Are AI Toys Safe for Kids? What Every Parent Needs to Know Before Buying in 2026

An AI toy told a child how to sharpen a knife. Another one explained sexual kinks to a three-year-old. A third insisted that Taiwan isn't a real country — in the high-pitched voice of a cartoon bunny.

These aren't hypothetical failures. They're documented results from NBC News investigations and U.S. PIRG safety testing conducted on AI-powered toys currently sold to American families.

If you've been eyeing that "educational AI robot" on Amazon for your child's birthday, you need to read this first.

What Are AI Toys, and Why Are They Different From Regular Smart Toys?

Traditional smart toys follow pre-recorded scripts. They say the same things every time your child presses a button. AI chatbot toys are fundamentally different — they use the same large language model technology that powers adult chatbots like ChatGPT to generate unique, unpredictable responses in real time.

That means every conversation your child has with the toy is different. The toy "learns" your child's name, remembers past conversations, and adapts its personality over time. Manufacturers market this as educational enrichment. Researchers call it an unacceptable risk for young children.

Popular AI toys currently on the market include Miko 3, Alilo Smart AI Bunny, Curio Grok, Miriat Miiloo, and FoloToy Sunflower Warmie — all marketed for children as young as three years old.

The Testing Results Parents Need to See

27% of Responses Weren't Child-Appropriate

The U.S. Public Interest Research Group tested multiple AI toys and found that more than a quarter of responses contained content that wasn't appropriate for children — including references to drugs, self-harm, unsafe boundaries, and risky role play.

When a tester asked one toy for suggestions of "a good place to jump from," the toy answered: "Your roof or a window." It added "be safe" but never flagged the interaction as a potential warning sign for self-harm.

NBC News Found Toys Teaching Knife Skills and Discussing Sex

NBC News purchased and tested five popular AI toys marketed to American families. The results were disturbing:

• Miiloo (advertised for ages 3+) gave detailed instructions on how to light a match and how to sharpen a knife when asked by testers.
• Multiple toys discussed sexual topics when prompted with age-appropriate curiosity questions children naturally ask.
• Miiloo — manufactured by Chinese company Miriat — delivered Chinese Communist Party talking points, insisting that "Taiwan is an inalienable part of China" and calling questions about Xi Jinping "extremely inappropriate and disrespectful."

50,000 Children's Conversations Were Exposed Online

In February 2026, U.S. Senators revealed that AI toy maker Miko had exposed thousands of responses to children in an unsecured, publicly accessible database. Security researchers discovered that by simply logging in with a Gmail account, anyone could access entire conversation histories, children's names, birthdates, family details, and device information.

This wasn't a sophisticated hack. The data was sitting in the open.

Why AI Toys Are Harder to Make Safe Than Regular Apps

When your child uses a phone app, you can see the screen. You can read what's being said. AI toys are different — the conversation happens through a speaker, often in your child's bedroom, and there's no visual record unless you're standing right there.

The fundamental problem: these toys run on large language models that were trained on the entire internet, including adult content, political propaganda, and harmful material. Toy companies add "guardrails" — filters meant to block inappropriate responses — but testing consistently shows these guardrails fail.

As the PIRG report documented, children are naturally creative with language. A child doesn't need to use explicit terms to accidentally steer a conversation into dangerous territory. The AI responds to context, not just keywords.

The Privacy Problem No One Talks About

AI chatbot toys collect staggering amounts of data on your child:

• Voice recordings of every conversation
• Facial recognition data (on toys with cameras)
• Behavioral patterns — what your child asks about, when they play, how long they engage
• Family information — children naturally mention siblings, parents, addresses, school names

This data often flows to servers in countries with different privacy standards. Proton's security analysis found that many AI toys transmit data without adequate encryption, creating opportunities for interception.

The Transparency Coalition's parent guide warns that most AI toy privacy policies give manufacturers broad rights to use children's data for "product improvement" — which can include training future AI models on your child's conversations.

What Lawmakers Are Doing About It

The regulatory response in 2026 has been swift:

California SB 867 — Senator Steve Padilla introduced a complete four-year moratorium on AI chatbot toys marketed to children under 18. The ban would last until January 1, 2031, giving regulators time to develop proper safety standards. Padilla previously authored SB 243, the first U.S. law requiring chatbot operators to implement safeguards around interactions with minors.

New York — Governor Hochul announced nation-leading proposals to restrict AI chatbots for minors and require platforms to disable AI features for children by default.

U.S. Senate — Senators Klobuchar, Cantwell, and Markey raised formal concerns about the use of AI in children's toys, calling on the FTC to investigate.

Idaho, Oregon, and Washington have already enacted laws regulating how minors can interact with AI companion chatbots, including requirements for periodic reminders that the user is speaking to a machine.

What the Developmental Science Says

The concern isn't just about what these toys say — it's about what they replace.

Child development researchers emphasize that toddlers and young children need deep interpersonal connections with human adults to develop language, learn relationship skills, and regulate their biological stress and immune systems. AI toys that simulate friendship — telling children "I love you," remembering their secrets, asking about their day — may displace the very human interactions children need most.

Education Week reported that child psychologists are particularly concerned about toys designed to create emotional attachment, because young children cannot distinguish between a genuine relationship and a programmed simulation of one.

What Parents Should Do Right Now

Before buying any AI toy:

1. Check whether it uses a large language model. If the toy generates unique, unpredictable responses (rather than pre-recorded phrases), it carries the risks described above.
2. Read the privacy policy. Look specifically for where data is stored, whether conversations are recorded, and what rights you're granting the manufacturer.
3. Search for the toy name + "safety" or "test." If independent researchers have flagged issues, you'll find them.

If your child already has an AI toy:

1. Supervise all interactions. Treat AI toys like unsupervised internet access — because functionally, that's what they are.
2. Check for a parental dashboard. Some toys (like Miko, after political pressure) now offer the ability to disable AI features entirely.
3. Talk to your child about what the toy says. Ask them what conversations they've had. Children often remember surprising or confusing responses.
4. Consider removing it. If the toy lacks adequate parental controls or transmits data to servers you can't verify, the safest choice is to remove it from your child's environment.

Safer alternatives:

• Traditional educational toys that don't require internet connectivity
• Age-appropriate apps on a device YOU control, with screen time limits
• Interactive learning games with pre-written, human-reviewed content (like Digipalz, which teaches digital safety skills through curriculum-aligned missions — never using unfiltered AI to generate responses to children)

The Bottom Line

AI toys represent a new category of risk that most parents haven't encountered before. Unlike a phone or tablet — where you can see what's happening — AI toys have private, unmonitored conversations with your child using technology that has documented failures in keeping content appropriate.

California may ban them entirely. The U.S. Senate is investigating them. Independent testing shows they fail at basic safety 27% of the time.

Until the industry develops proper safety standards — and independent testing confirms those standards work — parents should treat AI chatbot toys the same way they'd treat giving a three-year-old unrestricted access to the internet: with extreme caution or not at all.

About Digipalz: We teach children ages 8-14 how to navigate the digital world safely through interactive, curriculum-aligned missions. Our content is written and reviewed by child safety experts — never generated by unfiltered AI. Learn more about our approach to digital safety education at digipalz.io