Danish school platform builds AI rules into the individual assignment

AI has already become part of many pupils' schoolwork. For schools, the question is therefore no longer only whether pupils may use AI, but when, how and under which rules.

In the Danish school platform SkoleElev.ai, each school sets its own shared AI policy. Within that overall framework, the teacher chooses how much AI assistance is permitted in a given subject, at a given grade level and in the individual assignment.

The assignment itself states whether the pupil may not use AI at all, may use it for things such as idea generation and outlining, or may use it more extensively with a requirement to disclose how it was used.

AI use in the platform is logged, so the school can follow up and create consistent rules across teachers, subjects and classes.

"Schools do not need yet another general statement of intent about AI. They need rules that translate into everyday practice and are clear to both teachers and pupils. That is why the AI level is tied directly to the individual assignment."

AI should guide – not deliver the answer

Since launch, the pedagogical starting point for SkoleElev.ai has been that AI must not act as the pupil's answer machine.

Instead, the AI should ask questions, explain difficult material and help the pupil move forward in the learning process. It should support the pupil's independent thinking rather than take it over.

"We have not built a machine to solve the assignments for pupils. AI should help the pupil understand the problem, ask better questions and move forward. The teacher remains the academic authority."

When the platform gives the teacher a suggestion, an assessment or a draft, it is presented as professional input – not as an automated decision. The final judgement is made by the teacher.

The platform is not designed to make decisions about pupils based solely on automated processing where the decision may have legal effect or similarly significant impact on the individual pupil.

Prepared for upcoming requirements for AI in education

The EU AI Act is being introduced in stages. Certain AI systems in education may be classified as high-risk systems. This includes systems intended to:

• •evaluate learning outcomes and use the evaluation to steer the pupil's further learning
• •determine access or admission to education
• •assess the level of education a person should have access to
• •monitor and detect prohibited behaviour during examinations

The detailed requirements for this type of high-risk AI system take effect on 2 December 2027.

SkoleElev.ai has chosen to design the platform around central principles of the regulation, regardless of whether a given feature would legally be classified as high-risk in a specific use. This includes transparency, traceability, documentation and human oversight.

Pupils, teachers and parents can see when AI has been involved and what role the technology played. At the same time, the school can require the pupil to account for the use of AI in a submission.

The school sets the boundaries

It is the individual school – not the platform – that sets the rules for pupils' use of AI.

The school can choose different levels of AI assistance and adapt them to each subject, grade level and assignment type. A Danish teacher may allow AI for idea generation but not for the final wording, while a teacher in another subject may allow more extensive support.

The teacher therefore does not have to reinterpret an overall AI policy each time. The assignment states directly which assistance the pupil may use.

"Schools have very different pedagogical approaches to AI. A vendor should not dictate one particular policy. Our job is to give the school the tools to set its own line and enforce it clearly and consistently."

Core infrastructure and backup in the EU

According to SkoleElev.ai, the platform's core infrastructure is located at Hetzner in Falkenstein, Germany. The environment is mirrored in Nuremberg, and encrypted daily backups are made in Helsinki.

A data processing agreement is concluded under GDPR article 28, and schools get access to an up-to-date list of the sub-processors involved in operations.

Identifying information about pupils is pseudonymised before data is sent for processing in language models. According to the company, pupil data is not used to train models, and no advertising is shown to children.

Pupils must also learn to think critically about AI

Responsible use of AI is not only about technical safeguards and data processing. It is also about teachers and pupils understanding the possibilities and limitations of the technology.

SkoleElev.ai therefore includes material for teachers on responsible use of AI. Pupils work with source criticism, bias in AI systems and the risk that a language model can deliver incorrect or fabricated information.

Pupils must learn that a well-worded AI answer is not necessarily correct, and that answers from AI must be assessed in the same way as other sources.

"Regulation is often portrayed as a brake. For us, the AI Act is a quality standard we welcome — it describes exactly the accountability schools have been asking for all along."