Scotland recently became one of the first school systems to publish formal AI guidance for schools — and whether you work in Edinburgh or Essex, the five principles it sets out are worth reading carefully.
On 25 March 2026, the Scottish Government published Guidelines and Guardrails for the Use of Artificial Intelligence (AI) in Schools — the first national guidance of its kind in Scotland. It runs to 45 pages, was co-developed with the EIS, and aligns with Scotland's AI Strategy 2026–2031. The press coverage focused, as press coverage tends to, on a handful of quotations from policymakers about technology never replacing teachers. The actual document deserves closer reading than that, because it is both more thoughtful and more limited than the headlines suggest, and because the questions it leaves unanswered are at least as important as the ones it addresses.
What the guidance contains
The document is structured around seven sections: core principles and guardrails, frequently asked questions, a checklist with exemplification, data protection guidance, key definitions, ethical considerations, and a resource list. It is non-statutory — schools and local authorities are not required to follow it, and it is explicitly designed to allow local flexibility in a field that is evolving too fast for prescriptive regulation to keep pace.
The five core principles are worth stating plainly, because they do genuinely useful work in framing how schools should approach AI:
The first principle is that AI use must ensure the safety and privacy of children, young people, and staff — prioritising children's rights, teacher judgement, ethical standards, and data protection. The second is that AI must be underpinned by equity and fairness. The third is that AI must support the aims of the curriculum. The fourth is that AI must foster human connection and inclusivity. The fifth is that AI must support teachers.
Read together, these principles establish a clear hierarchy: the child's rights and wellbeing come first, equity second, curriculum alignment third, human connection fourth, and the teacher's professional role fifth. This ordering is not accidental. The guidance was published shortly after the formal incorporation of the UNCRC into Scots Law, and children's rights language runs through the document in a way that is more than cosmetic — it is accompanied by a Children's Rights and Wellbeing Impact Assessment (CRWIA) informed by engagement with 70 children and young people through the Scottish Youth Parliament, Young Scot, Children's Parliament, and Children in Scotland.
The document is also clear on what AI must not do: it must not make decisions on behalf of teachers. It must not be used to assess pupil progress without teacher oversight. It must not replace peer-to-peer interaction or act as a substitute for teacher involvement. And it must not be deployed in ways that widen the digital divide between well-resourced and under-resourced schools.
What's genuinely useful here
Three things stand out.
First, the guidance names the data protection question explicitly and dedicates a full section to it. This matters because data protection is where most schools' thinking about AI breaks down. The General Data Protection Regulation and the Data Protection Act 2018 apply to AI in schools just as they apply to any other processing of personal data, but the specific implications for AI tools — where data goes, who processes it, whether outputs constitute automated decision-making under Article 22 of the GDPR — are not well understood by most school leaders. The guidance doesn't solve this entirely, but it does make clear that local authorities bear specific legal responsibilities and that schools need to understand how any AI tool they adopt handles data before they use it with children. This is, at minimum, a necessary starting point.
Second, the guidance explicitly addresses the equity question. A survey conducted by the Children's Parliament in 2025 found that 71% of teachers had no or low confidence in addressing AI with their pupils, and 79% had received no guidance on its use. These are not figures from schools in challenging circumstances — they describe the profession at large. The risk, stated plainly in the guidance, is that AI adoption follows existing patterns of advantage: well-resourced schools with digitally confident staff adopt tools that genuinely enhance learning, while less well-resourced schools either avoid AI entirely or adopt it without the understanding needed to use it well. The guidance's insistence that AI adoption must be equitable does not, in itself, produce equity — but naming the risk is a precondition for addressing it.
Third, the resource signposting is useful. The guidance points schools to the TRAILS.Scot platform — the AI literacy curriculum framework developed by Judy Robertson and colleagues at the University of Edinburgh — and to the Digital Education and AI Literacy Hub hosted by the Data Education in Schools team, as well as the Centre for Teaching Excellence's forthcoming Digital Education and AI Hub. These are serious resources, developed through sustained research programmes rather than assembled hastily in response to ChatGPT. For a teacher who has been told to 'get to grips with AI' but given no starting point, the resource list alone justifies reading the guidance.
Where the gaps are
The guidance is honest about its limitations in a way that is commendable but also, frankly, a bit sobering. It acknowledges that the evidence base on AI in education is still developing. It acknowledges that the risks — to data security, to children's rights, to pupil development, to the quality of human interactions in classrooms — are real but not yet fully understood. What it does not do, and what its authors would probably concede it cannot yet do, is tell schools what to do on Monday morning.
The most significant gap is practical. The five principles are sound, but translating them into decisions about specific tools, specific classroom contexts, and specific pupils requires a kind of operational guidance that a 45-page national framework cannot provide. Consider a concrete scenario: a secondary English teacher wants to use a generative AI tool to help S3 pupils draft and redraft persuasive essays. The guidance tells the teacher that AI must support the curriculum, must not replace teacher involvement, must ensure data privacy, and must be deployed equitably. These are the right principles. But the teacher's actual questions are more immediate: Which tool? What does the school's data protection officer say about the terms of service? How do I ensure the pupil who doesn't have a device at home isn't disadvantaged? How do I assess work that has been produced in collaboration with an AI? What do I do when the AI produces confident-sounding nonsense and the pupil doesn't notice? The guidance gestures toward these questions — particularly through its FAQ and exemplification sections — but does not, and probably cannot, answer them with the specificity that classroom practice demands.
