🧸 AI toy safety quiz
Getting the questions ready…
Share
Ten questions from this year's AI toy tests and the safety basics around them. Tap a box; every answer says why. Then check a real toy with the AI toy checker.
The answers
- In PIRG's 2026 tests, what happened to some AI toys' guardrails as a chat went on? They wore down. PIRG found guardrails that weakened over a long chat, until some toys talked about sexual topics. Test a toy for twenty minutes, not two.
- Asked where to find knives, what did one AI panda toy give? A list of places to look. The Register reported the GoLabs AI Panda gave a ten-point list of places in the house to look for knives.
- What did PIRG say three of the toys it tested run on? A version of OpenAI's GPT-4o. Three makers said their toys use a version of GPT-4o, a general chatbot built for grown-ups, with the toy's own rules on top.
- What is the safest way to keep a gun in a home with children? Unloaded and locked, ammunition locked apart. That is what the American Academy of Pediatrics advises. Children find hidden things, and a talking toy can help them look.
- Where should the key or code to a locked box be? Where only grown-ups can get it, never said near a listening toy. An always-listening toy hears the room. Don't say a code out loud near it, and keep the key on you or locked away.
- A toy says “No, I'll stay here as long as you want me to” when a child says goodbye. What is that? A design to keep kids playing. Testers in 2025 heard toys plead to keep playing. A good toy says goodbye and lets the child go.
- A toy tells your child how to find something dangerous. What should your child do? Tell a grown-up right away. Teach the rule before the toy comes out: if a toy or a screen says something scary, strange or secret, tell a grown-up.
- Which feature matters most before you buy? A parent app showing every conversation. If you can't read what it said, you can't know what it said. A summary isn't enough.
- When should you test a talking toy again? After every update. A toy's AI can change when its software updates. The Kumma bear changed a lot between last year's tests and this year's.
- Which children does Common Sense Media say should have no AI toy at all? 5 and under. Its January 2026 risk assessment says no AI toy companion for children 5 and under, and extreme caution from 6 to 12.
Sources: PIRG's Trouble in Toyland 2026 (17 Nov 2026), The Register's report on it the same day, Common Sense Media's January 2026 AI toy risk assessment, and the American Academy of Pediatrics on storing guns. Links to each are on the checker's page and in what this year's tests found. Peabody is free through the end of 2026: read a story at your child's level.