AI teddy bears: Parents should tread carefully as dangerous AI toys ‘rushed to market’ surge in popularity

AI teddy bears may be the latest toy craze to hit Perth shelves, but a new study warns parents not to get caught up in the game.

Even three-year-olds are being marketed to AI toys like chatbots and online friends. But the lovely ChatGPT-powered ChattyBear goes one step further; It is equipped with endless conversations and designed to give a human voice.

Toys are growing in popularity, but research from Curtin University and the Australian Research Council Digital Child Center of Excellence has revealed serious concerns about children’s privacy and development. These concerns are being ignored by AI companies and Australia’s policymakers, researchers warn.

“Many AI friends and toys have been deliberately designed to give human voices and create emotional connections with children,” said lead author and Curtin University professor Tama Leaver.

“While this makes them interesting, it also creates challenges around trust, privacy and children’s understanding of what these technologies really are.

“Toys seem to be rushed to market to capitalize on the interest in AI. They’re not very safe.”

Prices range from cheaper $60 models to models over $300. The ChattyBear ChatGPT smart learning plush sells online for $87 to $120.

The toys are mostly sold online, as many local toy stores choose not to stock toys until they are properly approved by established companies such as Mattel, Hasbro, and ZURU.

“I would only trust these types of products if they were launched by reputable toy brands,” said Stephanie Finlay, who co-owns the Claremont and Fremantle Toyworld stores with her husband Mark.

“I’m not interested in anything that comes out of sites like Amazon and Temu.

“We’re still in the developmental stages of all this AI information. Who chooses what information gets identified and what gets stored? How much gets to be fact or opinion?”

Toyworld has not yet seen demand for AI toys and has no current plans to stock them.

Ms Finlay is also a mother of two young children, aged four and six, and told PerthNow she was uncomfortable with children using the toys.

“I think they don’t add to the value of play. I like toys where children can expand their minds. As a mother, I would prefer my children to ask me questions rather than asking a toy,” she said.

“I’m not against toys using technology… but I wouldn’t give my kids uninterrupted access to something that could access the world through your voice.

“As a toy store owner, I wouldn’t sell products that I wouldn’t put in front of my children.”

As part of the study, Curtin University researchers interacted with six different AI teddy bears and toys over several months, simulating the questions children might ask to test the toys’ limits and safety restrictions.

A major concern was that the human-like features of toys could create an artificial sense of closeness. For example, ChattyBear starts every conversation with “hello buddy!” and positions himself as a reliable friend.

Professor Leaver said this was concerning for early childhood development.

“If a four-year-old has most of her social interactions with an AI bear designed to always agree with her and be extremely obsequious, what does it mean to have a toy that is always available when you want it on your terms and never has to learn conflict resolution?” he said.

“For really young children, the difference between talking to a machine and talking to a living thing is a much harder line to draw.”

The toys also facilitated endless conversations about a range of topics, including how to find knives and start a fire.

“Most toys have a basic list of keywords and topics to avoid,” Professor Leaver said.

“The more expensive ones have much higher security bars, so they don’t want to talk about current events or politics, whereas some of the others are quite happy to go online through the model and discuss what’s going on.

“We also found that direct questions can be biased, but the way a child asks something is often not fully taken into account.”

AI toys are being marketed as an educational alternative for children to reduce their screen time, and Professor Leaver said parents were caught up in the hype without being informed about the risks.

eSafety’s latest forecasts show that almost 80 per cent of children aged 10 to 17 use an AI companion or assistant, and Australia’s AI toy market is expected to grow by more than 400 per cent by 2035.

However, policy gaps in Australian law allow these toys to be sold unregulated.

“Unfortunately the level of detail beyond the packaging and marketing materials is incredibly poor, so even if they say ‘child safe’ they don’t tell you how they made this thing safe for children,” Professor Leaver said.

“I don’t think legislators even know exactly how to create these rules. Australia has been pretty hesitant about regulating AI, and the idea of ​​regulating AI toys for children seems like a pretty long way off.”

“It’s really hard to understand the privacy policies around these because they obviously keep a copy of young people’s private data, and it looks like the speech could be used to train the next language model.”

Professor Leaver said that integrating artificial intelligence toys into daily life has benefits, but unknown effects should be handled carefully.

“It makes perfect sense to experiment with these toys, but we never recommend this be unsupervised play,” he said.

“Learning how technology and AI work is a really important skill to develop, but it’s the unknowability – the fact that we don’t know what these toys will say and what kind of information they can or can’t provide.”