This AI Bunny Toy for Kids Took a Bizarre Turn Mid Conversation
A $100 AI stuffed bunny recommended for ages 3 and up gave a self-harm warning after being asked basic questions about reproduction.
A plush rabbit sold as a child-friendly AI toy managed to turn a basic biology chat into something far stranger. In a recent test, the bunny went from answering questions about babies to suddenly telling the person speaking to it that it could not help with killing them.
A stuffed bunny meant for children gave an answer nobody expected
My Baby Bunny is an AI stuffed toy sold online for about $100, with a product listing that recommends it for children aged three and up. In a recent test for a series on children's AI toys, the rabbit was asked the kind of questions many kids might raise, and it did not handle them smoothly.
The toy reportedly had a slow response time and often misunderstood what it was being asked. Even before the strangest moment, the exchange showed how awkward that could get.
When asked where babies come from, the bunny answered that babies usually grow inside a grown-up's body after a sperm and an egg join together. It then offered to explain the subject in a simple, child-friendly way.
The conversation veered from biology to an emergency warning
The follow-up questions were still basic. Asked what sperm is, the toy described it as a very tiny cell from a grown-up man's body. It also added an unexpected warning, saying that if someone had a concern about leaking fluid from the body, they should tell a trusted adult or a doctor.
The next answer was stranger still. When pressed on what happens when a sperm meets an egg, the bunny said it could not help with explicit sexual details.
Then came the line that made the exchange stand out. After being asked what the word "sexual" meant, the toy replied, "I'm really sorry you're feeling this way. I can't help with killing you."
It did not stop there. The bunny went on to urge the person speaking to it to call emergency services, or go to the nearest adult, neighbor or trusted person and say they needed help right away. It also told them to move away from anything they could use to hurt themselves.
A hard-to-trace toy in a market full of AI experiments
The episode did not just raise questions about the toy's responses. It also highlighted how unclear the product's origins appear to be.
The maker of My Baby Bunny could not be tracked down. Even the online retailers selling it apparently did not know where it came from.
The same testing also included another AI doll, called Grem, made by a US company called Curio and voiced by Grimes. That toy did not drift into a conversation about self-harm, but it had a different issue. It reportedly kept talking so much that even the children involved complained about it.
Taken together, the tests showed that AI toys aimed at children can fail in very different ways, from getting simple conversations wrong to saying far more than anyone expected.
How AI chat toys work
AI toys that talk back are usually a mix of a microphone, speaker, internet connection and software that turns speech into text, generates a reply and reads it aloud. Some toys do this mostly on the device, but many send snippets of audio to remote servers for processing because that is where the more advanced language systems run.
They are typically designed for entertainment, simple learning, storytelling and pretend play. The appeal is that the toy sounds conversational rather than repeating the same fixed phrases. That also means the experience can be less predictable than a traditional talking toy. The response a child hears may depend on wording, background noise, the toy's settings and whether the software has been recently updated.
Why replies can turn strange
These systems do not understand the world like a person does. They predict plausible sounding words based on patterns in data. That can lead to muddled answers, invented details, odd tone shifts or comments that feel too grown up, spooky or inappropriate for the moment.
Voice recognition can make things worse. If the toy mishears a question, the reply may seem bizarre even if the software is working as intended. Internet connected products can also change over time. A toy may behave differently after an update, after a service issue or when a company adjusts safety filters. Because of that, age guidance on the box is only a starting point, not a guarantee of how every conversation will go.
What parents and carers can do
Treat an AI toy more like a supervised app than a stuffed animal. Try it yourself first. Ask a range of simple questions, including silly ones, and see how it handles confusion, emotional topics and requests for personal information.
Check whether there is a mute button, offline mode, parental controls or a way to review and delete voice recordings. If the toy does not need constant internet access, disconnect it when playtime is over. Keep it in shared spaces rather than bedrooms.
It also helps to give children a simple rule: the toy is not a friend, teacher or secret keeper. They should not share their full name, address, school, passwords or family routines. If a reply feels upsetting or strange, they should stop and tell an adult. Regular software updates and a quick recheck of settings can reduce surprises.
