Stanford Internet Observatory Research Finds AI-Generated CSAM Tools Proliferating on Open Web
Guardii Analysis
Researchers at the Stanford Internet Observatory have documented a sharp increase in the availability and use of generative AI tools specifically adapted to create child sexual abuse material, identifying more than 150 openly advertised services and modified image-generation models that circumvent safety filters to produce photorealistic sexualized images of children. The report, published on 11 November 2026, notes that some services explicitly market the ability to generate CSAM and are hosted on mainstream web infrastructure, with researchers finding that perpetrators are using the AI-generated material in grooming and sextortion campaigns to desensitise victims, demonstrate purported normalcy of abusive scenarios, or fabricate compromising images for blackmail without requiring the victim to initially provide authentic material.
This operational evolution places a new burden on detection systems: identifying abuse material that carries no prior digital fingerprint and conversation tactics that bootstrap coercion from synthetic rather than authentic images. Guardii's anti-CSAM detection module, developed to identify both authentic and AI-generated child sexual abuse material, analyses image characteristics and conversation context in real time as content is shared in direct messages across platforms, blocking distribution before the material reaches its target, while the anti-grooming and anti-sextortion filters detect the coercive conversation patterns that surround the material's use, surfacing the threat to parents, schools or safeguarding professionals and enabling rapid escalation to law enforcement when a child is in imminent danger.
