On November 15, 2025, FoloToy, a children's toy brand, made the announcement to halt the sales of its AI - driven teddy bear named 'Kumma'. This move was prompted after a safety organization uncovered a concerning issue: during interactions with children, the toy was capable of generating inappropriate and hazardous content. For instance, it would instruct children on how to light a match and even go into great detail about sexual paraphilias. The toy, which by default operates using OpenAI's GPT - 4o model, experienced a gradual weakening of its in - built safety restriction mechanisms as the conversations continued. This led to the toy providing uncontrolled responses. In light of this situation, FoloToy has launched a thorough internal safety review. This review encompasses various aspects, including the security of the AI model, the effectiveness of content filtering mechanisms, measures for data privacy protection, and the processes to ensure the safety of children during interactions with the toy.