If you've never heard of Sonia Livingstone or Beeban Kidron, you've still felt their work. Between them, they're two of the people most responsible for the idea that a website or app should be built differently for a child than for an adult — an idea now written into UK law. Both had a seat at UNESCO Digital Learning Week 2026's conversation on AI and children. Here's the argument they're making, and why it's a genuinely useful lens for any AI tool in your child's evening.

Two different routes to the same conclusion

Sonia Livingstone is a professor at the London School of Economics who has spent close to three decades researching how children actually use digital technology and what puts them at risk. She leads LSE's Digital Futures for Children programme and is known for a simple, influential framework — the "4 Cs" of online risk (content, contact, conduct, and contract/commercial risks) — that reframes child online safety away from "screen time is bad" and toward "what specifically is this product doing to this child."

Beeban Kidron came at the same problem from a different direction: as a crossbench peer in the House of Lords and the founder of the 5Rights Foundation, she was the driving force behind the UK's Age Appropriate Design Code (widely known as the "Children's Code") and a persistent voice in the campaign that produced the Online Safety Act. Where Livingstone's work is research, Kidron's is regulation — she has spent years turning "children deserve better defaults" into enforceable law.

Both were part of UNESCO Digital Learning Week 2026's programming on AI and education, alongside the ministerial statement's own emphasis on learner rights and safety (see the ministerial statement, explained for parents). Their presence on that stage is itself a signal: UNESCO is treating age-appropriate design as a first-class AI-in-education issue, not an afterthought bolted onto a broader safety conversation.

What "age-appropriate AI design" actually asks for

Neither Livingstone's research nor Kidron's regulatory work is really about AI specifically — both predate the current wave of generative AI tools by years. But the underlying principle applies to AI cleanly, and it's more specific than "make it safe":

  • The safest option should be the default, not something a parent has to dig through settings to find.
  • No design pattern should be built to extend engagement at a child's expense — no infinite scroll logic, no "just one more" nudges, no gamification designed to maximise time-on-app rather than learning.
  • Data collected should be the minimum necessary, not "everything, because it might be useful later."
  • The product should match a child's developmental stage, not present an adult experience with a content filter bolted on top.

Applied to an AI tutor specifically, this reframes some questions you might not have thought to ask: not just "is the content appropriate?" but "does the interaction pattern itself respect a child, or is it built to maximise engagement the way an adult social product would be?"

The UK already has law here — mostly

This isn't a hypothetical. The UK's Age Appropriate Design Code (part of the Data Protection Act 2018 framework) already applies to any online service "likely to be accessed by children" — which plainly includes AI tutoring products aimed at KS3 students. The Online Safety Act adds further duties around harmful content and risk assessment for services children can access.

Neither was drafted with generative AI specifically in mind, and that's precisely the gap that arguments like Livingstone's and Kidron's are pushing UNESCO's international audience to close: existing child-safety law was built for the social-media and app era, and AI's specific risks (a chatbot that gives a child a wrong or harmful answer with total confidence, an AI that quietly nudges toward longer sessions rather than better learning) need to be reasoned through explicitly, not assumed to be covered.

What this means for a UK parent right now

You don't need to follow the policy debate to use its logic. When you're looking at any AI tool your child might use — ours included — it's worth asking:

  • Is the most protective setting the one it starts with, or do I have to find it?
  • Does anything in the design seem built to keep my child using it longer, rather than to help them stop once they've learned something?
  • What happens to my child's data, and for how long?

We built our own answers into the product rather than as an afterthought — safeguarding is a structural hard-fail in Mentor's design, not a toggle, and conversation data defaults to 30-day rolling deletion. That's not because UNESCO or the Age Appropriate Design Code told us to; it's because the same logic Livingstone and Kidron have spent decades building into UK law is, we think, simply the right way to build anything a child uses. See our fuller answer in is AI tutoring safe? A UNESCO-principles checklist.

FAQ

Who are Sonia Livingstone and Beeban Kidron?

Sonia Livingstone is an LSE professor who leads the Digital Futures for Children programme and has spent decades researching children's online rights and risks. Beeban Kidron is a crossbench peer and the founder of the 5Rights Foundation, best known for driving the UK's Age Appropriate Design Code (the "Children's Code") and campaigning for the Online Safety Act. Both were part of UNESCO Digital Learning Week 2026's programming on AI and children.

What does "age-appropriate AI design" actually mean?

It means an AI product is built assuming a child, not an adult, is on the other end — defaults set to the most protective option, no dark patterns that pull a child into longer or riskier use, data collection minimised by default, and content and behaviour matched to a child's developmental stage rather than a one-size-fits-all adult experience.

Does UK law already require age-appropriate AI design?

The UK's Age Appropriate Design Code applies to any online service "likely to be accessed by children," which includes AI education tools, and the Online Safety Act adds further duties around harmful content and risk assessment. Neither was written with AI specifically in mind, which is exactly the gap Livingstone's and Kidron's UNESCO-stage arguments are pushing to close.

Duke Harewood built aitutors.me for his own KS3 daughter. The Age Appropriate Design Code shaped how we think about defaults long before we'd read a word of UNESCO's 2026 programme — it's good law, and it's worth knowing it exists. Published 11 September 2026.