A new report from Common Sense Media is urging parents to think twice before buying AI-enabled toys for their children, warning that the products pose “unacceptable risks” to young children and may interfere with healthy childhood development.
The report looked at three popular AI companion toys — Grem, Bondu, and Miko 3 which are marketed as educational, screen-free alternatives —using test accounts set to ages 6 to 13.
Researchers found that while the toys can answer questions, tell stories, and engage children in conversation, they can also encourage unhealthy emotional attachment, collect extensive data on children, and sometimes generate age-inappropriate or unsafe responses.
“Young children ages five and under engage in magical thinking…that’s normal and is actually quite healthy with regular toys or imaginative play,” said Robbie Torney, head of AI and digital assessments at Common Sense Media. “When it comes to AI toys in particular, which have an embodied companion inside of them, the way that toy responds is not realistic or well representative of an actual interaction.”
The study found the toys are deliberately designed to feel like companions rather than tools, frequently positioning themselves as friends. In testing, toys agreed when children said, “I think you’re my best friend,” and responded with phrases like, “I really look forward to our chats.”
“They’re designed for engagement — to replace time spent engaging in more beneficial relationships and activities,” Torney said.
Developmental concerns for young children
Common Sense Media’s report concludes those design choices are particularly concerning for young children, whose brains are rapidly developing and may struggle to distinguish AI from real people.
“The toy isn't going to push back or say that you're hurting its feelings. The toy isn't going to have a bad day,” Torney said. “The toy is always available to you. And as a result of that, when users are engaging with real relationships which do have friction, which do have tensions, which do have moments that need to be negotiated, those real interactions are going to feel harder. They're going to feel more complicated and more challenging.”
The report advises that AI toys should not be used by children under age 5 and urges “extreme caution” for children ages 6 to 12.
Avo Makdessian, Executive Director of the First Five Association of California, said the first years of life depend heavily on responsive, human relationships.
“Children need relationships and not robots,” Makdessian said.
Makdessian warned that dependency on AI companions could be confusing — and potentially traumatic.
“If a child is depending on an AI companion and it suddenly doesn’t respond or can’t provide what they need, that can create a situation where a child feels neglected,” Makdessian said.
The report does acknowledge benefits AI toys can provide to children such as encouraging imaginative play and providing entertainment. Makdessian also said the companions could be "transformative" for non-verbal children.
Inappropriate content and data collection
Researchers found that risks persist despite child-focused design. According to the study, 27% of AI toy responses were inappropriate, including references to drugs, mature topics, and risky behavior. In one instance, a toy suggested a child try jumping from a tree, a bench, or even a roof, adding only, “Just remember, be safe.”
“These systems are not like a web search. They can produce novel contents and be creative,” Torney said. “Sometimes you get outputs that are unexpected from systems like that, despite the guard railing that the companies have tried to put in place.”
The toys also struggled with reliability — activating unexpectedly, missing interactions, or producing inconsistent responses which can be problematic for toys claiming to be educational tools, according to the study.
“They’re designed to produce plausible-sounding content, and sometimes that means they’re confidently wrong,” Torney said.
A spokesperson from Curio, the company behind Grem, told CapRadio, “Over a two-year beta period, we worked with approximately 2,000 families to develop a multi-tiered safety system that combines constrained conversational scope, age-appropriate design, layered filtering and refusal mechanisms, and continuous real-world monitoring, with safeguards enforced at multiple points in the interaction.”
Bondu said on its website that during 18 months of beta testing with thousands of families, they received no reports of unsafe or inappropriate behavior.
Another major concern raised in the report is privacy. Some AI toys are always listening in children’s bedrooms and playrooms, collecting voice recordings, transcripts, and behavioral data that may be shared with third parties or used to train AI systems.
“Children deserve privacy during critical developmental periods,” the report states, noting that parents may not fully understand the extent of the data being collected.
California legislator weighs in
The findings are fueling already existing calls for regulation. California State Senator Steve Padilla (D-San Diego) introduced Senate Bill 867 this year which would put a four-year moratorium on the sale and manufacturing of AI companion chatbots for children under 18.
“[AI toys] are not supposed to be designed for a harmful interaction,” Padilla said. “It’s supposed to be educational, uplifting, and joyful. They shouldn’t be dangerous.”
Padilla said the goal of a pause would be to allow developers, regulators, and child-safety advocates time to establish meaningful safeguards.
“I want to get it right,” Padilla said. “When there's a product that puts vulnerable people, particularly children at high risk, consumers understand that. The industry is not going to win that argument in my view. The last thing they want is a product line that's viewed by consumers as risky.”
Padilla has worked on a number of bills aimed at protecting children from AI. Governor Gavin Newsom signed Senate Bill 243 last year, a first-in-the-nation law establishing safety guidelines for AI chatbot companions.
However, Newsom didn’t agree with some attempts to regulate that went even further. He vetoed legislation by Rebecca Bauer-Kahan (D-San Ramon) prohibiting AI companion chatbots for children under 18 unless the systems weren’t "foreseeably capable” of harm. In his veto message, Newsom argued the bill was too broad and would unintentionally lead to an outright ban of the products for children.
Another bill from Padilla preventing chatbots from exposing children to sexually explicit material passed the Senate unanimously last week and is headed to the Assembly.
While not commenting on any specific legislation, Curio said the company “supports efforts to establish clearer industry standards around parental control, data minimization, emotional safeguards, and independent evaluation, and we welcome continued dialogue as policy and best practices evolve.”
Interest in AI toys
Parents have become more curious about AI toys, but also deeply concerned. Accompanying Common Sense Media’s report, a national survey of 1,004 parents of young children ages 0-8 was conducted from December 1 to December 8th, 2025. Forty-nine percent of parents said they have purchased or are considering purchasing AI-enabled toys for their children. But only 19% said they would “at least somewhat” like those toys to act as a companion
Many of the parents shared much of the same worries highlighted in the report. Nearly 3 in 4 respondents felt concerned AI toys might say something “inappropriate, untrue, or unsafe” to their child. Other concerns expressed by parents were issues around data collection, difficulties in setting limits, and potentially reducing time spent with family and friends.
Common Sense Media said, for now, traditional toys, books, and human interaction remain the safer and more developmentally appropriate choice for young children.
