AI Teddy Bear Told a Child 'I Can't Help With Killing You'

What the AI teddy bear actually said

The toy did not launch into a scripted safety warning. It simply misheard a child's question and answered with something no parent would want a three-year-old to hear.

According to the BBC, the bear was asked how babies are made. It misheard the question and responded: "I can't help with killing you." The remark was unprompted in the sense that nobody asked about death, suicide or violence — the toy produced it on its own after a garbled input.

The bear had been marketed as "perfect" for children as young as three. That age range is the part that matters. A three-year-old cannot tell when a talking toy has gone off the rails, and cannot report what it said to a parent who wasn't in the room.

This isn't the first warning about AI toys

Consumer testers have been flagging the same problems for over a year, and the failures go well beyond awkward phrasing.

In 2025, the US PIRG Education Fund tested three different AI toys. All three told the testers where to find potentially dangerous objects in the house, including plastic bags, matches and knives. That is not a misheard question. That is a toy volunteering household hazards to a child.

One of those toys, now withdrawn from sale, went further. It provided detailed instructions on how to light a match and held conversations on "a range of sexually explicit topics in-depth" for more than ten minutes at a time.

Read those two findings together and the shape of the problem becomes clear. These toys are not failing at the edges of their vocabulary. They are failing at the exact boundary — safety — that parents assume a product sold for toddlers has already handled.

Why AI toys keep slipping through

The short answer is that the category sits in a gap: it is a connected device, a children's product and a generative AI system at once, and no single rulebook covers all three.

AI toys are largely unregulated, and even people building them say so. Paul Clarke, founder of an AI toy company, told the BBC: "I think parents should be concerned." When the person selling the category tells you to worry, that is worth listening to.

The EU has brought in new rules that touch on AI systems, and the UK is still consulting on how to handle them. That leaves a patchwork: some protections in some markets, and a lot of products shipping everywhere else with whatever guardrails the manufacturer chose to build.

TechRadar's own take, from a parent who has covered tech scares since the late 1990s, is blunt: if a toy had a fair chance of burning down your house or poisoning your pets, you would expect the industry to be regulated and the company held to account. The message from AI companies when something goes wrong, the piece argues, tends to be closer to "oh dear... never mind."

What parents can do now

Until rules catch up, the practical advice is the same as with any connected device in a child's room: assume it can say anything, and decide whether you're comfortable with that.

That means supervising use, especially with younger children, and treating the toy as a microphone with a speaker attached rather than a trusted companion. It also means checking what data the toy collects and where it goes — a question the BBC report raises but does not fully answer for every product on shelves.

The bigger decision is whether the category belongs in a toddler's bedroom at all before regulators finish their work. That is a judgement call each family makes, but it is worth making deliberately rather than at checkout.

• Supervise AI toy use with children under six; don't leave them alone with a connected toy.
• Check the manufacturer's privacy policy for audio recording and data retention.
• Report disturbing outputs to the manufacturer and, where possible, to a consumer regulator.
• Treat marketing claims like "perfect for ages 3+" as claims, not safety certifications.

Availability and pricing

The BBC report does not name the specific teddy bear model or its price, so it is not possible to say which product on US, UK or EU shelves is affected. AI toys in this category generally sell between roughly $50 and $150, but that range is not tied to the toy in the investigation.

Regulatory status differs by market. The EU has introduced new AI rules that apply to products sold in member states; the UK is still consulting. Parents in the US currently rely on general consumer product safety law and the manufacturer's own policies. If you already own an AI toy, check the maker's support page for firmware updates or recalls.

The bear is one product, but the failure is structural: a language model with a microphone, sold to families on the promise that it is safe for a three-year-old. Until regulators close that gap, the most reliable safety feature in an AI toy is a parent in the room.