Your child will learn about the world from an AI. Whose world is it?
A mother in Kenya told us she was surprised.
Her child had been exploring Kenya in our app. She said the content was correct and respectful. She said that hearing Una tell a Kenyan story reminded her of listening to her grandparents.
It is the kindest thing anyone has said about this work. I have also not stopped thinking about the first word.
Surprised. Nobody is surprised by something they expected. Surprise is a measurement. It tells you, quite precisely, what a family has learned to expect from everything else their child has ever been shown.
What does an AI tutor actually know?
In August 2026, Meta published an essay called "The Future is for Everyone." One promise in it has stayed with me. "Everyone will have a personalized tutor and coach with a PhD in every subject and unlimited patience to help you learn anything you want."
It is a lovely sentence. It is also worth reading slowly, because your child may well grow up with that tutor sitting in their pocket.
A PhD is awarded by an institution, inside a tradition, examined by particular people. So when a system offers a PhD in every subject, the fair question is a simple one. Whose subjects, and whose sources?
An AI knows the languages and cultures that sit in its training data. Some places are represented richly. Others are represented by a handful of travel articles and a documentary. Where the data is thin, though, the answer does not come back thin. It comes back fluent, warm, and completely confident.
That combination is the thing I would want every parent to understand before their child starts asking a machine about the world. Not thin knowledge. Thin knowledge delivered with total confidence.
Why confidence is the risk, not ignorance
Children are not sceptical readers. They are not supposed to be yet. A patient voice that answers every question without ever getting tired is enormously persuasive to an eight-year-old.
So a story about Kenya, told confidently, built mostly on sources written by people who have never lived there, does not arrive as one version among many. It arrives as Kenya. It becomes the picture in your child's head, and pictures formed at eight are stubborn things.
The companies building the largest models also have little reason to close that gap. Their stated goal is competitiveness. Getting a Kenyan classroom right does not serve it, it costs real money, and almost nobody is asking them for the receipt.
I say that as an observation about incentives rather than an accusation.
Incentives are simply what they are. But they are worth knowing about when the product in question is going to teach your child what the world looks like.
What we do about it, and where we still get it wrong
Unomundi is a cultural-discovery app where children aged 6 to 12 explore countries and cultures through stories, games, and conversations with Una, a warm AI guide.
We work with a cultural anthropologist. We work with local fellows in the places we make content about. It costs more than generating a story and moving on, and it is slower. We chose it anyway, because Una is warm and warmth makes accuracy matter more, not less.
Here is the part I want to be honest about, because the alternative is a promise I cannot keep. We get things wrong. We do not catch everything before it reaches a child.
What we have built is a way to hear about it. When someone from Kenya tells us we have got Kenya wrong, we change the story, and we update the knowledge the app draws on so the correction sticks rather than being fixed once and forgotten.
That is a small promise next to a personal superintelligence for everyone. It is also one a mother in Nairobi can hold me to, and that matters to me more than the size of it.
One question worth asking at home
You do not need to audit anybody's training data. But there is one small habit worth building, and it costs nothing.
When your child comes back from an app, a video or a lesson with a fact about a country, ask them who told them. Not to make them doubt it. Just to plant the idea that information has an author, and that authors come from somewhere and see from somewhere.
Children who grow up asking that question tend to become adults who still ask it. In a world where most of what they read will be generated, that habit may be the most useful thing we hand them.
The mother in Kenya was surprised because her child had spent years meeting versions of the world drawn by people a long way from it. Her surprise was earned, and it was not earned by us.
She should not have been surprised. That is the standard I would like all of us building for children to be held to.
FAQs
How do I know whether an AI is teaching my child accurate things about other cultures?
Ask where the information came from. A tool built for cultural accuracy will tell you who reviewed its content and how corrections are handled. If nobody named is accountable for what the AI says about a country, treat its confidence as a style rather than a signal of accuracy.
Why does AI get some countries wrong?
An AI knows the languages and cultures that sit in its training data. Some places are represented richly, others by a handful of travel articles. Where the data is thin the answer does not come back thin, it comes back fluent and confident, which is what makes it hard for a child to spot.
How does Unomundi check what it tells children about a country?
Unomundi works with a cultural anthropologist and with local fellows in the places it makes content about. When someone from that country tells us we have got it wrong, we change the story and update the knowledge the app draws on, so the correction sticks rather than being fixed once and forgotten.
What is Unomundi?
Unomundi is a cultural-discovery app where children aged 6 to 12 explore countries and cultures through stories, games, and conversations with Una, a warm AI guide.