The year of implementation. AI policies come off the page in AY 2026-27.

Three years of AI-in-schools policy work is now being enacted in classrooms. UAE schools, like their counterparts in Estonia, adopted the technology early. The next stretch of the road looks different. It requires teachers and school leaders to be considered rather than fast, and it puts a very old question back in the middle of the debate. What is education actually for.

Between 2023 and 2026, AI policy in education moved from working papers to published frameworks. UNESCO issued the Guidance for Generative AI in Education. The OECD ran consultations on AI literacy. Estonia rolled a national programme. Singapore did the same. In the UAE, ADEK, KHDA, and the Ministry of Education moved from advisory notes to formal policies for AY 2026-27. This is the academic year the policies become classroom practice.

The UAE was an early adopter. So was Estonia.

Two systems moved before most others did. Estonia treated AI in schools as a natural continuation of a long-standing digital learning agenda. The UAE moved because teachers and students moved on their own. Every serious survey of UAE classrooms in the past two years has shown the same picture. Students are already using AI to draft, revise, translate, tutor themselves, and generate visuals. Teachers are using it to plan, differentiate, and mark. Adoption ran ahead of policy because the tools were free, fluent in English and Arabic, and available on every phone.

That head start is a genuine asset. Systems that have never let AI into the classroom now have to teach both the tool and the discipline of using it. Systems that let AI in early have already learned which uses feel worth keeping and which have to be corrected. What the head start does not confer, though, is a strategy.

From early adopter to considered adopter.

The signal in AY 2026-27 is a shift in what is required of schools. Early adoption was permissive. Considered adoption is deliberate. The regulator now expects a school to have a written policy, a training pathway for teachers, and a framing for students about when and how to use AI. Inspection frameworks are beginning to reference AI practice directly. Parents are asking harder questions than they did two years ago. Schools that ran ahead have to consolidate. Schools that ran late have to catch up on a compressed timeline.

Being an early adopter is a position, not a strategy.

Two risks the policies are trying to prevent.

The regulator publications converge on two concerns, and both matter.

The first is over-restriction. The temptation for a school reading its new policy for the first time is to lock everything down. Ban AI in assessment. Confine it to a computing lesson. Require teachers to declare every use. This posture protects the school on paper. It also shuts off the specific uses of AI that actually help teachers and students. The dissertation research we have published previously on culture and disruption is directly relevant here. Innovation is a controlled-adhocracy question, not a hierarchy question. Regulation should enable considered use, not force schools underground into the shape of use they already had before the policy arrived.

The second is cognitive offloading. This is the risk of using AI in ways that quietly do the thinking a student was supposed to do. It shows up as strong first drafts with no evidence the student wrestled with the argument. It shows up as vocabulary a student cannot define. It shows up as confident answers that a student cannot defend when questioned. The offloading is invisible in the finished product and visible only when the student has to think about the same problem again, without the tool.

The Fadel and Center for Curriculum Redesign work catalogues the trade-offs precisely. On the cognitive side, AI use improves short-term task performance and reduces extraneous load. It also erodes retention, transfer, metacognition, and foundational skills such as writing when used without discipline. On the affective side, AI use raises confidence and engagement in bounded settings. It also risks attachment, dependence, and displacement of human connection. A school policy that names those trade-offs, and puts them in front of teachers and students on a regular basis, does more work than one that only lists permitted tools.

The most important AI skill is not digital.

What the policy year makes clear is that the AI question in schools is not a technology question. It is a cognition question, and behind that, a purpose-of-education question. The most important skill a school can teach in AY 2026-27 is not how to prompt a model. It is the capacity to notice when a piece of thinking has been outsourced, to know which parts of thinking are worth doing by hand, and to have a defensible view of what an educated person can do without assistance. Those are ancient teacherly instincts, and AI raises the price of not having them.

Two questions belong in every teacher's toolkit this year. Does this technology deepen the learning, or replace it. Would the analogue version of this task teach the same concept better. If a teacher can answer those questions in front of a class, on the fly, the policy has been internalised.

What a considered-adopter school looks like this year.

The picture is straightforward once the concerns are named.

• A school policy that permits and structures AI use rather than only prohibiting it.
• Teacher CPD that treats AI as an instructional tool the teacher must judge, not as an add-on the teacher tolerates.
• A student framing repeated across subjects. When AI helps. When AI is not to be used. Why. The reasoning matters more than the rule.
• An assessment approach that makes the student's thinking visible. The prompt, the draft, the change the student made. The change is the learning.
• A protected set of skills every year that is taught without AI, chosen deliberately, and named to students.
• A short list of AI uses the school evaluates each term for whether they still earn their place.

Where the UAE goes from here.

The UAE has an unusual advantage. Its schools moved before the policy did. Its regulators ae now publishing policies designed by people who watched what happened in the intervening years. The country's ambition on AI is national in a way that few school systems can match. What matters this year is whether AY 2026-27 becomes a policy year or an implementation year. Publishing frameworks is not the same as changing what happens in a classroom on a Tuesday morning. That change involves leadership, teaching, and cognition together, and it is the change every school is now being asked to make.

Sources include: UNESCO, Guidance for Generative AI in Education and Research (2023); ADEK AI in Schools guidance (2026); KHDA policy publications on AI (2026); UAE Ministry of Education AI-based learning platform documentation; OECD AI literacy framework consultations (2024-25); Fadel and Center for Curriculum Redesign, AI in the Classroom (2023, 2024); Nasserghodsi (2026), Cultural Agility and School Culture Through a Disruption.