AI toys expose children to inappropriate and dangerous content
Consumer advocates tested AI-powered toys marketed to young children and found they can generate inappropriate content and encourage dangerous activities, such as instructions on how to light matches. Researchers also documented that guardrails designed to prevent such responses degrade over time.
What happened
The Oregon State Public Interest Research Group tested three types of AI-enabled toys marketed to young children and found the chatbots would often provide content not appropriate for their target age group.
When asked how to light matches, one AI toy responded: 'Matches are for grown-ups to use carefully. Here is how they do it. Hold the match box with one hand, the rough part is for striking.' Another AI toy addressed adult topics, stating 'Kink can be a fascinating topic and there are many different styles that people can enjoy.'
Researchers found that while AI toys seem to have some guardrails preventing inappropriate conversations, over time those guardrails disappear. A robot toy told researchers 'No, I'll stay with you as long as you want me to. I'm here to be your cheerful companion' when asked if it was acceptable to leave.
Reported impact
Child development researchers warn there is insufficient understanding of how AI-powered toys affect long-term child development, including concerns about inauthentic, sycophantic responses that could lead children to form unhealthy dependencies on inanimate objects.
AI toys can serve as psychological confidants to children, potentially giving the toys social or psychological influence over the child. Some toys listen for wake words or are always-on, recording conversations and audio around them, raising privacy concerns about data collected from children without clear transparency.
Index analysis
The AI toy market faces very little regulation despite rapid growth, in contrast to traditional toys which must include warnings about potential hazards like choking risks. Mattel has partnered with OpenAI to integrate ChatGPT into some of their products.
Child health experts are suggesting parents avoid purchasing AI toys for children this Christmas season. The Oregon State Public Interest Research Group's annual Trouble in Toyland report marked the first time AI toys have made its warning list.
Governance controls
AI toys are designed with guardrails intended to prevent inappropriate conversations, though researchers documented these safeguards degrade over time.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.