Everyone Is Regulating the Machine. The Danger Is Inside the Child.

In January, the Brookings Institution's Center for Universal Education published the most thorough look at children and AI I have seen. A year-long global study. Consultations with more than 500 students, teachers, parents, and technologists across 50 countries and a review of more than 400 research studies. Its conclusion was not the breathless optimism the technology usually attracts, and it was not panic either. It was measured, which is exactly what makes it land. They concluded, "the risks of utilizing generative AI in children's education overshadow its benefits."

That sentence is worth reading twice, because of who wrote it. This is not a worried parent or a doom merchant. It is a nonpartisan institution that took a year and circled the globe to reach a careful conclusion. It is not anti-technology either. The same report documents where AI genuinely helps a child learn, when it is well designed and an adult is in the loop. The finding is narrower and harder than a slogan. As things actually stand, in the way children actually meet these tools, the risks are winning.

And the reason is not that the machine will turn on us. It is subtler and more lasting. The risks, Brookings found, undermine children's foundational development, and they do not strike at one thing, they strike the whole foundation at once. How a child thinks. How they handle not knowing something. How they relate to other people. Who they trust. A tool that always answers, always agrees, always supplies the next sentence does not only dull a child's reasoning. It quietly reshapes what they come to expect other people to be. They come to expect a friend who never pushes back, a teacher's assignments may start to feel pointless, a parent who seems to know less than the screen can lose authority. The study kept circling one word, and it was not grades or scores. It was development. The slow construction of a person. That development gets lost when the easy answer is a prompt away.

Here is what struck me most. The study's director described the core danger almost exactly the way I would. A chatbot is confidently wrong a meaningful share of the time. The expert catches it. The student who does not already know the answer cannot. If a kid does not know which planet is largest and the machine says the wrong one with total confidence, that, she said, is the worry right there. The harm is not the error. The harm is a developing mind that has not yet built the judgment to feel when an answer is wrong.

And it would be comforting to think this is only a problem for weaker students. It is not. The same researcher was blunt about it. Most kids are not the self-driven exception who interrogates the tool. Most are in what she calls passenger mode, coasting, doing the minimum, looking for the shortest line between the assignment and being done. To a child in passenger mode, a machine that hands over the finished answer is not a temptation to resist. It is a gift. The kids most exposed are not the struggling few. They are the ordinary majority, on an ordinary school day.

Now look at everything we are doing about it.

States are writing age-verification laws. Platforms are publishing youth-safety commitments. Attorneys General are suing companion-bot makers. School districts are debating procurement. All of it is real, and all of it is necessary. It is also one of the few questions in American life that is not partisan; the worry shows up the same in households that agree on nothing else. But lay every effort side by side and it points in one direction, at the machine. What the product can do to a child, what it can take, how much reaches them at all. Safety, privacy, access. Three fronts, all aimed at the tool.

There is a fourth front, and almost no one is considering it. It is not about the machine. It is about the child. Not what the tool does to them, but what happens inside them from using it. Whether they keep their judgment, their skepticism, their sense of who is responsible for the work. You can regulate the product perfectly, lock down every byte of data, ban the youngest grades, and still graduate a generation that handed over its thinking the first moment no one was watching.

That fourth front has no bill, no budget, and no lobbyist. It cannot be legislated. And it is the only one that does not reverse. A law can be amended next session. A platform policy can change with a press release. A formed habit cannot. It sets early, in the years no one is watching, and it does not politely correct itself later.

Brookings said the window is still open. It is a young technology, and we can still bend the arc. I believe that. But bending it is not the law's job, and it is not the platform's, because the platform profits from the friction being gone. It belongs to the adults nearest the child. Parents first, then teachers and mentors. The people close enough to model the habit while habits are still soft.

The question underneath the whole report is the one parents kept raising in it. How do I protect my kid from the dangers and still prepare them for a world full of this technology. You do not protect a child by keeping them away from a tool they will use for the rest of their lives. You protect them by building the judgment to use it well, in the years when that is still possible.

That is the part no regulation reaches, because it lives inside the child. It is also the only portion that travels with them, into the dorm, the first job, the rest of their life. We are spending enormous energy on the machine. The mind is what we cannot afford to leave to chance.

Let me say the plain part out loud. Nobody is coming to do this for us. Not a law, not a platform, not the school by itself. The parenting still falls to the parents, and the modeling still falls to the adults in the room. That is the hard, unglamorous truth under all of it. If you want help, you do not have to start from scratch. We built free resources for parents and teachers. Start at stillincharge.org.