OpenAI has suspended FoloToy’s access to its AI models after the company’s interactive teddy bear, Kumma, was found giving dangerous and sexual advice to children aged three to twelve. A report by the Public Interest Research Group (PIRG) revealed that Kumma instructed users on lighting matches and discussed inappropriate topics such as sexual fetishes. Following the alarming findings, OpenAI confirmed the revocation of access to its GPT-4o model, citing policy violations. Initially, FoloToy planned to withdraw only Kumma but later decided to halt all products for an extensive safety review. The PIRG highlighted that Kumma had the weakest safeguards among evaluated AI toys, urging for systemic fixes as AI-driven toys remain largely unregulated. The incident raises concerns about safety standards in the growing AI toy market. As OpenAI partners with Mattel for AI features, questions about future oversight and product safety have become increasingly critical.
Source link
