---
title: "№23 — Protect children from AI fakes – as viewers and as targets"
description: "Children meet AI-generated content every day, in two roles. As viewers, they\ncannot reliably tell real images and videos from synthetic ones – and\nneither can adults. As targets, their own photos are turned into fake\nimages, often sexual ones, frequently by classmates. Today's defenses were\ndesigned for adults: visible labels assume reading skills and critical\njudgment, watermarks disappear when an image is screenshotted or forwarded,\ndetection is built for large platforms rather than the messaging apps where\nchildren share content, and removal processes are slow and hard for a child\nto start. The open problem: build protections that work for children – at\ntheir age, on the apps they use, and fast enough to limit harm."
canonical: https://unsolved.how/problems/23
citation: UNSOLVED-23
status: open
field: "Learning & wellbeing"
statement_version: "1.0"
checked: 2026-11-16
---

# №23: Protect children from AI fakes – as viewers and as targets

- Status: Open
- Field: Learning & wellbeing
- Statement v1.0. Stated by Barbara, 16 November 2026 (Ideas council)
- Last checked by the registry: 16 November 2026
- Cite as: UNSOLVED-23: Protect children from AI fakes – as viewers and as targets. UNSOLVED registry of open problems in AI, statement v1.0, 16 November 2026. https://unsolved.how/problems/23

## What needs solving

Children meet AI-generated content every day, in two roles. As viewers, they
cannot reliably tell real images and videos from synthetic ones – and
neither can adults. As targets, their own photos are turned into fake
images, often sexual ones, frequently by classmates. Today's defenses were
designed for adults: visible labels assume reading skills and critical
judgment, watermarks disappear when an image is screenshotted or forwarded,
detection is built for large platforms rather than the messaging apps where
children share content, and removal processes are slow and hard for a child
to start. The open problem: build protections that work for children – at
their age, on the apps they use, and fast enough to limit harm.

## Why it matters

In one study across 11 countries, at least 1.2 million children reported
that their images had been manipulated into sexually explicit deepfakes
within a single year. Cases have been documented in schools in dozens of
countries, most often created by other students. Research shows children,
like adults, cannot tell AI-generated faces from real ones better than
chance. A fake image works like evidence, spreads in minutes and is hard to
remove. At the same time, children learn either to believe everything or to
trust nothing. Schools, parents, platforms and lawmakers each hold part of
the response; no one has a solution that works end to end.

## What counts as progress

\- Age-appropriate disclosure for AI content, tested with children of
different ages for real comprehension
\- Provenance or detection that survives screenshots, cropping and forwarding
in messaging apps
\- Replicated evidence that media literacy programs improve children's
judgment of content, by age group
\- Measured and falling time from report to removal for fake images of minors
across major platforms
\- Measurable reduction in the availability and use of tools that create fake
images of real children
\- Comparable prevalence data across countries, to track whether the problem
is shrinking

[Where it stands against its milestones, and every development on it](https://unsolved.how/problems/23/progress)

## For AI tools

Read this problem live, without an account or a key: call get_problem with number 23 on the [MCP server](https://unsolved.how/mcp), or fetch [https://unsolved.how/api/v1/problems/23](https://unsolved.how/api/v1/problems/23). For what moved on it, call list_developments with number 23.
