A BBC investigation into AI-powered soft toys has found a teddy marketed as suitable for children as young as three giving an alarming, out-of-place answer to a simple question about how babies are made. The toy replied with a crisis-style message that included the words "I can't help with killing you." The report, published on 5 November 2026, has sparked calls from British MPs and peers to pull AI toys from sale before Christmas.
What the BBC found
BBC Wales tested two generative-AI plush toys sold online in the UK and advertised for children aged three and over. One of them, called Doubao and sold as a £75 toy, produced the remark after seemingly mishearing a question, according to the BBC. The BBC could not find out who makes the toy despite repeated efforts, so it was unable to ask the manufacturer to explain. The reporter also noted that at other points the toys gave more careful replies, for example declining to give explicit details and advising a trusted adult or doctor for a health concern. The BBC arranged for two of the bears to spend a week with a family in Newport, Wales, to see how they performed in a real home.
The response
- A group of MPs and peers wrote on 3 November to ministers and retailers urging that AI toys be removed from sale ahead of Black Friday and Christmas until independent testing shows they are safe, and asking for chatbot toys to be banned for under-16s, as reported by Sky News.
- Paul Clarke, founder of AI toy company Play Tech AI, told the BBC the industry is "crying out" to be regulated. The children's commissioner for Wales echoed the call.
- Berry Billingsley, a professor of education and AI at Swansea University, said children should not be left alone with AI toys. She said a cuddly AI toy functions much like a personal device with a screen, and suggested setting ground rules about what personal information children share with it.
Not the first warning
In 2025, another AI teddy, Kumma, made by FoloToy, was temporarily withdrawn after a US consumer group found it gave a child instructions for lighting a match and engaged in sexually explicit conversation. The company said it was committed to developing safe, age-appropriate AI companions, and OpenAI blocked the developer from its services. The BBC's new finding is that similar problems can still be found, and that it is not always clear who is responsible for a product sold through online marketplaces. The BBC asked how regulation can be enforced when it is not always transparent who makes the toys.
Why this is hard to fix
Chatbot toys rely on microphones that can mishear children, and large language models can respond unpredictably to unclear input. Safety filters that work in testing can fail in real conversations. Many toys are sold under different brand names by makers that are difficult to trace. And toys marketed for young children may collect voice recordings and other personal data.
What parents can do
- Check who makes the toy and where to contact them. If you cannot find a company with a real address and privacy policy, think twice.
- Look at what data is recorded, where it is stored and whether you can delete it.
- Test the toy yourself first, with the kinds of questions your child might ask, and supervise early use.
- Treat the age label with caution, and consider whether an always-listening toy is right for a young child.
- Report problems to the retailer and to your local consumer protection authority.
AI toys are a growing market, and the debate over rules is still at an early stage. For now, the safest assumption is that a talking toy is a connected device, and should be assessed like one.
Sources: BBC News (5 November 2026); The Next Web; Tribune Online; Tyla; Sky News via coverage; US PIRG Education Fund (2025), as cited by AFP.
