We Make Sure Kids’ Pajamas Are Safe. Why Not Their AI?

Now even the Trump White House wants AI tested before it ships. A year after Vice President J.D. Vance dismissed AI safety as “hand-wringing” at the Paris Summit, the administration is reviving advanced review of cutting-edge models in the wake of Anthropic's Claude Mythos release. So why are we still letting AI chatbots into kids' lives with less attention to safety than we give kids’ pajamas?

Last week in Copenhagen, Common Sense Media convened European leaders at a summit to address that gap. We announced a new Youth AI Safety Institute that will crash-test AI chatbots and companions as rigorously as we crash-test cars.

The stakes are already visible at the kitchen table. Nearly three-quarters of American teenagers have used AI companions, and more than half use them regularly. A third say they'd just as soon talk to an AI companion as to a human. Too many of those conversations have led to emotional dependency, self-harm, or even suicide.

In the four years since the launch of OpenAI’s ChatGPT ushered in a new era of AI chatbots, the United States, the United Kingdom, and other nations have established a dozen safety institutes around the world. But those institutes focus primarily on national security threats, not the threat to kids and teens. The Youth AI Safety Institute is the world’s first independent lab to make sure the AI kids use is safe, age-appropriate, and in their best interests. From national security to youth safety, the key to AI safety is clear: Don’t trust — verify.

The Institute will bring significant new resources and technical expertise to set rigorous youth safety standards, test whether AI products meet those standards, and publish the results for parents to see. Several philanthropic and industry leaders — including the Walton Family Foundation, Anthropic, the OpenAI Foundation, and Pinterest — have helped launch the effort.

America has a long, proud national tradition of ensuring child products are safe. The Food and Drug Administration tests pediatric drugs and medical devices. The Consumer Product Safety Commission tests kids’ toys, toothbrushes, and pajamas.

It’s about time we set that same standard of safety for AI, the product teens use — and parents fear — the most.

History shows how independent testing can benefit consumers and companies alike:

• For half a century, the Insurance Institute for Highway Safety’s crash-test ratings have shown consumers which cars are safe and spurred automakers to make vehicles safer. Last year, the US had the second-lowest traffic fatality rate in history.
• Within a decade of the Consumer Product Safety Commission testing the fire resistance of pajamas, child-clothing-related fire deaths in the US plunged from one a week to two per year.

Underwriter Laboratories (UL) has saved countless lives since it started setting standards and testing for electrical shocks and fires in 1894. When UL introduced tougher standards for hairdryers in the early 1990s, hairdryer-related electrocutions dropped from an average of 13 deaths a year to one death over 7 years.

The Youth AI Safety Institute will apply a similar model — testing the AI products children use most, showing parents the results, and spurring a race to the top on safety.

The Institute will also do extensive research on another nagging worry: AI’s impact on education. A recent Common Sense Media national survey found that by a 52%-32% margin, American parents think it’s wrong to use AI in the classroom — while by a similar margin, children under 18 think it should be encouraged.

Standards, research, and testing will help us better understand AI’s risks and potential. But the work of the Institute is evaluation, not regulation. It will be up to states and the federal government to enact and enforce guardrails for youth AI products.

States are already showing the way. Last year, California passed a law requiring device makers and app stores to enforce age restrictions through a simple age signal that preserves child privacy. States red and blue have passed laws to protect minors from harmful AI companions. Earlier this month, the US Senate Judiciary Committee unanimously passed a bipartisan bill to do the same.

In this new era, kids and parents need all the help they can get. For two decades, kids and teens have been the crash-test dummies for social media. Let’s not make the same mistake on AI.