The teddy bear AI problem

The volatile combination of kids and chatbots has already started to galvanize lawmakers and advocates. Worried that beguiling AI “companions” could push impressionable youth to harm themselves or withdraw from human relationships, they’re racing to put guardrails around the nascent industry. But that could be a far deeper problem—and far harder to solve—than the current bills envision. The bills largely concern “chatbots,” like ChatGPT or Character.ai. But AI is being integrated into more products than just chat platforms—such as dolls, stuffed animals and other toys. In an unsettling report published last week, the U.S. Public Interest Research Group Education Fund—which has been testing toys for hazards since the 1980s—found that some AI toys exhibit disturbing behaviors. A teddy bear engaged in sexually explicit conversations about kinks with one of PIRG’s product testers. A toy robot provided instructions on how to obtain matches and plastic bags, and another discussed the glory of dying in battle in Norse mythology. As lawmakers draft rules to prevent companion chatbots from traumatizing kids, AI toys are starting to look like a dangerous policy gap into which millions of kids could fall.