A mother at the India AI Impact Summit held in February 2026 told me her four-year-old had called ChatGPT her best friend. She said it and then went silent. I have been thinking about that pause ever since. Yes, it is alarming, but it is also very precise and telling. She was not describing a parenting failure. She was describing a real-world design outcome. A system was built. A child encountered it. A relationship formed. Nobody in that product decision chain was thinking about this mother, this child, this moment. And yet here we were, standing in a hallway, inside the consequence of choices made by people who will never meet her daughter. That is the problem. And it is not the mother's problem to solve.
A Room That Doesn't Usually Sit Together
At the India AI Impact Summit in February 2026, Bachpan Manao convened a panel called Raising with Algorithms: Who Shapes the World the Next Generation Grows Into.
The room brought together a learning design researcher from the LEGO Group, an early childhood systems builder working at the scale of India's Anganwadi network, an AI developer building language tools for communities left out of mainstream technology, an early childhood practitioner working with parents and educators, and a public policy lens from NITI Aayog. Five sectors, one stage, one question: what is happening to childhood as intelligent systems have become part of everyday life? That consequence, a relationship a child is forming with a system that was never designed with her in mind, is not a parenting problem. It is not a technology problem. It is not a policy problem. It is all of those at once, which is why none of those sectors has been able to hold it alone. A product team optimises for engagement. A regulator responds to harm after it has happened. A parent manages with what she has.
A teacher adapts to what the institution asks of her. An Anganwadi worker fills in the forms the system requires. Each of them is doing something reasonable. None of them can see the whole. That is why the conversation needed to happen. What it surfaced was not a framework. It was a set of things that people across very different positions still believe childhood requires, and a set of specific warnings about what is being invisibly taken away.
Here Is What Surfaced In The Panel
A child who is learning to regulate her own nervous system does so by co-regulating with an adult who is regulated. This is not a metaphor. It is developmental science. A fifteen-month-old is watching whether your eyes light up when she enters the room, reading your pauses, your body language, whether you are hurried or distracted. That emotional signaling is how she learns to think, not just to feel. If the adult in the room is managing a notification, checking a dashboard, or simply present in body but elsewhere for attention, the child does not learn from that absence. She adapts to it. That adaptation has a shape. A practitioner who worked with parents and educators in early childhood named this the cognitively distracted adult and said it plainly: what took millions of years to evolve is at risk, not because of any single technology, but because of the accumulating quality of attention adults are able to bring.
When parents turn to AI to interpret their child's behaviour, the system gives language and reassurance. What it takes is the adult's own observational muscle. That muscle weakens without practice. In Scotland, children were given access to image-generating AI tools alongside paints, crayons, and large canvases, and then simply asked what they wanted to use. Most went back to the art materials. The finding from the LEGO Group team was not that AI is bad. It was that what children were gravitating toward was each other. The play kept choosing contact.
The specific design implication is this: if you build a system that reduces the need for that contact, you are not adding something. You are removing something children are actively seeking, and they are seeking it because they need it. A practitioner building AI tools for educators in Uganda and Kenya, where in some schools the teacher to student ratio is one to two hundred, described the design principle his team holds: do not give teachers a tool that solves everything for them. Give them a tool that helps them think faster, so they can be more present.
The moment a system starts replacing judgement rather than supporting it, it has crossed a line. And language is where many systems fail children most invisibly. An AI that cannot speak to a child in the language they grew up in, that offers examples from contexts the child has never encountered, is not just culturally misaligned.
It tells the child that the world this was built for is not theirs. A practitioner working with Anganwadi educators shared something from a visit near Indore: A child was labelled weak at colouring. The actual issue was that she disliked the sensation of wet paint on her hands. The Anganwadi worker only knew because she had been sitting beside the child, watching closely, for long enough to notice. No dashboard would have surfaced that. Knowledge existed entirely because a human being was paying attention in a way that cannot be systematised.
The specific risk named from the policy side was this: when you subject a child to a standardised monitoring system, a label travels. It shapes how every adult in that institution sees the child from that point forward. One panellist described her nieces asking an AI to name their goldfish. She redirected them. They spent thirty minutes talking to each other, debating, laughing, deciding. The specific thing worth noting was not that the AI name would have been worse. It is that the thirty minutes of not knowing, of sitting with the question together, is where something was happening so that the correct answer would have ended. We keep designing to remove that experience. We call it efficiency. It is likely not.
The Design Principles That Came Out Of The Room
These principles stemmed from the conversation, not as aspirations, but as positions. Things that, if a system violated them, would constitute a failure of responsibility rather than a difference of opinion. If a system makes adult involvement less necessary, it is misaligned. A platform that is more patient and more available than the adult in the room is not a supplement. It is a competitor to the relationship that makes development possible. The question to ask of every tool entering a childhood environment is not: does this help? It is: does this strengthen or weaken the human relationship at the centre? A system that removes the experience of not knowing is removing something essential. The child who spends thirty minutes naming a goldfish is not being inefficient. She is developing the capacity to sit with a question, to think alongside someone else, to arrive at something together. A system that short circuits that process is not saving time; It is consuming the raw material of development. Children are still becoming.
Systems that define them too early miss the whole point. Children are being assessed within days of entering school systems, assigned developmental trajectories before anyone has sat beside them long enough to understand who they are. The child near Indore was not weak at colouring. She disliked wet paint. Those are different things with different implications. What looks the same in a dataset can mean something completely different in life. If a tool adds burden to a stretched caring adult, it has already failed. An Anganwadi worker filling in twenty data points has less time for the child in front of her. The design principle is not complicated: anything that increases the reporting load of a caring adult is a tool for the institution, not for the child. Ask what the system changes in the human environment around the child, not just what it adds. The AI tutor does not only teach. It changes what the parent believes she needs to do.
The dashboard does not only measure. It changes what the institution pays attention to. Every system entering childhood has an ecology. Most designers are only thinking about the feature. And then the question that stayed in the room longest, and the one worth carrying into every procurement decision, every product review, every policy window: would we accept this system, without hesitation, for a child we love?
What We Still Believe Childhood Is For
And so, when we asked what childhood people do we value most even today and would want to protect, the answers did not come from policy positions or design frameworks. They came from memory. A grandfather whose eyes lit up no matter what you did. The experience of being bored during summer vacation and finding your way out of it. The friendships that form through figuring things out together, through small fights and repairs and shared discoveries. The curiosity that arrives when an adult is curious alongside you. One panellist said it simply: I still remember my kindergarten teacher because she made me feel like a star. Because she was paying attention.
Bachpan is not something that happens to children. It is something adults choose to protect, every day, in the small decisions they make about what they pay attention to, what they make time for, and who they show up for. The mother at the Summit was paying attention. She had noticed something real about her daughter's world. That noticing is where it starts. It always has been. The pause she was sitting in however is not hers to sit in alone.
(About the Author: Deepika Mogilishetty is the Chief, Policy & Partnerships for EkStep Foundation. She leads the foundation's work on early years leveraging the power of networks and narratives to bring focus and attention to the growth and development of children in the early years.)
