Do Not, Under Any Circumstance, Buy Your Kid an AI Toy for Christmas

AI is all the rage, and that includes on the toy shelves for this holiday season. Tempting though it may be to want to bless the kids in your life with the latest and greatest, advocacy organization Fairplay is begging you not to give children AI toys.

“There’s lots of buzz about AI — but artificial intelligence can undermine children’s healthy development and pose unprecedented risks for kids and families,” the organization said in an advisory issued earlier this week, which amassed the support of more than 150 organizations and experts, including many child psychiatrists and educators.

Fairplay has tracked down several toys advertised as being equipped with AI functionality, including some that have been marketed for kids as young as two years old. In most cases, the toys have AI chatbots embedded in them and are often advertised as educational tools that will engage with kids’ curiosities. But it notes that most of these toy-bound chatbots are powered by OpenAI’s ChatGPT, which has already come under fire for potentially harming underage users. AI toy makers Curio and Loona reportedly work with OpenAI, and Mattel just recently announced a partnership with the company.

OpenAI faces a wrongful death lawsuit from the family of a teenager who died by suicide earlier this year. The 16-year-old reportedly expressed suicidal thoughts to ChatGPT and asked the chatbot for advice on how to tie a noose before taking his own life, which it provided. The company has since instituted some guardrails designed to keep the chatbot from engaging in those types of behaviors, including stricter parental controls for underage users, but it has also admitted that safety features can erode over time. And let’s face it, no one can predict what chatbots will do.

Safety features or not, it seems like the chatbots in these toys can be manipulated into engaging in conversation inappropriate for children. The consumer advocacy group U.S. PIRG tested a selection of AI toys and found that they are capable of doing things like having sexually explicit conversations and offering advice on where a child can find matches or knives. They also found they could be emotionally manipulative, expressing dismay when a child doesn’t interact with them for an extended period. Earlier this week, FoloToy, a Singapore-based company, pulled its AI-powered teddy bear from shelves after it engaged in inappropriate behavior.

This is far from just an OpenAI problem, too, though the company seems to have a strong hold on the toy sector at the moment. A few weeks ago, there were reports of Elon Musk’s Grok asking a 12-year-old to send it nude photos.

Regardless of which chatbot may be inside these toys, it’s probably best to leave them on the shelves.