københavns Universitet (University of Copenhagen, KU) expresses concern that the university is too slow to integrate AI into teaching. He fears it will split students into an A- and a B-group, where some seek out knowledge about AI themselves while others do not receive it as part of their degree.
The university acknowledges the lag and has set up a working group to develop proposals for how AI can be integrated into teaching.
According to Gustav Rudolph Røgild, it is not only about being ready for the labour market. It is also about legal safeguards and about lawyers’ ability to understand and navigate complex factual and technological issues.
Read the opinion piece here:
Freja Lillevang Sørensen
A year after Styrelsen for Undervisning og Kvalitet (the Danish Agency for Education and Quality) issued its recommendations for the use of AI in teaching, Uddannelsesmonitor asked three schools how AI is being used as a pedagogical tool. IBC, Vestfyns Gymnasium and ZBC say the focus has shifted from control to learning and reflection.
At IBC they are working on the project Hybrid intelligens, where AI is combined with human intelligence. The project explores how learning can be supported by breaking tasks into subprocesses and using AI for ideation, structure and sparring.
Six first-year classes at Vestfyns Gymnasium are taking part in a project where students work with portfolio-based writing both with and without AI. The aim is to restore confidence in their own writing and strengthen students’ ability to reflect critically on AI’s output.
At ZBC they have for several years worked to equip teachers to use AI through workshops and now also via asynchronous materials. The school sees large differences in the use of AI, but notes a growing understanding that the technology is not a threat but a tool that requires conscious and critical use.
The three institutions share a desire to develop a teaching practice where AI supports students’ learning and motivation while the teacher retains their role as mentor and subject guide.
LEGO Education is launching a new teaching concept in April with a focus on technology comprehension and artificial intelligence. The materials are developed for primary and lower secondary school and are divided into three sets for the early years, middle years and lower secondary. The solution is based on hands-on, collaborative learning, where pupils work together to investigate, build and program.
The idea behind LEGO Education is to create a balanced approach that combines physical activities with digital programming, without any requirement for pupil logins or data storage.
The accompanying Coding Canvas app enables block-based programming and supports pupils in developing their technological understanding.
LEGO Education
Associate Professor Kim Toft Hansen from Aalborg Universitet (Aalborg University) has received a grant of 18 million Norwegian kroner (NOK) from Nordforsk for the GAIYA project.
The project will examine how upper secondary school (gymnasium) students in the Nordic and Baltic countries use generative AI, and how this affects teaching and digital competences. It has a particular focus on digital literacy and gender equality.
A central aim is to analyse how digital differences between genders arise and how they can be counteracted. Previous research shows that boys often adopt new technology faster than girls, which can lead to digital inequalities in both school and working life. The project therefore focuses on playful learning, digital pedagogy and the development of inclusive teaching strategies.
The GAIYA project views generative AI as a cultural and social technology, not merely a technical tool. It will result in policy recommendations, open educational resources, and a strengthened understanding of how the technology can be used responsibly and equitably in future teaching.
The project runs from 2026 to 2030.
The Independent reports that a growing share of teachers in the United States are using AI detection tools to identify cheating among students, but research and experience show that these tools are often unreliable and can harm innocent students.
A survey by the Center for Democracy and Technology shows that 43 percent of teachers in Grades 6 to 12 used AI detection in the 2024/25 school year, even though the technology often produces both false positives and false negatives.
Students such as Marley Stevens and Ailsa Ostovitz are examples of young people who have been falsely accused and have experienced serious consequences, both academically and mentally. Marley Stevens lost her scholarship after being penalised for using Grammarly, while Ailsa Ostovitz has been accused three times, even though she wrote her assignments herself.
A study from the European Network for Academic Integrity concludes that even the best AI detection tools are not suitable as evidence of academic misconduct. The tools are easy to circumvent and often misjudge both human- and machine-written text.
It is therefore recommended to focus on prevention and on developing alternative forms of assessment, where the learning process is central rather than just the finished product.
Mike Bedigan
In The Conversation, Professor Pablo Sanguinetti writes that AI is not just technology, but is also shaped by the words and images we use to describe it. According to Pablo Sanguinetti, we risk misunderstanding, using and regulating AI incorrectly because the public narrative is shaped by metaphors and myths that do not align with reality.
He particularly criticises the idea of AI as autonomous and human-like, a view rooted in both mythology and science fiction. This approach distracts from the fact that AI is built on human labour, data and infrastructure and has neither will, consciousness nor feelings. When we talk about AI as an autonomous being, we risk forgetting that it is, in reality, a tool under human control.
He therefore suggests that we change the way we talk about AI, both linguistically and culturally. We should describe AI with more precise terms such as "data processing" and "computation" rather than words such as "thinks", "dreams" and "creates". This is especially important in education, the media and politics, where the narrative about AI influences decisions and expectations.
If we do not change the narrative, we risk public debate about AI, including ethical and regulatory questions, being built on misunderstandings.
Pablo Sanguinetti
Reading this week’s selected news, a paradox emerges.
On the one hand, we see teachers in the United States trying to solve the AI challenge with more technology. The article shows that more innocent students are being punished, and that this leads to mistrust.
We see signs of the same trend in Denmark.
Last week, STIL (the Danish Agency for IT and Learning) published a mini-guide on the lawful use of AI. The challenge is that, as a school leader, you may get the impression that you can easily procure technology that will uncover cheating. But in reality, the tools do not work particularly well. Today’s language models are so numerous and so advanced that it is, technically speaking, impossible to detect patterns with certainty. And, as the mini-guide itself emphasises, such a system must never stand alone, which means the benefit disappears because all assignments must still be read manually to ensure due process.
On the other hand, we see a pedagogical and didactic approach gaining ground. Schools such as IBC, Vestfyns Gymnasium and ZBC are shifting the focus away from the product and towards the process. In my view, that is the path everyone should take instead of simply investing in new technology. But we must also bear in mind that technology is deeply integrated into everything we do, and it therefore makes no sense to take it out of the equation.
We cannot buy a licence that will solve AI in schools. We have to do the hard work ourselves, which means we must reinvent the way we teach and assess. It is cumbersome, yes, but the alternative is worse: a two-tier system of students and a culture characterised by mistrust.
