There's only one way to make AI safe for kids

OpenAI has launched ChatGPT for Teens, complete with various safety features. But this implies that the main concern of kids using AI is their access to damaging content. If only.

Don’t get me wrong. The AI-assisted suicides, deepfakes and sexually explicit conversations with and for children are - to be frank - horrific.

But as a mother, my deepest concern is what AI use does to teens’ ability to think critically and develop socially. The only way to address those worries is not to toggle with the app, but to limit kids’ exposure to it entirely. 

The new version of ChatGPT for Teens is targeted for kids ages 13 to 17.  It includes content restrictions around sexual, violent and self-harm topics. It also promises to curb AI use for homework, guiding students towards next steps instead of providing straight-up answers.

This is all well and good. I’m with the 80% of American parents who support more controls on AI for kids, including parental permissions and alerts.  

But when it comes to the efficacy of such protections, I’m not holding my breath. We’ve seen how hard it is to impose age controls on social media. And age verification for ChatGPT for Teens has no legal force behind it. 

Even more concerning, a version of the app marketed towards teens may make parents (or educators) feel more comfortable with children accessing it. That could increase the use of AI by adolescents. They are already talking too much with AI chatbots as it is. 

And of course, whatever OpenAI blocks is still available elsewhere. A new survey by Common Sense Media’s Youth AI Safety Institute found that 63% of teens are using AI explicitly to get homework answers. Testers asked Google Search to do their homework (math sets and essays) and it completed every assignment. And if you or your child are wondering, there is no way to turn Google AI’s assistant off. 

Already, our children’s brains are being changed by exposure to these tools. A recent study out of the University of Hong Kong and Stockholm University looked at 27,000 students ages 12–18. The researchers found that after six months of using AI, students saw their average homework score go up. But come exam time, the same kids scored 20% below their classmates who had not utilized AI. 

This comes atop other early research suggesting a decline in cognition from AI use. A 2026 literature review of 400 studies by the Brookings Institution, where I am a fellow, concluded “the risks of utilizing generative AI in children’s education overshadow its benefits.” 

This is to say nothing about what AI is doing to children’s relationships with their parents and with each other. In a 2025 survey by Thorn, a child safety nonprofit, nearly a third (29%) of minors had turned to AI chatbots for guidance on a sensitive or intimate topic - a friction-free alternative to seeking their parents’ and peers’ counsel.

To be sure, we are at the very beginning of understanding what AI might do to human cognition. Maybe there will be some incredible upside, some unique advantage for kids in particular.  

But try telling that to this battle-weary bunch of moms and dads. In the last decade, we’ve had to navigate kids’ relationships with smartphones, social media and screen-based schooling. And now artificial intelligence.  

Each time, policymakers and Big Tech and schools have erred on the side of sorry rather than safe. Each time, reams of after-the-fact data have confirmed what common sense could foretell: Addictive technology and developing brains are a dangerous mix. 

So yes - if OpenAI wants to help address the problem they created, that’s progress. I’ll cheer on any attempt to put more guardrails on Big Tech when it comes to kids. 

But the most surefire way in schools and homes to protect children from AI - in this critical time of development for the technology and a critical time of development for our children - is to limit access to it completely.  

In school, it makes sense to return to taking exams with pen and pencil. Homework should be done on paper, or at the very least, a device that can block AI - which means companies will have to grant parents and teachers the ability to do so.

At home, parents should keep trying to delay personal screens from being used by their children as long as possible. Bring back reading, time outside, and playing with friends instead. 

And call me reactionary, but look at apps designed for children with a particular level of skepticism.

This column reflects the personal views of the author and does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners. Abby McCloskey is a columnist, podcast host, and consultant. She directed domestic policy on two presidential campaigns and was director of economic policy at the American Enterprise Institute.