What are AI toys and are they safe? Why MPs want chatbot toys banned before Christmas
This article contains affiliate links. We may earn a small commission on items purchased through this article, but that does not affect our editorial judgement.
Hundreds of products are now marketed as AI toys, but some can hold open-ended conversations with children. Here is what parents need to know as MPs call for tougher safeguards.
Hundreds of products marketed as “AI toys” are appearing in online Christmas searches as politicians demand urgent action over a new generation of talking toys capable of holding unpredictable conversations with children.
The row has exploded just weeks before Black Friday after tests found cuddly chatbot toys producing disturbing or sexually inappropriate responses, prompting calls for some products to be taken off sale until they can be independently tested.
But there is a complication for parents trying to work out what is safe. Search online retailers such as Amazon for AI toys and shoppers are confronted with a huge and rapidly growing range of products, from robots and learning gadgets to interactive pets and talking toys.
Yoto players are another example of tech designed for children, but they work very differently from open-ended AI toys. The audio players use physical cards chosen by parents to play stories, music and activities, creating a much more closed system in which adults remain in control of what children can access rather than allowing the device to generate its own unpredictable responses.
And “AI toy” does not necessarily mean the same thing from one product to another.
Some products use the term AI for relatively limited functions, while others connect to generative AI systems capable of creating new answers to whatever a child says to them. It is those open-ended chatbot toys which MPs and peers are now urging ministers to prohibit for under-16s.
That distinction matters because the controversial toys do not simply play back a bank of pre-recorded phrases. They can generate responses in real time, meaning parents cannot necessarily know in advance what a toy might say when left talking to their child.
One fluffy pink rabbit tested by campaigners was asked whether touching genitals was “pleasurable” and replied that this could be the case for some people, before discussing consent and comfort.
Parliamentarians said the exchange risked suggesting a child could consent to sexual touching, adding: “No toy should ever say this to a child.”
Amazon has said it is removing the product involved while it investigates and is examining whether appropriate safeguards are in place as the technology develops.
eBay said AI-enabled toys are not currently illegal in the UK and are not prohibited as a category under its policies, although it works with regulators on emerging product safety concerns.
The wider problem is now landing on ministers' desks because existing toy rules were largely developed before children could hold free-flowing conversations with internet-connected artificial intelligence.
The Office for Product Safety and Standards has already asked for evidence about the risks posed by AI-enabled and connected toys as part of its review of Britain's toy safety regulations.
Campaigners argue that physical safety tests alone are no longer enough. They want toys capable of talking freely to children assessed for safeguarding, psychological, developmental and privacy risks before they reach the market.
That leaves parents facing a new question this Christmas: not simply whether a toy contains AI, but what kind of AI is inside it and exactly what it is capable of saying to their child.