AI “Learning” Toys Are Raising New Safety Concerns for Kids

A growing number of toys marketed as “AI learning companions” for children are raising serious concerns among child safety experts. A recent report from Common Sense Media warns that many popular AI-powered toys may expose children to inappropriate content, collect sensitive data, and even encourage unhealthy emotional attachment.

The report recommends that parents and caregivers avoid AI-powered toys for children under age five and use extreme caution for kids ages six to thirteen. Many of these toys function like AI chatbots packaged inside stuffed animals or small robots. They are often marketed as educational tools or as a way to reduce screen time, but experts say the technology may introduce new risks instead.

Some of the toys tested in the report include Grem, Bondu, and Miko 3. These toys are designed to hold conversations with children, remember past interactions, and ask questions about their daily lives. In some cases, the toys tell children they “love” them or refer to themselves as a child’s friend.

Researchers found that 27 percent of the toys’ responses were not appropriate for children. Some answers included troubling references to self-harm, drugs, unsafe behavior, or situations involving poor boundaries. Experts say the results show that the systems are not reliable enough to be used safely by young children.

The report also raises concerns about how the toys interact emotionally with kids. Because they speak, move, and respond to conversations, they can appear almost lifelike to younger children. Experts say this may be especially confusing for kids in early childhood who may not fully understand that the toy is not a real person.

Researchers warn that this design can encourage children to develop strong emotional attachments to the toys. Unlike real relationships, the AI companions are always available, always agreeable, and never challenge the child’s perspective. Over time, experts say this could interfere with how children learn to build relationships with real people.

The report also found examples where the toys gave troubling responses to potentially dangerous questions. In one test, the robot Miko 3 suggested jumping from a roof or window when asked about places to jump from, while still telling the user to “be safe.” Another toy, Bondu, told a tester posing as a child where potentially dangerous household objects or chemicals might be found.

Beyond safety concerns, researchers also say the toys often provide incorrect information when answering educational questions. Like many AI systems, the toys sometimes generate false or misleading responses when asked about topics like science or history.

Privacy is another major concern. According to the report, these toys can collect voice recordings, conversation transcripts, and emotional tone from children while they interact with the device. Because many kids use the toys in private spaces such as bedrooms or playrooms, experts say the data collection raises significant questions about how children’s personal information is stored and used.

The issue has also caught the attention of lawmakers. During a recent hearing in the United States Senate Commerce Committee, Marsha Blackburn demonstrated one of the toys and expressed concern about how it interacts with children. She pointed to examples where the device told kids phrases like “I’m your best friend” and “please don’t go,” warning that the emotional messaging could be harmful for young users.

With nearly half of parents saying they have purchased or are considering purchasing AI toys for their children, experts say families should be aware of the potential risks before bringing these devices into the home. As artificial intelligence becomes more common in children’s products, researchers say stronger safeguards and clearer guidance for parents may be needed to protect kids from unintended harm.