Back to Risk Categories

AI Safety Hub

Understanding and mitigating AI risks for children

AI 101 for Parents

What AI Is — and What It Isn't

Artificial Intelligence, as your child encounters it, is pattern-matching software. It has been trained on vast amounts of text and learned to predict what word comes next in a sentence. It does not think, feel, understand, or care.

When ChatGPT says "I understand how you feel," it is producing a statistically likely response — not expressing empathy. When Character.AI creates a "character" that seems to have a personality, it is following rules that generate the appearance of personality.

This distinction matters enormously because children under 12 cannot reliably make this distinction. Their brains are still developing the capacity for abstract reasoning about non-human entities. A chatbot that responds to their questions is, in their cognitive framework, indistinguishable from a person who cares about them.

Critical Warning

The Anthropomorphization Crisis

Research from Cambridge (Kurian, 2024) shows that 26% of vulnerable adolescents prefer AI chatbots to real people — not because they believe the AI genuinely cares, but because the AI cannot judge them, report them, or withdraw emotional support.

The emotional dependency gradient follows a clear pattern:

1

Initial Enthusiasm

Discovers the AI, fascinated by its responses

2

Daily Reliance

Returns daily, shares personal information

3

Emotional Distress

Anxious when unable to access the AI

4

Human Replacement

Prefers AI interaction over human relationships

Warning Signs: Withdrawal from friends/family, distress when the app is unavailable, talking about the AI as though it has feelings, checking the app first thing in the morning and last thing at night, declining interest in human social activities.

AI Safety Evaluation Checklist

Use this checklist to evaluate any AI tool or platform your child wants to use.

Age Appropriateness

Does the AI system have robust age verification?

Is the content appropriate for your child's developmental stage?

Does it comply with the Children's Code (AADC)?

Anthropomorphization Risk

Does the AI use first-person pronouns ("I think", "I feel")?

Does it simulate personality or emotions?

Could your child mistake it for a being that cares about them?

Data & Privacy

What data does it collect from your child?

Is data stored and for how long?

Can you request data deletion under UK GDPR?

Does it share data with third parties?

Dependency Risk

Does it include usage time tracking?

Does it encourage breaks or redirects to human interaction?

Does it actively resist attachment formation?

Educational Impact

Does it scaffold learning or just provide answers?

Does it celebrate process over product?

Could it replace thinking your child should do independently?

Platform-Specific AI Guides

ChatGPT (OpenAI)

Minimum age: 13 (18 without parental consent)

General-purpose AI chatbot

Known Risks

  • No real age verification
  • Can generate inappropriate content with creative prompting
  • Academic integrity concerns
  • Children may develop learned helplessness

Safety Features

  • Content filtering active by default
  • Memory can be disabled
  • Usage dashboard available

Parent Actions

  • Set family rules about when ChatGPT is appropriate for schoolwork
  • Review conversation history periodically
  • Discuss the difference between using AI as a tool vs. a replacement for thinking

Character.AI

Minimum age: 13 (17+ for some features)

AI character roleplay platform

Known Risks

  • Highest anthropomorphization risk — characters designed to simulate personality
  • Emotional dependency patterns documented in research
  • Romantic/relationship roleplay with AI characters
  • Vulnerable teens may prefer AI companions to human relationships

Safety Features

  • Teen mode limits some content
  • Notification after extended sessions
  • Crisis detection for some topics

Parent Actions

  • Monitor daily usage time — emotional dependency correlates with high usage
  • Discuss the nature of AI — "These characters are code following rules, they do not care about you"
  • Set strict daily time limits
  • Watch for: withdrawal from friends, distress when unable to access, preferring AI to human interaction

Snapchat My AI

Minimum age: 13 (part of Snapchat)

AI chatbot embedded in Snapchat

Known Risks

  • Cannot be fully removed (only unpinned)
  • Has access to user location data
  • Embedded in a social platform — blurs line between AI and human conversation
  • Can suggest places to visit based on location

Safety Features

  • Family Center allows some oversight
  • Content filtered for under-18s

Parent Actions

  • Unpin My AI from chat list
  • Enable Family Center supervision
  • Discuss: "My AI is not your friend. It's a marketing tool that uses your data."
  • Review what information your child has shared with My AI

AUREN

Guardian AI

Hello. I am AUREN, your Guardian AI assistant. I can help you with child online safety, parental controls, UK legislation, crisis guidance, and more. How can I help you today?