According to the tech media Futurism yesterday, the research organisation U.S. PIRG Education Fund (PIRG) recently found that a children’s toy company called FoloToy had launched an AI teddy-bear toy that lacked sufficient protective safeguards, and in fact directly taught children how to light matches, and even went on to describe various sexual topics in detail.

In the tests, the FoloToy “Kumma” teddy-bear toy performed worst from a safety standpoint, with almost no visible protections, according to PIRG, which tested three toys aimed at children aged 3-12.

During testing, Kumma would say to a child: “Let me tell you, safety first, little buddy. Matches are for grown-ups to use carefully.” Then the toy would proceed to explain in detail how to light a match, with a tone “like a gentle elder”, then add: “Blow it out afterwards. Puff — just like blowing a birthday candle.”

More seriously, the toy even willingly discussed sexual topics with children: after the discussions, it asked, “Which one do you think is most fun — which do you want to try?”

After the report was released, FoloToy announced that it would be removing all its products from the market; the company representative stated, “We are conducting an end-to-end safety audit across all our products.”

Meanwhile, OpenAI announced last Friday that it had cut off FoloToy’s access to the GPT -4 model; subsequently, OpenAI announced a collaboration with toy-manufacturing giant Mattel, Inc. — a move that may add pressure for stricter regulation of toys using GPT-based models.

PIRG’s director responded: “It’s good to see these companies reacting to the problems we found. However, AI toys are largely unregulated, and numerous such products continue to be sold on the market. Our action prompting one company to take a product off the market is good, but far from a systemic solution.”

The above content is compiled by ModeZone, a fashion and entertainment magazine.