UNESCO's Digital Learning Week 2026 ministerial statement is a diplomatic document, signed by 25+ education ministers, written in the careful language of international policy. But underneath the diplomacy, it names five things — public accountability, deliberative governance, teacher agency, learner rights, safety — that translate surprisingly well into a practical checklist. Here's how to actually use it, on any AI tutoring product you're considering, ours included.

Why this checklist exists

Every AI tutoring product markets itself as safe. That word alone tells you almost nothing — it's the education-tech equivalent of "all natural" on a food label. What's useful isn't the claim, it's a specific enough set of questions that a vague answer becomes visible for what it is.

UNESCO's ministerial statement gives you exactly that: five named principles, developed through a year of international consultation, that are specific enough to ask real questions with. Here's each one, translated from diplomatic language into something you can actually check.

1. Public accountability → "Who's responsible when it's wrong?"

The principle, in UNESCO's language: AI in education must come with public accountability — someone answerable, not a black box nobody can question.

The question to ask: If the AI gives your child a wrong answer, a harmful suggestion, or mishandles a sensitive disclosure, is there a real, named process for what happens next? Or does the product's terms of service quietly disclaim all responsibility for anything the AI says?

2. Deliberative governance → "Did anyone actually think this through?"

The principle: decisions about how AI is used in education shouldn't be made unilaterally by whoever built the technology — they should involve genuine deliberation among educators, researchers, and the people affected.

The question to ask: Was this product designed with input from actual teaching methodology, or is it a general-purpose chatbot with an education skin applied afterward? Does the company publish anything about the pedagogical reasoning behind how the AI behaves, or is that reasoning proprietary and invisible?

3. Teacher agency → "Does this protect judgement, or override it?"

The principle: AI should support a teacher's — and by extension, a parent's — professional judgement, not replace or override it.

The question to ask: Does the product make you (or a teacher) more informed and in control, or does it quietly make decisions you never see — choosing what your child practises, how hard it is, when to escalate a concern — with no visibility into why? We look at this specific principle in more depth in teacher agency in the AI classroom.

4. Learner rights → "Does my child's own thinking stay theirs?"

The principle: AI in education must protect learners' rights — including, implicitly, the right to actually learn rather than have learning outsourced to a machine.

The question to ask: Does the AI ever just hand over the final answer to a problem, or does it structurally refuse to — using hints, questions, and worked examples on a different problem, so your child still has to do the thinking? This is, in our view, the single most important question on this whole list, because it's the difference between a tool that builds capability and one that quietly erodes it. We go deeper on learner rights specifically in learner rights and AI safety: UNESCO's principles.

5. Safety → "What actually happens to my child's data, and when things go wrong?"

The principle: safety, in the structural sense — not a feature, a foundation.

The question to ask: Is there a stated, specific safeguarding policy — what happens if your child discloses something concerning? Is there a stated data-retention policy, and does it default to deleting data rather than keeping it indefinitely "just in case"? Is the safest setting the default, or something you have to go find?

How aitutors.me answers these questions — honestly, not as a sales pitch

We didn't design aitutors.me against this checklist — UNESCO's 2026 statement came years after our own principles were set. But reading our own answers against it is a useful honesty check, and we'd rather show our working than just claim to pass:

  • Accountability: Mentor's safeguarding flow hard-fails to a direct Childline referral on any safeguarding indicator — not a soft nudge, a structural stop.
  • Learner rights: our hint ladder never hands over the final answer to the problem in front of a child — the top rung is a fully worked similar problem, then "now redo yours." That's the product's core commercial promise, not a marketing line.
  • Safety: conversation data defaults to 30-day rolling deletion, and we don't train third-party models on children's conversations.

That's our own answer. The point of this article isn't to ask you to trust it — it's to give you the same five questions to ask of us, or of anyone else.

FAQ

What are UNESCO's core principles for AI in education?

The Digital Learning Week 2026 ministerial statement, adopted by 25+ countries on 9 September 2026, centres on five ideas: public accountability, deliberative governance, teacher agency, learner rights, and safety — with education framed as something that must "remain a common good" rather than be shaped unilaterally by whoever builds the AI.

How do I check if an AI tutoring product is actually safe, not just marketed as safe?

Look for structural evidence, not marketing copy: a stated, specific safeguarding policy (not just "we care about safety"), a clear answer on what happens to your child's data and for how long, transparency about why the AI behaves as it does (why it gives a hint instead of an answer, for instance), and a design that doesn't quietly maximise engagement at the expense of actual learning.

Does any UK AI tutoring product officially meet UNESCO's principles?

UNESCO doesn't certify or endorse individual products, so there's no official badge to look for. The principles are a lens for parents to apply themselves — the questions in this checklist are how you apply that lens to any specific product, whichever one you're considering.

Duke Harewood built aitutors.me for his own KS3 daughter. If a checklist like this one makes any product — ours included — look worse than its marketing suggests, that's the checklist doing its job. Published 11 September 2026.