When Swedish schools talk about AI, the loudest questions are about strategy and curriculum. But there is a quieter, more specific document worth reading: guidance from the one national authority whose entire mandate is students who learn differently. In November 2025, SPSM, Specialpedagogiska skolmyndigheten (the Swedish National Agency for Special Needs Education and Schools), published a page on AI from a special-education perspective. It is measured, it is honest about the risks, and it lands on a principle that matters for every neurodivergent student in the country.
The starting principle
SPSM does not open with the technology. It opens with inclusion. In its words, “För att vara inkluderande måste användandet av olika AI-tjänster i skolan genomgående ske med hänsyn till elevers olika förmågor och förutsättningar.” To be inclusive, the use of AI services in school must consistently take account of students’ different abilities and conditions.
That is the whole argument in one sentence. AI in a classroom is not automatically good or bad for a student with ADHD, autism, or dyslexia. It depends entirely on whether it is built and used with those students in mind.
What SPSM says AI can actually do
The guidance is concrete about the opportunities. AI can, in SPSM’s assessment, “i flera fall förtydliga sökningar på ett sätt som en traditionell sökmotor inte kan” (in many cases clarify searches in a way a traditional search engine cannot). It points to creating custom images and geometry figures, supporting language development through text suggestions and refinement, and generating targeted exercises in a specific subject.
It also lists the accessibility functions that AI can bring together: speech-to-text input, text-to-speech output, image description, keyboard navigation, screen-reader support, and translation.
Read that list against what actually holds neurodivergent students back day to day: working memory that drops the thread, reading that costs so much effort there is nothing left for comprehension, a blank page that will not start, instructions that arrive faster than they can be processed. These are not the same barrier, and a single accommodation does not clear all of them. That is the case for tools that adapt per student, which is exactly the ground SPSM is pointing at.
The cautions, taken seriously
SPSM is not a cheerleader, and neither are we. It warns that unreflective use can stop students from practising foundational skills. It flags data privacy plainly: “AI använder sig av de uppgifter som den ‘matas’ med i andra sammanhang” (AI uses the data it is fed in other contexts). It notes that biased training data produces biased output, and that AI can create a false appearance of correctness.
These are the right cautions, and they shape how a support tool should work. Support that reaches the same learning goals is not the same as a shortcut that removes the thinking. The aim is to lower the barrier to starting and staying with the work, not to hand over the answer. Data about a child’s difficulties is some of the most sensitive data a school holds, and it has to be treated that way, not fed into a general-purpose model that reuses it elsewhere.
The risk that is easy to miss
The most interesting part of SPSM’s guidance is a risk that rarely gets airtime. A protective, well-meaning “care culture” can end up excluding certain students from learning to use AI at all, and by doing so “fördjupa kunskapsklyftorna”, deepen the knowledge gaps. Alongside it, SPSM notes that AI tools which are not built to accessibility standards simply shut out students with disabilities.
Put those two together and the conclusion is uncomfortable but clear. Shielding neurodivergent students from AI, out of caution, is not the safe option. It is another way of leaving them behind, on top of the support gap they already face. The inclusive move is not less AI for these students. It is AI that is built to include them from the first line of code.
Where Nuro fits
Nuro sits in exactly the lane SPSM describes. The support that students with ADHD, autism, and dyslexia are owed is not optional goodwill, it is what Skollagen already requires. The debate about AI strategy and curriculum can take its time. Using AI to deliver individualised support, with or without a formal diagnosis, does not have to.
We build for these students first, not as an afterthought: accessibility by default, adaptation per student, and a firm line that the goal is the same learning outcome, reached in a way that fits how a given student works. SPSM sets the bar for inclusive AI in Swedish schools. That bar is the product.