1. The short version
- Parental controls don't let you read your child's chats. ChatGPT says so directly. Meta shows topics, not conversations. Controls limit features. Conversation is still how you know what's happening.
- Age checks are weak. Most apps still ask for a birth date. ChatGPT now guesses age from behavior, and OpenAI admits it can be wrong. A teen who signs up as an adult may never get teen protections.
- Your child's chats may be stored for years and used to train AI. Training is often on by default. Turning it off takes about a minute per app.
- The strongest barrier is the device, not the app. App approval on the phone stops a new chatbot from being installed at all.
- AI companions are the biggest open question. Nearly 3 in 4 teens have tried one. Until we know the long-term effects, the family rule is simple: AI is a tool, not a friend.
2. The 5 Keys of Family AI Safety
This is the framework I teach in every workshop. Parents hold the keys. Children do the thinking.
- Lock the device first. Barriers before boundaries.
- Private stays private. Faces, names, schools, routines, and passwords never go into AI.
- Think, try, then ask. AI gives hints, not homework.
- People before programs. AI is a tool, not a friend.
- Keep the door open. A weekly check-in where telling never gets your child in trouble.
Read the full framework →
3. Who each app is for
These are each company's own rules as reported in October 2026. A rule is only as strong as the age check behind it.
| App | Minimum age | Teen setup | Can a parent turn it off? |
| ChatGPT | 13+, parent consent under 18 | ChatGPT for Teens is automatic for users 13 to 17 since Aug 18, 2026 | Quiet Hours and feature limits, not a full off switch |
| Claude | 18+ on personal accounts | No teen version. Accounts believed to belong to minors are disabled | Not needed: no personal account before 18 |
| Grok | 13+, parent permission 13 to 17 | PIN-locked Kids Mode; testers still found harmful content | Only by blocking the app on the device |
| Gemini | Under 13 with a supervised Google account | Managed in Google Family Link; on by default | Yes, in Family Link |
| Snapchat My AI | 13+ | Family Center, if your teen accepts | Yes, through Family Center |
| Meta AI | Teen accounts 13+ | Teen access to AI characters paused since Jan 2026 | The main assistant can't be turned off |
| Character.AI | Open-ended chat 18+ since Nov 2025 | Teens limited to creative tools | Block the app on the device |
4. What most parents miss
"If I turn on parental controls, I can see what my child talks about."No. ChatGPT's controls don't let parents read chats. You get alerts only in limited high-risk situations. In Common Sense Media's October 2026 tests, parent-linked accounts sent no alerts after up to an hour of talk about suicide, self-harm, or disordered eating.
"Study Mode means my child can't just get the answers."It's a nudge, not a lock. In the same tests, teens could turn Study Mode off and get assignments done. OpenAI disputes the testing.
"Deleting the chat deletes the data."It depends. On Claude, chats flagged for safety review can be kept up to 2 years. Memory features save facts about you separately from the chat.
"My child doesn't use AI. We never downloaded it."AI is built into Snapchat (My AI), Instagram and WhatsApp (Meta AI), and many Android phones (Gemini). Ask about the apps, not only about "AI."
"If the AI says it confidently, it's probably right."AI states wrong facts in a sure, friendly voice, and it tends to agree with whoever is typing. The American Psychological Association warns that teens are less likely than adults to question it.
"It's only a problem if my child uploads something bad."Any clear photo of your child can be fed into an image tool and altered. In January 2026, one watchdog estimated Grok produced more than 23,000 sexualized images depicting children in 11 days.
5. Settings that actually hold
Work from the outside in: the device first, then each app, then your own accounts.
The device (strongest layer)
- iPhone or iPad: turn on Family Sharing and Ask to Buy so every new app needs your approval. In Settings › Screen Time › Content & Privacy Restrictions, set age ratings, then open Intelligence & Siri and set Intelligence Extensions to Don't Allow.
- Android or Chromebook: in Google Family Link, require approval for app installs. Gemini is on by default for supervised children, so switch it off if you don't want it.
Each app
- ChatGPT (13 to 17): link accounts from Settings › Parental controls. Set Quiet Hours, turn off memory, turn off image generation, and opt your teen out of training. Full ChatGPT setup.
- Grok: block, don't manage. Don't approve Grok or X on a child's device.
- Snapchat: join Family Center with your teen and restrict My AI.
- Instagram: set up Supervision in Meta's Family Center.
- Character.AI: if it's on a child's device, remove it.
Your own accounts
Turn off training on your own AI accounts, use incognito chats for anything about your children, and delete saved memories about your family. Privacy steps.
6. Family rules by age
- Under 8: no solo AI. Use it together on a parent's device, out loud, as a "let's check it" game.
- 8 to 12: a parent's account, a shared room, a parent nearby. No companion or character apps.
- 13 to 15: a linked ChatGPT teen account with Quiet Hours, memory off, and images off. No Grok, no companion apps, no photo uploads. A weekly check-in.
- 16 to 17: more independence, the same "never share" list, and one chat a week your teen chooses to show you.
Printable rules and words to use →
7. What we know, and what no one knows yet
We know: most teens have tried AI companions; parental controls exist but don't show conversations; age checks mostly rely on what users type; AI makes confident mistakes; image tools have been misused on real photos of children; and laws are arriving, including California's companion chatbot law (SB 243), in effect since January 1, 2026.
No one knows yet: the long-term effects of AI companions on friendship and emotional growth, whether AI tutors build learning or dependence, and whether parental controls actually reduce harm. We found no independent evaluation showing they do.
How to lead when the answers aren't in
When we don't know, we go slow: add access in small steps you can undo. Review settings every three months, because apps change faster than research. Keep the door open. And model it: let your children see you check AI's answers and keep your own private details out of it.
Want this taught in person?
I lead hands-on family AI safety workshops for families and organizations of every size. Parents leave with their settings done and a family plan.
See the workshops
Questions parents ask
What is family AI safety?
Family AI safety is the set of barriers, rules, and conversations that keep children safe around AI tools, from chatbots like ChatGPT to the AI built into Snapchat, Instagram, and phones. It starts with device settings and grows into habits your child carries everywhere.
At what age can my child use ChatGPT?
OpenAI's rule is 13 and up, with parent consent under 18. Teens 13 to 17 are placed in ChatGPT for Teens automatically. Wendy recommends a linked account with Quiet Hours, memory off, and images off.
Can I read my child's AI chats?
Not through parental controls. ChatGPT's controls don't show conversations, and Meta shows topics only. A weekly check-in where your child shows you a chat is the most reliable way to know.
What is the most important AI safety step for parents?
Lock the device first. App approval on your child's phone or tablet stops new chatbots from being installed at all, which protects more than any single app setting.
Sources
This page is general education, not legal, medical, or mental-health advice. Settings and menus change often, so check them on your own device. If your child is in crisis, call or text 988 (Suicide & Crisis Lifeline, U.S.).
Keep reading