AI firms should have to prove their products are safe for kids

Kaley sat in a Los Angeles courtroom this March while 12 jurors tried to calculate what her childhood was worth. They found Meta and YouTube liable for designing features that addicted her as a child, contributing to depression, body dysmorphia, and suicidal thoughts, and awarded her $6 million — half of it punitive, after deciding the companies had acted with malice.

To set the compensatory half, one juror explained, they estimated what she might have earned over a lifetime, weighing “what she’s gone through and what job she would be able to hold down.” Both companies are appealing.

The jury compensated Kaley for a diminished adulthood — for the person she would never become. That is the true cost of these platforms.

The verdicts keep coming.

In March, a New Mexico jury ordered Meta to pay $375 million for failing to protect children from sexual exploitation. In May, Meta, TikTok, Snap, and YouTube agreed to pay about $27 million to settle the first federal test case, brought by a Kentucky school district.

The tide is turning against Big Tech, but we should not be celebrating.

These verdicts arrive after children have already been irreparably harmed, and they do nothing for the children who will be harmed by whatever technology launches next. As new platforms capable of greater harm enter the market and are adopted at an accelerating pace, a reactive legal system is profoundly inadequate.

Instagram took two and a half years to reach 100 million users. TikTok took nine months. ChatGPT took two. By the time a jury assigns liability, the product has already scaled past the reach of any remedy. Smartphones entered classrooms two decades ago, and Pennsylvania lawmakers are only now passing legislation.

But to understand what is actually being taken, start with how a child’s mind develops.

Humans have the longest childhood of any species on Earth. That stretch of dependence gives our mental capacities time to develop through specific experiences. But that window does not stay open forever. The neural pathways solidify. What doesn’t form in childhood may never form at all.

So the more urgent question about any technology a child uses is not what it puts in front of them, but what it crowds out.

The average teenager now spends half their waking life looking at screens, and the number keeps climbing. That time has to come from somewhere.

Sleep has fallen to the lowest level on record. Only one in four adolescents meets the recommended federal minimum for physical activity. Over two decades, face-to-face socializing has been cut in half.

Yes, these platforms expose children to violence and predators, but the greater threat is the algorithms draining necessary developmental inputs.

Kaley started on YouTube at 6, Instagram at 9, Snapchat at 11. She is 20 now and living with the consequences, but she became an adult before artificial intelligence reached the market. The next child will not be that lucky.

I started teaching in 2009. My students made eye contact, talked to each other, and could hold their attention on a hard problem. By the end of that decade, they came in sleep-deprived and anxious, reaching for their phones at every moment. Earbuds in, eyes down.

Now, I watch students reach for the answer before they think for themselves.

Some technologies don’t just change how we live; they change who we are.

A car lets you travel farther; it does not make you incapable of walking. GPS goes a step further. Lean on it long enough, and your own sense of direction fades. But GPS takes only one skill. AI takes the rest. Offload your thinking to it, and the skills weaken. Offload it during childhood, and your capacity to think may never fully develop.

We learn through struggle. Our survival depended on the ability to solve problems, synthesize information, and imagine new possibilities. Remove the struggle, and with it goes the growth the struggle produces.

Students who lean heavily on ChatGPT show a sharp decline in cognitive effort and weaker memory retention — what Massachusetts Institute of Technology researchers call “cognitive debt.” A generation is offloading its thinking capacity, and with it, losing the stamina, curiosity, and desire for hard problems that only come from solving them yourself.

AI does the same to the capacity for connection. AI companions are patient, compliant, never bored, never hurt, never inclined to push back. Nearly two-thirds of teenagers now use AI chatbots. Nearly one in five high schoolers has had a romantic relationship with an AI, or knows someone who has.

But human relationships require struggle. A child who learns intimacy without friction will find it difficult to connect with a real person.

This is why the conversation needs to move beyond content moderation.

We already know how to handle products that endanger human health. We do not let a pharmaceutical company ship a drug and promise to fix the lethal side effects once the bodies turn up. We require proof of safety first, through years of testing, because we decided that corporate speed does not outrank human life.

For the systems now shaping the developing brain, we decided the opposite.

The man building the most powerful of them has said so plainly. The best way to make AI safe, Sam Altman has written, is “by iteratively and gradually releasing it into the world, giving society time to adapt and coevolve with the technology.” The adapting is done by users. Many of them are children. Do we want him experimenting on the minds of children?

The fix is not complicated.

Make companies prove their products are safe for developing brains before release, the way we make drug companies prove it before the pill reaches the shelf. Put the burden of proof where it belongs — on the trillion-dollar firm, not on the 10-year-old.

We already know how to handle products that endanger human health.

The objection to this shift is that a slowdown will hand the future to China.

But this year, Beijing moves to bar AI companies from offering virtual companion and intimate relationship services to anyone under 18. China has decided the best way to compete for the future is to protect the people who will inherit it.

There is progress at home.

In June, Florida sued OpenAI and Altman, alleging the company released a product that addicts minors, erodes their critical thinking, and counterfeits human compassion.

Pennsylvania should join the fight.

But these lawsuits won’t undo the damage done. Nothing short of proving these products are safe before they reach a child will be sufficient.

This should no longer just be a debate over shielding children from harmful content. We need to defend the baseline capacity to think, to connect, to be human. The opportunity cost for children is too steep.

A jury can put a dollar figure on an unfulfilled life, but it can never restore what is lost. If we don’t act fast, another generation will suffer.

AJ Ernst worked as a teacher and administrator in Philadelphia for 13 years and holds a doctorate in educational leadership from the University of Pennsylvania’s Graduate School of Education.

Inquirer Opinion Newsletter

Future product

Be the first to hear about a roundup of Inquirer columnists’ perspectives on what’s happening now in our city, our nation, and our world.