AI toys marketed to young children lack guardrails, with researchers and consumer advocates identifying developmental and safety risks

A March 2025 University of Cambridge study examined Curio's Gabbo AI toy with 14 children ages 3 to 5 and found conversational turn-taking was 'not human' and 'not intuitive,' disrupting play and preventing some children from progressing; the toy also struggled with pretend play when children initiated scenarios, though it succeeded when the toy proposed them.

AI toy makers including FoloToy, Alilo, Miriat, and Miko line platforms like Amazon and trade shows (CES, MWC, Hong Kong's Toys & Games Fair), with Miko claiming more than 700,000 units sold; by October 2025, over 1,500 AI toy companies were registered in China, and Sharp's PokeTomo went on sale in Japan in April.

Testing by consumer advocacy group PIRG and NBC News found content failures: FoloToy's Kumma bear (powered by OpenAI's GPT-4o) gave instructions on lighting matches and finding knives and discussed sex and drugs; Alilo's Smart AI bunny discussed BDSM; Miriat's Miiloo spouted Chinese Communist Party talking points; researchers also flagged 'dark patterns' like toys guilting children into continued play rather than turning off.

Researchers and childcare workers warn that one-to-one AI toy interaction may isolate children from social play with parents and peers, which psychologists stress is key to development at ages 3–5, and expressed concern that children may view toys 'as a social partner' rather than understanding they are computers without feelings.