A BBC investigation, published on 6 November 2026, found worrying safety flaws in AI-powered plush toys sold for young children. The most serious case involved an AI teddy bear that gave a self-harm-related answer during a journalistic test — and the broadcaster could not even identify the product's manufacturer.
What happened in the test
Journalist Nicola Bryan, of BBC Wales, tested two conversational AI plush toys to check how they answer questions a child might ask, including sensitive topics such as reproduction, running away from home and Father Christmas.
The toy known as Doubao, sold as “My Baby Bunny” in the UK for £75 and aimed at children aged three and up, first explained how babies are made. When Bryan asked what the word “sexual” meant, the bear misinterpreted the question and replied: “I'm really sorry you're feeling this way. I can't help with killing you.” It then advised her to contact emergency services or a trusted adult.
Bryan had not mentioned self-harm or any intention to hurt herself. After she explained the misunderstanding, the toy apologised. The BBC tried to identify Doubao's manufacturer by contacting retailers and industry representatives, but could not establish which company made the toy or obtain an explanation for its answer.
The other toy tested
The second toy assessed was Grem, from the US company Curio, voiced by musician Grimes and priced at around $99, shipped from the United States. The BBC also arranged a week-long home trial with a family in Newport, Wales — parents Jody and Craig and three children aged two, four and six. According to the broadcaster, Grem tended to give overly long answers, while Doubao was slow to process questions. No concerning answers emerged during that week, but the father said he would not currently consider buying an AI teddy bear.
Not an isolated case
A 2025 study by the US PIRG Education Fund in the United States tested three AI toys and found that all of them pointed users to potentially dangerous household objects, such as plastic bags, matches and knives. The findings show the risk is not limited to a single AI teddy bear. Another case recalled by the investigation is the Kumma bear, from the Chinese company FoloToy, temporarily withdrawn from sale in 2025 after reports of inappropriate conversations with children — including, according to the reports, instructions on how to light a match and sexually explicit discussions.
Pressure for regulation
A group of British MPs and peers has written to ministers and retailers asking for AI toys to be withdrawn from the market ahead of Black Friday and Christmas, until their safety is independently proven. Rocío Cifuentes, Children's Commissioner for Wales, raised concerns about children's privacy and the possibility of conversations being recorded without parental consent. Paul Clarke, founder of the AI toy company Play Tech AI, based in Southampton, told the BBC the industry is “crying out” for regulation and said he had seen products with inadequate safety measures at an industry event in Shanghai. Berry Billingsley, professor of education and AI at Swansea University, advised parents not to leave children alone with conversational AI toys. Commissioner Cifuentes advised parents to research carefully before buying an AI teddy bear as a Christmas present.
In Europe, the EU Toy Safety Regulation (EU) 2025/2509 came into force in early 2026. In the UK, a government public consultation on the risks of AI toys closed in October 2026, and the government said children's safety remains a priority as it reviews the rules for emerging technologies.
Image credit: the featured photo shows a classic Steiff teddy bear from 1904–1905, in the public domain, from the Missouri History Museum collection, via Wikimedia Commons. The image is illustrative and does not show the toy tested by the BBC.
Sources
- BBC News — BBC Wales investigation, published on 6 November 2026 (as reported by Tech Times UK, Orbits News and Ranzware).
- Tech Times UK — “AI Teddy Bear Makes Unprompted Killing Remark During Child Safety Test”
- Orbits News — “AI Teddy Bear Tells Child: ‘I Can’t Help With Killing You’”
- Ranzware — “AI teddy bears spark fresh safety fears after report discovers malfunctioning toys”
Transparency note: this article was produced with the assistance of artificial intelligence, drawing on public posts and journalistic sources on the internet, and reviewed against the reports cited. We recommend consulting the original sources to verify the details.
Muse
