An AI toy aimed at children as young as three has prompted major questions about AI, child safety and corporate responsibility following a BBC investigation which revealed a disturbing response and an unknown manufacturer.
An AI teddy bear sold to UK consumers has apparently been programmed to say to a BBC journalist “I can’t help with killing you” when asked about human reproduction. The toy, Doubao or My Baby Bunny, is aimed at children from as young as three and is available online for around £75.
The incident (reported by BBC News on 6th November 2026) became even more troubling when it was found that the company that made the toy could not be established. The central issue is far more than just an inappropriate AI response. If an AI product can talk directly to a child, and the manufacturer can’t be readily identified, who is then responsible when the technology fails?
What did the AI teddy bear actually say?
During the course of the investigation BBC journalist Nicola Bryan was testing how the AI toys would respond to some of the more sensitive questions a child might ask, for instance how babies are made. When Bryan asked Doubao what the word “sexual” meant, Doubao completely missed the point and just gave a completely unrelated answer about self-harm.
It reportedly said “I’m really sorry you’re feeling this way. I can’t help with killing you.”
The toy then suggested calling an adult or contacting emergency services. Bryan had absolutely not had any suicidal ideation or any thoughts of self-harm at all. Once she’d clarified that, the toy apologised. The response is consistent with an AI safety feature which has mistakenly taken a normal query to be an emergency. We have no idea what the actual cause is as we don’t have the manufacturer’s technical explanation.
Character.AI bankruptcy exposes why millions of users couldn’t save California-based company →
Perhaps the most significant finding of all was that the BBC couldn’t work out who actually made Doubao. Investigators did a bit of online searching, rang the retailer, spoke to a number of industry people and got in touch with a number of companies who might have made the toy. Nothing gave them a definitive answer as to who the manufacturer was.
This doesn’t prove the manufacturer deliberately hid its identity or broke the law. It does though highlight a very serious accountability issue in a developing consumer tech industry. Traditional toys are subject to the normal product safety requirements. Conversational AI toys present an additional set of risks as their responses can vary depending on the questions the child asks, the context in which they are asked and the software underneath.
A product that seems innocuous in one instance can give an inappropriate response in another.
UK lawmakers demand stronger AI toy regulations
The probe is happening at the same time as UK MPs are pushing for greater regulation of AI products aimed at children. BBC reports that a group of MPs and Lords members have called on the Government and retailers to pull AI toys off the market ahead of Black Friday and Christmas until their safety can be independently verified.
Paul Clarke, founder of Play Tech AI in Southampton, told the BBC the industry was “crying out” for regulation. His warning is symptomatic of a wider commercial issue. A company investing in responsible AI development could find itself up against manufacturers who are not necessarily operating with the same level of transparency when it comes to safety.
The UK Department for Business and Trade ran a review in July 2026 looking at how the existing toy safety rules apply to new technologies, in particular AI-enabled products. The consultation closed on 6th October, though the regulatory issues are still very much open.
These Dangers of Roblox Are Scaring Parents Across the Country →
The incident doesn’t establish that all AI toys are unsafe. Manufacturers maintain that conversational technology can be used to support learning, storytelling and interactive play. Well designed systems can offer a range of educational benefits whilst incorporating appropriate age-related safeguards.
However Berry Billingsley (Professor of Education and AI at Swansea University) was warning in the BBC investigation that kids shouldn’t be left alone with conversational AI toys. Unlike traditional stuffed toys these can process children’s speech, generate unpredictable responses and potentially pass on information via the connected services.
Parents therefore need to look at what a toy might say as well as who is behind the technology that runs it and how the children’s information is protected.
The bigger question of course is who is to be held accountable for AI?
Doubao’s rather odd comment may in the end just be a case of a technical misreading rather than a genuine safety failure. The fact that it can’t identify its manufacturer is a far more fundamental issue though.
As AI moves into children’s bedrooms, into classrooms and into all sorts of everyday products, regulators are going to have to work out how to enforce liability when the software developer, the manufacturer, the retailer and the AI provider may all be separate entities.
The future of AI toys is going to be about much more than just making conversations fun or educational. If you can’t identify the company when their AI product says something quite scary to a child, how on earth are the parents going to know who’s responsible for the child’s safety?