On 9 September 2026, at UNESCO's Digital Learning Week in Paris, more than 25 education ministers and representatives signed a joint statement putting a name to something a lot of teachers have been quietly worried about: "teacher agency." AI in education, they agreed, has to keep the human professional in charge — not sideline them into supervising an algorithm. It's a principle written for classrooms and governments. But if you've ever watched your child open a homework app after tea, the same question is sitting in your living room too.

What UNESCO actually said

The ministerial statement, adopted alongside the rest of Digital Learning Week 2026's proceedings, set out a small number of conditions education systems should insist on before letting AI further into schools: public accountability, deliberative governance, learner rights, safety — and teacher agency. The idea behind that last one is straightforward, even if the phrase sounds like policy-speak: a teacher should still be the one deciding how a lesson goes, still be the one exercising professional judgement about a struggling student, still be able to override or ignore what an AI system suggests. The tool serves the teacher. Not the other way round.

It's a reasonable line to draw, and not a new worry — teachers have raised versions of it since the first "personalised learning" platforms arrived a decade ago. What's new is a room full of governments agreeing, on the record, that it matters enough to put in a joint statement.

The gap the statement doesn't cover

Here's the thing about a ministerial statement on AI in education: it's aimed at education systems — schools, curricula, national policy. It has nothing to say, directly, about the app your Year 8 opens on their own laptop at 7pm to get unstuck on a maths problem. That tool was never a classroom decision. No teacher chose it, registered a class for it, or reviewed how it works. You did — or your child did, and you're finding out about it now.

That's not a loophole in UNESCO's thinking so much as a boundary of what a statement about education systems can reach. But the principle underneath teacher agency doesn't actually stop at the school gate. It's really a principle about where human judgement sits relative to an algorithm — and at home, the adult in that position isn't a teacher. It's you.

Teacher agency, translated for the kitchen table

Swap "teacher" for "parent" and the statement's logic holds up surprisingly well:

  • You should be able to see what happened. A teacher can look at what a classroom AI tool told their students. Can you look at what a homework tool told your child? If the answer is "not really, it's just a chat window that clears itself," you've lost the visibility the whole principle depends on.
  • The tool should support your judgement, not replace it. A teacher decides whether a lesson plan an AI generates is actually right for their class. You're in the same position when your child says "the app just explained it" — was that explanation any good, and did it leave your child able to do the next one alone?
  • It should know its limits. Teacher agency assumes an AI tool defers to the human on anything it shouldn't handle solo — a struggling student, a safeguarding concern, a judgement call about effort versus difficulty. At home, that means a tool that notices when something's off and points back to you, rather than quietly ploughing on.

None of this requires a homework tool to be run by a government or reviewed by a ministry. It just requires the same basic shape: a human stays informed, stays able to override, stays the one making the actual call.

Why this matters more than it looks like it does

It's tempting to read "teacher agency" as an abstract governance concern that has nothing to do with a 20-minute homework session. But the failure mode it's guarding against is exactly the one parents worry about most concretely: a tool that gets so good at producing the answer that nobody — not the teacher, not the parent, not eventually the child — is exercising any judgement at all. A maths app that just states "x = 5" has, in miniature, done precisely what the ministerial statement is warning against: removed the human from the loop.

That's also the reasoning behind why a Socratic approach — a tutor that asks questions and hands the thinking back rather than handing over answers — matters for reasons well beyond "cheating." It's a design choice that keeps a human (in this case, your child, working things out) doing the actual cognitive work, instead of an algorithm quietly doing it for them while everyone else stays out of the loop.

What to actually check

You don't need to read UNESCO's statement to apply this at home — you need three habits:

  1. Ask to see it. Sit with your child for one session and actually watch what the tool says. Not to police them — to know what "help" looks like in this app.
  2. Ask whether it explains or just answers. A tool that shows its reasoning, or refuses to give the final step outright, is doing the work of keeping a human — your child, thinking it through — in the loop. One that just states the answer isn't.
  3. Ask what happens when something's wrong. Does the tool ever say "I'm not sure" or "ask your teacher about this one"? Or does it always sound confident, on everything, regardless of whether it should be?

Those three habits are teacher agency, translated for a room UNESCO's statement was never written to reach — but whose logic it describes almost exactly.

FAQ

What does "teacher agency" mean in UNESCO's AI guidance?

It means teachers stay the decision-makers when AI enters a classroom — choosing how and whether a tool is used, retaining professional judgement over a child's learning, and never being reduced to a passive supervisor of an algorithm. UNESCO's 9 September 2026 ministerial statement names it as one of the non-negotiables education ministers agreed on for AI in education.

Does teacher agency apply to AI tools my child uses at home, outside school?

UNESCO's statement is written for national education systems and classrooms, so it doesn't literally regulate a homework app. But the underlying logic travels well: at home, you are the adult closest to the decision, in the same position a teacher is in at school. A tool that keeps you informed and in control respects that logic; one that quietly does the thinking for your child, out of your sight, does not.

How can I tell if a homework AI tool respects the same principle?

Look for visibility, not just safety features. Can you see what your child asked and how the tool responded? Does it explain its reasoning or hand over a bare answer? Does it nudge your child back to you or their teacher on anything it shouldn't handle alone (wellbeing, a topic it's unsure about)? A tool built around those habits is extending teacher agency into the home rather than working around it.

Duke Harewood built aitutors.me for his own KS3 daughter — a tutor designed so a parent can always see what happened in a session, and so the tool hands the thinking back rather than the answer. Published 11 September 2026.