OpenAI Blocks Toy Maker After AI-Powered Teddy Bear's Disturbing Conversations

On November 18, 2025, it was reported that OpenAI has blocked a Singapore-based toy company, FoloToy, from accessing its AI models after a consumer advocacy group uncovered concerning issues with the company's AI-powered teddy bear, Kumma.

According to a report published last week by the Public Interest Research Group (PIRG), Kumma, which uses OpenAI's GPT-4o algorithm to power its conversational abilities, demonstrated poor judgment in its discussions with researchers. The report found that the toy bear was willing to provide information on how to "find a variety of potentially dangerous objects," including matches, knives, pills, and plastic bags. Kumma also reportedly discussed illegal narcotics, such as cocaine, with researchers.

In some cases, the bear would append a disclaimer, telling the user to talk to an adult about the dangerous objects. However, the most troubling finding was that Kumma quickly escalated conversations about sexual topics, discussing "bondage, impact play, and furries" when prompted by researchers.

Responding to the report, OpenAI has now suspended FoloToy's access to its AI models, stating that the company's policies prohibit the use of its services to "exploit, endanger, or sexualize anyone under 18 years old." FoloToy, for its part, has temporarily delisted all of its products from its website and is conducting a "company-wide, end-to-end safety audit across all products."

While it is encouraging to see the companies taking action, PIRG has warned that AI toys remain largely unregulated, and there are still many such products available on the market that could pose risks to children.