There is also a notable silence around assessment. The guidance states that AI used for assessment purposes should involve teacher oversight, which is correct as far as it goes. But the deeper question — how AI changes the nature of what can meaningfully be assessed — is barely touched. No jurisdiction has answered this question well. But for a country in the middle of a major curriculum and qualifications reform programme, with Qualifications Scotland reviewing assessment approaches and the Curriculum Improvement Cycle still in its early stages, the intersection between AI and assessment is not a future problem. It is a present one, and the guidance would have been stronger for engaging with it more directly.
AI guidance is one of several policy developments reshaping Scottish education — the Qualifications Scotland reforms raise equally fundamental questions about what we're preparing young people for.
Before schools rush toward AI-generated insights, they might want to ask whether they're using the data they already have.
The third gap is in the relationship between AI and school-level data. The guidance focuses, understandably, on the AI tools that teachers and pupils interact with directly — generative AI, adaptive learning platforms, digital tutors. But the more transformative and more quietly disruptive uses of AI in education are not pupil-facing at all. They sit in the systems that schools use to track attainment, monitor attendance, identify patterns in behaviour, and allocate support. These are the applications where algorithmic decision-making has the greatest potential to help — and the greatest potential to cause harm if the data is bad, the model is opaque, or the humans in the loop aren't asking the right questions. The guidance's principle that AI must not make decisions on behalf of teachers is exactly right here. But the principle needs operational teeth: what counts as a 'decision'? If a system flags a pupil as at risk of disengagement based on attendance and attainment data, is that a decision or an observation? If a teacher acts on that flag without interrogating how it was generated, has the AI made the decision in practice if not in principle?
What this means for schools
The guidance is best understood as a starting point rather than an answer. It gives schools permission to engage with AI — which some were waiting for — and it gives them a principled framework within which to do so. It does not give them the practical infrastructure they need: the vetted tools, the professional learning time, the data protection clarity, or the assessment frameworks that would make AI adoption genuinely safe and equitable at scale.
For school leaders, the most actionable thing in the document is probably the requirement to update existing digital technology policies to reflect AI use — carried out in consultation with unions, parents, and pupils. This is not a small ask, but it is a concrete one, and it forces the conversations that need to happen locally rather than waiting for further national direction.
For teachers, the most important message is the repeated insistence on professional autonomy: the teacher decides when and how AI adds value in their classroom. This is welcome, but it comes with an obligation that the guidance does not quite spell out. Professional autonomy requires professional knowledge. A teacher who decides not to use AI because they don't understand it is not exercising autonomy; they are constrained by a gap in their professional learning. And a teacher who uses AI without understanding how it works, what it does with data, or where its outputs should and shouldn't be trusted is exercising autonomy without the knowledge to do so wisely. The guidance points to resources. It does not — and probably, given the non-statutory nature of the document, could not — guarantee that schools will create the conditions for teachers to access them.
The bigger question
The most interesting thing about the guidance is what it reveals about where Scotland is, politically and educationally, on the question of technology in schools. The document was co-produced with the EIS, whose general secretary Andrea Bradley's position — that AI is a tool for teachers and must never replace professional judgement — is embedded throughout. This is the right instinct, and it produces the right principles. But it also shapes the limits of what the guidance can say, because it forecloses the more uncomfortable conversation about what happens when AI doesn't just support existing practice but changes what good practice looks like.
The guidance's foundational premise is that education in Scotland is, and must remain, a human enterprise. At the level of values, this is unassailable. At the level of practice, it is more complicated than the document quite acknowledges. The question is not whether AI will replace teachers — it will not — but whether it will change what teachers need to know, what they need to be able to do, and what counts as evidence that their pupils have learned. The guidance, by design, does not attempt to answer these questions. They will be answered, over time, by the schools and teachers who take its principles seriously and begin the work of figuring out what they mean in a classroom on a Tuesday afternoon with thirty young people and a curriculum to deliver.
Scotland has a document. What it needs next is practice — and the willingness to learn from it honestly, including when it goes wrong.
If you're weighing how schools turn classroom evidence into defensible insight (with or without AI), Learning Lens is available for Scottish schools — book a conversation if you'd like to see the approach in your setting.
Jamie Scobie writes from extensive experience in Scottish secondary education, including pastoral care, data for improvement, and school self-evaluation. This blog is an independent publication: he writes in a personal capacity as the creator of Learning Lens, writing about classroom observation, teaching evidence, and education policy. He speaks here only for himself and for Learning Lens, not for any employer or other organisation. He holds an MSt from Cambridge (Distinction) and a Masters from Stirling.