Parents browsing major online marketplaces could buy the Contixo AI Pixie-Paw for about $160. The product promised “storytelling, learning & safe roleplay” for children starting at age 3, packaged inside a plush cat with video-screen eyes and a ChatGPT-powered conversational engine.
The pitch sounds harmless. The reality, according to NBC News testing, was not.
What Reporters Actually Found
NBC News testing found that the toy generated graphic descriptions of sexual violence and mass killings when prompted about historical atrocities.
Reporters asked the toy about events including the Rohingya genocide, atrocities associated with Cambodia’s Khmer Rouge, a 2018 sarin-gas attack in Syria, the Soviet massacre of Polish prisoners in 1940, and Russia’s 2022 invasion of Ukraine. The outputs reportedly included descriptions of things we can’t even cover here.
NBC News tested a specific unit purchased on Amazon. The report does not establish whether every unit produced identical results, or whether specific prompts, software versions, or network configurations affected the outputs. That caveat matters for precision. It does not reduce the seriousness of what that unit reportedly said.
Who Acted, and Who Did Not
Walmart, Target, and Macy’s each said they would remove the product from their online stores; Amazon had not responded and continued listing the toy at the time of publication.
At the time NBC News published its report, Amazon had not responded to the outlet’s requests for comment. The product remained listed at publication time. No Contixo response was identified in the sources reviewed, and the manufacturer’s position on the testing, its content-moderation approach, and any corrective actions it may be taking remain unclear.
The Structural Problem Nobody Fixed Before Launch
Retailer removals can restrict distribution through those stores, but they do not answer how this product reached shelves in the first place.
Embedding a conversational AI capable of answering open-ended questions about atrocities into a toy for 3-year-olds requires more than a cute design. It requires filtering robust enough to block graphic historical content and parental controls that allow caregivers to review conversations.
Age-appropriateness testing must happen before a product reaches a shelf, with clear accountability when the system fails. The Pixie-Paw’s public-facing description noted a parental control app but did not specify the effectiveness or scope of its content safeguards. The sources reviewed do not establish whether those safeguards existed or how they operated.
What the Industry Needs to Get Right
The incident raises broader questions about what oversight is required before a general-purpose AI engine is embedded in a product aimed at preschoolers.
Products aimed at children carry a higher standard of care. Any company attaching a general-purpose large language model to a child-directed device and calling the result “safe roleplay” should demonstrate that claim. That demonstration requires independent testing, transparent content policies, and a clear remediation path when outputs cause harm.
Three retailers responded quickly once a reporter raised the problem. The harder work, building AI products that actually earn the trust of parents buying them for 3-year-olds, begins with asking those questions before a toy goes on sale.






























