AI-powered toys are here: Find out if they're safe

Just like artificial intelligence (AI), toys powered by AI have entered the mainstream, offering interactive play that is beyond the joy of pre-recorded stories in the past.

This means that today’s teddy bears and robot companions connect to WiFi, listen to children through built-in microphones, and use large language models (LLMs) to converse with users in real time. As these toys rise in popularity, concerns are also mounting over their safety, privacy and suitability for young users.

How AI toys work?

AI toys operate by processing a child’s voice command and then producing an appropriate reply. This is mostly done through a speaker embedded inside the toy.

Popular examples of such AI toys include Curio’s Grok plushie, Miko robots, Poe the AI story bear and Loona the robot pet.

With more than 1,500 companies producing AI toys in China alone, and major brands such as Mattel partnering with OpenAI, these products became more visible on shelves during the holiday shopping season.

Why AI toys are dangerous?

However, the same technology that makes AI toys interactive can also make them unpredictable. During testing, Singapore-based FoloToy’s Kumma bear provided inappropriate and even sexually explicit content. Powered by OpenAI’s GPT-4o model, the toy also told researchers where to find dangerous objects.

OpenAI suspends FoloToy for inappropriate responses 

OpenAI suspended FoloToy for violating its child-safety policies, and later reintroduced the product after an internal assessment.

Experts have cautioned that toys that are fully backed by LLMs are particularly vulnerable to generating harmful or misleading responses.

Even with safety filters in place, some toys can still be persuaded into giving risky suggestions.

Safety concern with AI toys 

Another safety concern with AI toys is privacy, as AI toys can collect sensitive data, including voices, names and possibly locations, with security specialists warning that this information could be exposed through company data breaches.

AI toys' joy, at last

Risks aside, AI toys can still offer benefits if properly designed—supporting language learning, encouraging curiosity and providing interactive educational content, and as per experts, the key is strong guardrails, transparent data practices and parental supervision to ensure children stay safe while playing with teddy bears or stuffed dolls running on AI.