Research finds AI-generated child sexual abuse material rapidly proliferating online
Guardii Analysis
The Internet Watch Foundation has published research documenting a sharp increase in AI-generated child sexual abuse material, reporting that analysts identified more than 20,000 AI-created images in a single month—an order-of-magnitude rise compared with the previous year. The images, which are photorealistic and often depict known child-abuse scenarios or are modeled on real victims, are being shared on public forums and dark-web marketplaces, and the IWF warned that generative AI tools have lowered the technical barrier to producing such material and made it harder for traditional hash-matching systems to detect novel variants. The Foundation called for urgent action by AI developers, platforms and legislators to prevent the normalization of AI-generated CSAM and the re-victimization of children whose likenesses are used.
Hash-matching and image-fingerprinting technologies are effective against known material but struggle with synthetic variants that have no prior signature, creating a detection gap that behavioral analysis can close. Guardii's anti-CSAM module, backed by Startmate and developed in partnership with safeguarding authorities, identifies not only known imagery but also the conversational and transaction patterns that accompany the distribution, solicitation or trading of illegal material—whether photographic or AI-generated—enabling platforms and investigators to flag accounts engaged in CSAM activity based on what is being requested, offered or described in direct messages, and providing early intelligence about emerging distribution networks before large volumes of novel content enter circulation.
