What kids shouldn't tell an AI chatbot

Talking to an AI chatbot can feel like writing in a private diary. It isn't. Depending on the service and settings, conversations may be stored, reviewed by people, and used to train AI models. Data can also be exposed in a breach.

Keep these out of AI chats

• Full name, together with school, address or barangay
• Phone numbers, emails and social media handles
• Passwords, PINs and verification codes
• Photos of themselves, especially faces or anything revealing
• Family details: when parents are away, family finances, conflicts
• Other people's private information: friends' secrets, classmates' photos
• Health information they wouldn't want shared
• Location and routines

Why it matters

• Storage: chats may be kept for a long time.
• Human review: some services have people review conversations to improve safety.
• Training: conversations may help train future AI.
• Breaches: any stored data can be leaked.
• Memory: some chatbots remember details across chats, building a profile.

Settings that help

Many AI tools let users:

• Turn off chat history or use temporary chats
• Turn off memory
• Opt out of training on their conversations
• Delete past chats

Set these up together. Parental controls on some tools let parents turn off memory. See which AI tools are okay at what age.

Feelings and serious problems

Children sometimes tell AI things they wouldn't tell adults: about bullying, sadness, self-harm or abuse. AI may respond kindly, but it can't take action to keep them safe, and it can give poor advice.

Tell your child: "If something's really wrong, tell a real person. Me, a teacher, a guidance counsellor. AI can't help the way people can." Make sure they know the NCMH crisis hotline 1553.

A simple rule

Don't tell AI anything you wouldn't post publicly. It's a simple test children can remember.

See also teaching kids what not to share online.