Three of the world’s more ambitious education systems have now looked at the same technology and reached three different conclusions about what to do with it. The UAE made artificial intelligence a compulsory school subject. New York City moved to restrict student use. England told teachers they could use it to mark quizzes — but not to decide a grade. Same tool, three answers.
I have been in education since 2000, long enough to have watched calculators, the internet, smartphones, and social media each arrive wearing the same costume: the thing that would either save the classroom or ruin it. AI is the loudest arrival yet. But the part worth paying attention to isn’t the technology. It’s that serious people, looking at the same evidence, disagree this sharply about how it should reach a child.
The Gulf decided to go first
The UAE made AI a formal subject across public schools, kindergarten through Grade 12, starting in the 2025–2026 year (WAM). Education Minister Sarah Al Amiri framed it as raising a generation that understands the ethics of the technology and can build with it, not just consume it (Arabian Business). Saudi Arabia moved on the same timeline, rolling out a national AI curriculum across general education.
There’s economic weight behind the ambition. AI is projected to add around $150 billion to Gulf economies by 2030. When the numbers get that large, ministries move quickly — and I understand the appeal. I have sat through enough meetings where a good idea died because nobody would commit to it. The Gulf committed. That deserves credit.
The question I keep coming back to is less about policy than about people: who delivers this, and how ready are they? A qualitative study based on interviews with 19 teacher educators in the UAE found gaps in their own familiarity with AI, alongside real concerns about bias, fairness, and students outsourcing their thinking. It’s one study, not a verdict on every teacher — but it names something real. A curriculum can be written in a year. The confidence to teach it well takes longer.
What’s striking is how directly this is already being addressed here. In Abu Dhabi, ADEK’s AI literacy initiative now runs across more than 170 private schools, built on a human-centred framework, a graded curriculum, a twice-yearly skills assessment, a school technology-and-AI policy covering data privacy and human oversight of AI-generated decisions, and — the part that matters most to me — a professional-development programme aiming to train 19,000 educators. On paper, that answers almost every concern I just raised. It borrows the best instinct from Britain (start with the teacher) and from America’s more careful states (draw clear lines around data and decisions).
So my question isn’t whether the training exists. It plainly does. It’s the harder, quieter one every reform faces: does it reach the classroom, and does it change what happens there? A framework measured twice a year is only as good as what a tired teacher can actually do with thirty students in front of them. That’s not a criticism of the policy. It’s the real work that begins after the policy is announced.
America let the states take different paths
Rather than one answer, the US produced fifty debates. In the 2026 session, MultiState counted 134 AI-in-education bills introduced across 31 states — introduced, not all passed — pulling in different directions. Boston launched a public-private initiative to make its high-schoolers AI-literate, though the city has been careful to call it a proficiency goal, not yet a formal graduation requirement (WBUR). Ohio required every district to adopt its own AI policy. New York City moved the other way, restricting generative-AI use for younger students while allowing limited, supervised use higher up.
The effect is that a child’s exposure to this subject now depends heavily on which district they happen to live in. I find that uncomfortable. We already know what happens when educational quality is left to local accident: it widens the gaps everyone says they want to close.
Britain started somewhere unexpected: with the tired teacher
Rather than ask what AI should do for students, England asked what it could take off teachers’ plates. Government guidance now lets teachers use AI for low-stakes marking, like quizzes, and for routine letters home — on the condition that teachers check the output, since the tools can invent facts (BBC). The motivation was practical: many teachers had rated their own AI confidence very low.
But the line is drawn firmly where the stakes rise. For formal qualifications, the regulator Ofqual is explicit: AI may support marking, but it cannot be the sole marker of a student’s work — a human has to make the judgement that counts. The phrase that keeps recurring across the guidance is “human in the loop.” It isn’t elegant, but it’s honest.
I’ll admit this is the approach closest to my own instinct. Let the software save time on the routine, and keep a person answerable for every decision that shapes a child’s record.
So where does that leave those of us teaching in the Gulf?
We have what the others lack: the will to move and the resources to do it well — and, as ADEK’s programme shows, a serious attempt to prepare teachers rather than just students. The task now isn’t to invent the policy. It’s to make sure the support behind it reaches the people standing in front of the students, and to keep measuring whether it actually does.
And the American lesson still stands: before any new tool reaches a classroom, someone should be able to answer two plain questions — who holds this child’s data, and which decisions will always stay human? Encouragingly, those are exactly the questions the better frameworks are now asking out loud.
Every few years a new machine walks into the classroom promising to do the teacher’s job. It never quite does. It does part of the job, sometimes well, and it leaves behind a harder question than the one it answered: now that the machine can do this part, what exactly was the human for?
That was never a question about AI.
The views expressed here are personal and do not represent those of my employer or any educational authority.
