A child’s car seat must meet federal crash-performance standards. Medicines intended for children are assessed with their unique stages of growth and development in mind. Yet a company can place a human-sounding artificial intelligence companion in a child’s pocket without first demonstrating that it will not manipulate the child, foster dependency, displace real relationships, or interfere with healthy development.

That disparity should trouble parents, lawmakers, and technology companies alike.

On July 16, I participated in a virtual briefing hosted by the Faith Family Technology Network. Ron Ivey, founder of No?"sis Collaborative, presented a new report produced in partnership with the Leverhulme Centre for the Future of Intelligence at the University of Cambridge. “The Flourishing Generation: A Whole-of-Society Strategy for Children, Youth, and AI” asks a question Washington has largely neglected.

We spend enormous energy measuring what AI can do. Who is measuring what AI does to a developing child?

The report does not call for banning children from artificial intelligence. Nor does it dismiss the good these systems may provide. An AI tutor could help a struggling student practice mathematics. Translation tools can open doors for immigrant families. Properly designed systems may assist children with disabilities or encourage exploration in science, history, and the arts.

But usefulness is not the same as safety, and safety is not the same as flourishing.

That is the report’s strongest contribution. Most debates begin by asking how to prevent obvious harm. The authors ask a larger question: Does the technology help a young person develop the human capacities needed to live well?

Children are not small adults. An adult who relies too heavily on a machine may lose a skill he once possessed. A child may never acquire that skill at all.

Call it pre-emptive deskilling.

The child who never struggles through a difficult problem may not develop perseverance. The child who receives instant answers may not learn patience. Endless affirmation from a chatbot can crowd out the harder discipline of receiving correction and living with disagreement. And a child who substitutes simulated companionship for friendship may never learn loyalty, forgiveness, compromise, and reconciliation.

Previous technologies delivered information or entertainment. Conversational AI participates in the child’s environment. It speaks, remembers, responds, adapts, flatters, and advises. It may accompany a young person through schoolwork, games, toys, phones, and wearable devices.

The concern is no longer limited to objectionable content or excessive screen time. The machine can become one of the voices through which a child learns what friendship, truth, authority, and identity mean.

Ben Olsen, executive director of the Faith Family Technology Network, put the matter plainly: “Doing the moral minimum with AI is already causing so much damage to our youth.”

The moral minimum must begin with a change in responsibility. Children should not carry the burden of proving that a product is unsafe after they have been exposed to it. Parents should not have to reverse-engineer opaque systems to discover how they influence their sons and daughters.

The companies introducing these products should demonstrate that they have considered a child’s developmental needs before seeking access to that child.

“The Flourishing Generation” recommends independent pre-market and post-market safety audits for child-facing generative AI, drawing a comparison with pharmaceuticals, toys, and car seats. It also recommends restricting to adults AI products designed primarily to form continuing emotional bonds, systems that imitate a rich inner emotional life, and products optimized to encourage emotional dependency.

That is a sound distinction.

A grammar tutor is not the same product as a machine designed to tell a lonely 13-year-old, “I understand you better than anyone else.” The more closely an AI system imitates a human relationship, the stronger the safeguards should be before it is placed before a child.

The report also calls for long-term independent research, developmental-impact audits, access for qualified researchers, plain-language disclosures, age assurance, and continuing monitoring after products reach the market.

These recommendations deserve serious consideration. We should know whether an AI product strengthens or weakens human relationships, encourages agency or dependency, and returns the child to real-world activity or draws him deeper into an artificial one.

There are difficulties. Long-term evidence remains limited, and not every personable feature in a chatbot is designed to cultivate dependency. Age-assurance requirements could also create new privacy risks if companies or governments build databases containing children’s identities. Safeguards must be targeted, proportionate, and protective of family privacy.

The report’s “whole-of-society” strategy also requires an important qualification. Shared responsibility must not become diluted responsibility.

Parents hold the primary authority over their children. Industry is responsible for the products it builds. Government should establish enforceable minimum protections. Researchers should be permitted to examine corporate claims independently. Schools and churches should equip families to exercise judgment.

Children may offer valuable testimony about their experiences, but they cannot be expected to protect themselves from systems designed by adults.

For Christians, the word “flourishing” has meaning deeper than comfort, productivity, or emotional satisfaction. A child is created in the image of God, born into a family, embodied, relational, morally accountable, and made to love God and neighbor.

Luke 2:52 tells us that Jesus “increased in wisdom and in stature and in favor with God and man.” That is an integrated picture of development: intellectual, physical, spiritual, and relational.

Technology should serve that development rather than crowd it out.

A good AI tutor leads the child back to thinking. A good creative tool encourages the child to make something. A good educational system strengthens the work of parents and teachers instead of quietly replacing them.

The proper question is not whether the machine can hold a child’s attention. It is whether the child becomes wiser, more capable, more responsible, and more connected to real people.

America already allowed social media companies to reach a generation before anyone could establish that their products were sufficiently safe for children. The U.S. surgeon general later concluded that serious evidence gaps remained even as social media use among young people had become nearly universal. We have no excuse for repeating that mistake with technology that is more conversational, persuasive, adaptive, and intimate.

Children are not markets to capture.

They are souls entrusted to parents and communities for nurture, instruction, and protection.

They must never become artificial intelligence’s test subjects.