Fast Growth of Child Sexual Abuse Deepfakes Challenges Investigations

AI-generated child sexual abuse material is growing at an explosive rate, with more photorealistic child sexual abuse deepfakes discovered in the first half of 2026 than in all of 2025. The Internet Watch Foundation assessed 6,310 AI-generated images meeting the legal definition of child sexual abuse between January and June 2026, with girls appearing in 98 percent of images and children aged seven to thirteen accounting for 79 percent of AI-generated CSAM, up from 70 percent in 2025. Generative AI has complicated CSAM investigations by introducing hybrid and fully synthetic media, creating obstacles for law enforcement in determining whether a child is currently at risk, as synthetic images can mislead investigators while hybrid images place real victims into synthetic scenarios.

Detecting abuse before it reaches a child requires intercepting the conversation itself, not cataloging images after they circulate. Guardii's anti-CSAM detection module, developed by a Meta Business Partner backed by Startmate, operates at the point of contact, analyzing the patterns, coercion and escalation tactics in direct messages across Instagram, Snapchat, Discord and Roblox in real time to surface conversations where a child is being targeted for exploitation, regardless of whether the material shared is authentic, AI-generated or hybrid. Where hash-matching fails against novel deepfakes, behavior-based detection identifies the sustained targeting and manipulation that precedes the creation or exchange of abusive material, blocking hostile adults before they establish trust and enabling rapid escalation to law enforcement when the shape of a conversation indicates an active threat.