UNICEF finds 20 million children abused through everyday apps in one year

UNICEF’s Innocenti office released Through Children’s Eyes on 3 September. Across 21 countries, about 20 million internet-using children aged 12 to 17, roughly one in five in those countries, were sexually exploited or abused through digital tools in a single year.

The report’s counts are blunt. More than 15 million children were shown sexual material they did not want. Nine million were pressed into sexual talk or into sending images. Four million had sexual images of themselves passed on without consent. Nearly 60 percent of the recorded cases happened on platforms the report names: Facebook, Instagram, Snapchat, TikTok and WhatsApp. Fourteen percent sat inside online games. The features those products use to hold attention were the same features used to reach the children.

In nine countries where the question was asked, about 1.1 million children said someone had made AI-generated sexual images or videos of them. That figure is new in UNICEF’s public file and is the part of the report that will follow product lawyers the longest.

Reporting is almost absent. More than 40 percent of children told no one. Fewer than 1 percent reached police, a social worker or a helpline. Children who had been targeted were about four times more likely to report suicidal thoughts or self-harm and about 11 percentage points higher on anxiety scores than peers who had not. In Mexico and North Macedonia the link to suicidal thoughts was more than eight times higher. In Pakistan the link to self-harm was more than seven times higher. UNICEF Executive Director Catherine Russell said many children “suffer in silence, unsure where to turn.”

The agency wants platforms to change design, not only write policies. It wants faster takedown, age-appropriate defaults, and laws that treat the company as a party to the harm when its tools are the room in which the harm happens. None of that is a new slogan. The new fact is the scale inside a defined set of countries and the AI number sitting beside the older categories.

What the report does not do is useful to say as well. It does not claim the 21 countries represent every child on Earth. It does not break out every platform’s exact share beyond the grouped 60 percent. It does not prove that any one design change would have prevented a given case. It does show that the abuse is not a dark-web specialty. It is happening on the apps already installed on household phones, and almost no one in authority hears about it in time to act.

Governments that regulate those apps now have a document dated September 2026 with a 20 million figure, a 1 percent reporting rate and a 1.1 million AI figure. The next test is whether any legislature writes those numbers into a duty that a company can be fined for missing.