TL;DR
Childlight, the Edinburgh-based Global Child Safety Institute, says reports of AI-related child sexual abuse material climbed by over 8,400% in two years. Its count for 2025 tops 400,000, against roughly 4,700 two years earlier, the Guardian reported on 16 November. The figures come from reports that tech companies filed with NCMEC, the US clearing house for missing and exploited children, and they include deepfakes produced with “nudify” apps.
What the research measured
This round of Childlight research covers North America, Latin America with the Caribbean, and east Asia with the Pacific; western Europe was the subject of earlier work. The Guardian notes that, fake or not, such images are used to extort children for money or more pictures.
Professor Debi Fry, Childlight’s global director of data, called the scale of exploitation “shocking”. She accepted that better detection and reporting may account for part of the increase, but said the findings show why governments and technology firms “must act urgently” to prevent these forms of abuse and take material down quickly.
The report also tracks wider trends. In the US, the rate of recorded reports and notices about abuse material grew 63% from 2024 to 2025; Canada’s rose 18%.
The UK picture
The Guardian set the findings beside recent UK data. In October the Internet Watch Foundation, a UK-based organisation that monitors abuse material globally, reported 6,310 AI images so far in 2026 that it judged to be illegal child abuse imagery. That is already 40% above its 2025 tally of just over 4,500, and the IWF says it has seen more photorealistic material in the first six months of this year than in all of last year.
Paul Stanfield, Childlight’s chief executive, said governments often work with “an incomplete picture” of where harm occurs. He wants them to treat child sexual exploitation and abuse as “a preventable public health emergency” and to invest in better evidence.
Looking Forward
In our view, Fry’s caveat matters: a jump in reports partly reflects better detection, so the size of the jump should not be read as a precise measure of new abuse. Even so, the IWF’s own assessments point the same way. We expect UK platforms and AI image tool providers to face growing pressure to show that detection and removal keep pace with what generative tools can now produce.