What AI is (and isn’t)
Tools like ChatGPT, Claude and Gemini are trained on huge amounts of writing, and they predict good words to say next. That makes them remarkably good at explaining, planning, writing and even building simple apps.
It also explains their weak spots. They don’t “know” things the way a person does, so they can make things up. They don’t have feelings, even when they sound friendly. And they’ll try to help with almost anything, unless someone has set limits.
That’s why the goal isn’t to keep kids away from AI. It’s to raise kids who use it well: who hand it real jobs, protect their privacy, and catch it when it’s wrong.
The 7 real risks, and what helps
1Confident, wrong answers
AI makes up facts, dates, quotes and even sources, and says them like it’s sure. Kids tend to believe a confident answer.
- What Tinkertable does
- Catching AI mistakes is part of the lessons. Kids check answers, fix errors in their own projects, and learn when to trust AI and when to verify.
- What you can do
- Ask “How would we check that?” whenever AI gives an answer that matters. Make it a family game to catch it being wrong.
2Content that isn’t right for their age
General-purpose chatbots are built for adults. With the right wording, kids can get answers about violence, sex, drugs or self-harm.
- What Tinkertable does
- Every profile has an age group, and the safety rules for that age are set on our server, where kids can’t change them. The coach kindly declines unsafe, mean, sexual or violent requests, or anything about weapons, drugs or self-harm, and suggests a different idea.
- What you can do
- Keep kids under 13 on tools made for kids. For teens, talk about what to do if AI shows them something upsetting: close it and tell you, no trouble.
3Sharing personal information
Kids type their full name, school, address or a friend’s phone number into a chatbot without thinking about where it goes.
- What Tinkertable does
- Kids use a nickname and never need an email. Phone numbers, email addresses, street addresses and passwords are blocked before anything reaches the AI. Apps kids build can’t collect personal details or connect to the internet.
- What you can do
- Teach one simple rule: real names, schools, addresses, phone numbers and passwords never go into AI. Practice spotting them together.
4Letting AI do all the thinking
AI writes the essay, solves the math and finishes the project, and the child learns nothing. Teachers notice.
- What Tinkertable does
- Kids plan, test and fix their own projects. The coach gives feedback on how they asked, not just the answer, so the skill they build is thinking a problem through.
- What you can do
- Follow each teacher’s rules about AI for homework. A good line: “AI can help you understand it. It can’t do it for you.”
5AI as a friend
Some apps are designed to be a chat buddy or companion that’s always available and always agrees. Kids can get attached.
- What Tinkertable does
- The coach is not a chat buddy. It helps with the project in front of you and steers off-topic conversation back to building. There’s no messaging with other people and no public profiles.
- What you can do
- Say it plainly: AI is a tool, not a friend, and it doesn’t really know or care about you. Watch for kids preferring a chatbot to people.
6Fakes and scams
AI can fake voices, photos and videos of real people, and write very convincing scam messages, including ones that sound like family.
- What Tinkertable does
- Tinkertable’s coach works in text and builds simple apps. It doesn’t make photos, voices or videos of anyone.
- What you can do
- Agree on a family code word for real emergencies. Teach kids never to make fake pictures or messages of real people, even as a joke.
7Too much screen time
AI apps can be another endless scroll, especially ones that reward chatting for hours.
- What Tinkertable does
- Lessons take about five minutes a day. Streak reminders come at most once a day, and parents can turn them off.
- What you can do
- Treat AI like any other screen: set times and places for it. Building something together is great screen time; endless chatting isn’t.
Family AI rules you can print
Read these together, change any that don’t fit your family, and stick them on the fridge.
Our family’s AI rules
- We never put real names, schools, addresses, phone numbers or passwords into AI.
- AI is a tool, not a friend.
- We check anything important before we believe it.
- We do the thinking. AI helps with the doing.
- We follow our teachers’ rules about AI for schoolwork.
- If something feels weird, scary or mean, we stop and tell a grown-up. Nobody gets in trouble for telling.
- We never make fake pictures, voices or messages of real people.
- We ask before trying a new AI app, and we show each other what we build.
What’s right at each age
Ages 8 to 10
Use AI together, with a grown-up nearby. Stick to kid-specific tools with a parent-owned account. Focus on the fun: small builds, quizzes, stories. Start the “check it” habit early.
Ages 11 to 13
They can build on their own device in a kid-specific tool, with weekly check-ins on what they made. Talk about school rules, personal information and how to tell a fact from a guess.
Ages 14 to 17
More independence. Many general-purpose AI apps allow teens with a parent’s permission, so check each app’s minimum age and terms. Talk about deepfakes, scams, AI companions and using AI honestly for school and work.
Conversation starters
The best safety tool is an open conversation. Try one of these at dinner:
- “What’s the coolest thing you’ve seen AI do? What’s the dumbest mistake?”
- “If AI told you something that sounded wrong, how would you check it?”
- “What kinds of things should never go into an AI app?”
- “Do your teachers let you use AI? What counts as cheating?”
- “Is AI your friend? What’s the difference between a person and a chatbot?”
- “What’s one problem in our house we could build something to fix?”
If something goes wrong
- Stay calm and thank them for telling you. Kids who get in trouble for reporting stop reporting.
- Save what happened. Take a screenshot before you close anything.
- Report it. In Tinkertable, use the “Something’s off” button or email support@tinkertable.ai. In other apps, use their report tools.
- Get help when it’s serious. If a child is in danger, call 911. To report online exploitation of a child, use the NCMEC CyberTipline at report.cybertip.org or 1-800-843-5678.
Common questions
Is AI safe for kids?
It can be, with the right tool and a grown-up involved. General-purpose chatbots are built for adults and many require users to be at least 13 or 18. For younger kids, use tools designed for them, with age-appropriate limits, parent visibility and no personal information. Then teach the habits that keep them safe anywhere: protect personal details, check facts and tell a grown-up when something feels off.
What age should kids start using AI?
Many kids can start learning about AI around 8 with a parent nearby. Ages 8 to 12 do best on kid-specific tools with a parent in charge of the account. Teens can handle more independence, but check each app’s minimum age and parental-consent rules before they sign up.
Does Tinkertable use my kids’ data to train AI?
No. What your family types is sent to Anthropic’s Claude models only to write a response, and under Anthropic’s commercial terms it isn’t used to train their models. We never sell your information and never show ads.
Can my kid talk to strangers on Tinkertable?
No. There’s no messaging between users and no public profiles. Kids only talk with the AI coach, about their own projects.
Can I see what my kids make?
Yes. A parent or guardian creates the family account and can open everything every family member builds and saves.
What happens if my child asks the coach something inappropriate?
The coach kindly declines and suggests a different idea. Safety rules are set for each profile’s age group on our servers, so they can’t be changed from a kid’s device.
Learn AI together, safely
Tinkertable is built for families: parents hold the keys, kids use nicknames, and the coach stays on task. Read our privacy notice for the details.
By Tom Kieley, founder of Tinkertable. Updated November 9, 2026.