The Pulse
IWF Finds 6,310 AI-Generated Child-Abuse Images in Six Months
The Internet Watch Foundation identified 6,310 AI-generated images that met the legal definition of child sexual abuse in the first half of 2026, 40% more than in all of 2025. The organization is urging EU policymakers to advance legislatio
AI.info Team · Written with AI ·
6,310 images in six months
The Internet Watch Foundation identified 6,310 AI-generated images that met the legal definition of child sexual abuse between January 1 and June 30, 2026. That is 40% more than the 4,512 images its analysts identified during all of 2025. The figures count material assessed by the IWF, not every image online, but the six-month comparison marks a steep rise in what the watchdog is finding.
The IWF says its analysts found photorealistic material depicting children under two, alongside a broader increase in images of younger children. The organization is calling on EU policymakers to advance long-delayed legislation intended to give technology companies legal certainty to detect and remove criminal material. The figures sharpen the debate over whether those rules should cover newly generated material as well as files already identified by authorities.
Girls and younger children make up most identified images
Among the 6,221 images for which analysts recorded both age and gender, girls appeared in 98%, according to the IWF. The organization counted 6,094 unique images depicting girls by the end of June, compared with 4,259 across all of 2025. These are counts of images, not a count of individual children.
Analysts identified 2,534 images depicting children aged seven to 10 and 2,369 depicting those aged 11 to 13 in the first half of 2026. Another 1,004 depicted children aged three to six, while 190 depicted infants and toddlers under two. The 4,903 images depicting children aged seven to 13 accounted for 78.8% of the 6,221 images for which both age and gender were recorded, up from 70% in 2025.
Manipulated real images can renew harm
The IWF warns that some offenders use AI tools to alter genuine images of victims, creating new material that can further traumatize survivors. To limit circulation, the organization turns confirmed criminal images into digital fingerprints, or hashes, which it shares with law enforcement and technology companies. Services can use those fingerprints to identify and block known files without storing or viewing the original images.
Hash matching has a clear limit: it works on material that has already been identified and fingerprinted. A newly created image has no existing hash, so it must first be discovered and assessed before the method can help stop its spread. That gap is central to the IWF’s argument for rules that address previously unseen material.
IWF presses the EU on detection rules
The IWF wants the proposed Child Sexual Abuse Regulation to let services use proportionate tools to detect, prevent and respond to both known and unknown material. IWF Chief Executive Officer Kerry Smith said:
“It is deeply concerning to see how quickly AI-generated child sexual abuse imagery continues to grow and spread online.”
The IWF’s data also shows a shift in the severity categories assigned under UK law. Category C images, which can include sexualized posing and nude imagery, numbered 5,557 in the first half of 2026 and made up 88% of AI-generated images, compared with 2,842 and approximately 63% across 2025. Analysts also identified 350 Category A images and 403 Category B images in the first six months of 2026; Category A is the most severe classification, while Category B can include non-penetrative sexual activity.
The policy distinction is concrete: a system built only to match known files cannot identify a new image by its hash. The IWF is asking EU lawmakers to give platforms a legal basis to detect both categories of material as negotiations on the regulation continue.