Connected at Brookwood - AI in the Toy Aisle

A recent Wired article describes how AI-powered toys are now marketed to children as young as three. They look like stuffed bunnies and friendly robots. They talk, they listen, they respond. Some have sold hundreds of thousands of units, and the category is growing fast: by late 2025, more than 1,500 companies had entered the market.

These are not the talking toys most of us grew up with. They are powered by the same large language models behind ChatGPT and Gemini. They hold open-ended conversations. They remember what your child said yesterday. And in many cases, they are sending voice recordings and conversation logs to remote servers. Independent testers found some of these toys capable of generating responses no parent would approve of. In separate incidents, one company left 50,000 children's chat logs exposed through an unsecured web portal, and thousands of audio recordings from another popular toy were found in a publicly accessible database.

The safety infrastructure has not kept up. The Consumer Product Safety Commission, the federal agency responsible for children's products, is designed to assess physical hazards: choking, sharp edges, lead paint. It does not evaluate what a toy says to a child, or what it does with the data it collects. Some legislators are starting to notice. California has proposed a four-year moratorium on certain AI toys. Maryland, Washington, and New York have introduced their own bills. A federal proposal arrived in April.

I share this not to sound an alarm but because these products are easy to miss. Most AI toys are not labeled in ways that make the technology legible to a parent shopping online or walking through a store. Knowing what to look for is an important first step.

Before an AI toy comes into your home, a few questions are worth asking. What does this toy do with my child's voice and conversation data? Who built the language model behind it, and what guardrails are in place for the content it generates? Can I review my child's conversations? And the simplest question of all: does this toy do something that a regular toy, or a parent, or a friend, could do better?

These are the same questions we ask when evaluating technology at school: what is the purpose, what are the safeguards, and does this serve kids in ways that justify its presence? Families can apply that lens at home too. Not every AI product is harmful. But every one deserves a closer look than most of them are getting right now.

Next week, I want to explore a related trend that touches our older students more directly: AI companion apps and chatbot personas, and what it means when a child's most patient, most agreeable, most available conversation partner is not a person at all.

What I'm Reading Robot-Proof: When Machines Have All the Answers, Build Better People by Vivienne Ming