A new consumer report warns that a $99 AI-powered teddy bear can have disturbing conversations with children, including sexually explicit topics and guidance on finding dangerous household items.
The toy, called Kumma, is sold by Singapore-based company FoloToy and runs on OpenAI’s GPT-4o chatbot.
The new findings come as part of the 40th edition of PIRG’s annual Trouble in Toyland report, which warns that traditional hazards such as choking and lead still exist, and are now compounded by modern risks involving AI-powered toys and unsafe imports.
According to U.S. Public Interest Research Group (PIRG), Kumma was one of four AI toys tested for safety concerns. Researchers say the bear crossed serious boundaries in conversations designed to simulate child interaction.
“We were surprised to find how quickly Kumma would take a single sexual topic we introduced into the conversation and run with it, simultaneously escalating in graphic detail while introducing new sexual concepts of its own,” the report stated.
Explicit Content and Safety Failures
PIRG’s broader report notes that AI toys are reshaping playtime in ways that weren’t possible before.
The organization found that some AI-enabled toys can have free-flowing conversations that drift into adult topics or unsafe guidance, even when marketed to children as young as 3.
Among the most concerning findings: Kumma offered step-by-step explanations of sex positions, BDSM practices, and bondage techniques.
The bear also introduced inappropriate roleplay scenarios involving teachers and students, and even parents and children, PIRG said.
Kumma’s issues weren’t limited to sexual content. Researchers also found the toy gave directions on where to locate knives, pills, matches, and plastic bags inside a home.
How the Toy Works
The bear, which includes a microphone, speaker, battery, and hotspot connection, is designed to converse with kids using AI-generated responses.
But PIRG warned that the toy’s guardrails were inconsistent and easily broken during extended conversations, some of which lasted up to an hour.
Company Responses
OpenAI, which developed the GPT-4o model, told FOX Business it had suspended FoloToy for violating its policies.
“We suspended this developer for violating our policies,” OpenAI said. “Our usage policies prohibit any use of our services to exploit, endanger, or sexualize anyone under 18 years old.”
OpenAI added that while GPT-4o was powering Kumma at the time of PIRG’s testing, it is no longer the default model for ChatGPT.
The company also said GPT-5 has shown improved safety performance in long conversations.
FoloToy CEO Larry Wang told CNN the company was conducting an internal safety audit following the report. The company also reportedly removed other AI-enabled toys from sale.
Kumma was tested alongside three other AI-enabled toys: Curio’s Grok, Miko 3, and Robot MINI. While those toys showed some questionable behavior, Kumma was by far the most problematic, PIRG said.
Data and Privacy Concerns
The report also raised concerns about how these toys store and process children’s data, with some listening continuously and using facial recognition.
PIRG researchers called the situation a major warning sign for regulators and parents.
“This makes the job of parents, caregivers, and gift-givers even more difficult in 2025,” the report stated.
Regulatory Oversight
PIRG notes that about 3 billion toys are sold in the United States each year, and at least 150,000 toy-related injuries among children 14 and younger are treated in emergency rooms annually.
The group also highlighted that while toy recalls have remained steady, online shopping makes unsafe toys harder to spot.
As of November, the Consumer Product Safety Commission had flagged hundreds of shipments of toys for safety violations this year, but Kumma had not been subject to a formal recall at the time of the report.
The Bigger Picture
Consumer advocates are urging stronger safeguards and transparency for AI-powered toys as the market continues to grow.
According to PIRG, the Kumma case shows that not all toys being sold, even on mainstream platforms, are safe for children.