# Should Children Use AI? Shielding vs AI Literacy

> Age limits and filters cannot keep children away from AI: most teens already use it. Why teaching children how AI works, with safe tools, protects them better.

URL: https://www.clarityailab.com/insights/ai-safety-strategy-shielding-vs-double-literacy-exposure/

Keeping children away from AI is no longer a realistic safety plan on its own: seven out of 10 American teens have already used a generative AI tool, often without their parents knowing. Age limits and filters still matter, but children are safer when they also learn how AI works and how to question it, and when the tools they use are built for them.

This article is for parents and educators deciding how to handle AI at home or in class. It compares two strategies, shielding and what we call double literacy, and sets out what works best together.

## How many children already use AI?

More than most adults think. Common Sense Media's 2024 survey, [*The Dawn of the AI Era*](https://www.commonsensemedia.org/press-releases/new-report-shows-students-are-embracing-artificial-intelligence-despite-lack-of-parent-awareness-and), found that seven out of 10 US teens had used at least one type of generative AI tool. Only 37% of parents whose teen had used one thought their child had.

Use goes beyond homework. In 2025 the same organisation reported that [nearly three in four teens have used AI companions](https://www.commonsensemedia.org/research/talk-trust-and-trade-offs-how-and-why-teens-use-ai-companions), chatbots designed for conversation and relationships, and half use them regularly.

## What does shielding mean, and why is it not enough?

Shielding means keeping children away from AI or blocking what it can show them. It has three main tools:

- **Age limits.** OpenAI's [terms of use](https://openai.com/policies/terms-of-use/) require users to be at least 13, and under-18s need a parent's or legal guardian's permission. UNESCO's [guidance on generative AI in education](https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research) also proposes an age limit for independent conversations with these tools.
- **Privacy law.** In the United States, the [Children's Online Privacy Protection Rule](https://www.ecfr.gov/current/title-16/chapter-I/subchapter-C/part-312) (COPPA) defines a child as under 13 and requires verifiable parental consent before a service collects a child's personal information. It protects data. It does not govern whether an AI's answers are true or suitable.
- **Filters and blocks.** Content filters and school network blocks reduce exposure to harmful material, but they do not catch every answer, and they stop at the school gate.

These are necessary. The survey figures show they are not sufficient: children reach AI anyway, through phones, friends and apps their parents have not heard of.

## What is double literacy?

Double literacy is the name we use for teaching children two things together:

1. **How AI works.** A chatbot predicts likely words from patterns in its training data. It can sound confident and still be wrong, and it does not understand or care about the child.
2. **How to judge what it says.** Check important answers against another source, notice when an answer flatters or pressures you, and ask an adult when something feels wrong.

*Analysis:* the aim is the same as teaching road safety. Children are not kept off roads for ever; they learn the rules and practise with an adult until they can cross alone.

## Does the evidence support teaching over blocking?

The strongest evidence is about how AI is used, not whether. In a field experiment with nearly 1,000 high school maths students, [Bastani and colleagues](https://doi.org/10.1073/pnas.2422633122) found that students given an unrestricted chatbot did better during practice but worse on a later exam taken without it. Students given a tutor version with safeguards, which gave hints rather than answers, did not suffer that drop. Design and guidance changed the outcome.

The risks of the wrong kind of AI are also clear. In April 2025 Common Sense Media concluded that [social AI companions pose unacceptable risks](https://www.commonsensemedia.org/press-releases/ai-companions-decoded-common-sense-media-recommends-ai-companion-safety-standards) to children and teens under 18 and should not be used by minors. That is a case for shielding from one category of product, alongside teaching about the rest.

## What is the best approach: shielding, literacy or both?

Both, matched to age:

| Age | Shielding | Literacy |
|---|---|---|
| Under 13 | Child-specific tools only, used with an adult; no general chatbots or companion apps | Simple ideas: AI can be wrong; it is a program, not a friend |
| 13 to 17 | Parental permission, no AI companions, clear household rules | Checking sources, spotting flattery and pressure, protecting personal information |

Tools built for children help. Our own product, Clarity Kids, is designed for ages 6 to 12, screens every interaction with a multi-stage safety filter, carries no ads or tracking, and never uses children's data for advertising. Whatever tool a family chooses, the same questions apply: who is it for, what does it filter, and what does it keep?

## What can parents and teachers do this week?

1. **Ask, do not assume.** Ask children which AI tools they use; the survey shows parents often do not know.
2. **Use AI together first.** Try a question side by side, and check the answer together against a book or trusted website.
3. **Name the limits.** Explain that AI can be wrong, can make things up, and is not a person.
4. **Rule out companion apps for minors.** Follow Common Sense Media's advice for under-18s.
5. **Check the privacy terms.** Prefer tools that ask for parental consent, keep little data and say when they delete it.

For a fuller framework for designing AI for children, see [Child-Safe AI: A Framework for Learning and Privacy](/insights/child-centric-ai-framework-cognitive-potential-digital-privacy/).

### Is ChatGPT safe for a young child to use?

Not on their own. OpenAI's terms require users to be at least 13, and under-18s need a parent's or guardian's permission. Younger children should use AI only with an adult, or through tools designed and filtered for their age.

### What is double literacy?

It is our name for teaching children two things together: how AI works, for example that a chatbot predicts likely words and can be wrong, and how to judge what it tells them by checking it against other sources and asking an adult.

### Does COPPA protect children from harmful AI answers?

Not directly. The US Children's Online Privacy Protection Rule governs personal information from children under 13, including verifiable parental consent and limits on keeping data. It does not regulate the quality or safety of what an AI says.

### Should teenagers use AI companion apps?

Common Sense Media concluded in April 2025 that social AI companions pose unacceptable risks to children and teens under 18 and should not be used by minors. Its later survey found nearly three in four teens had used one.