New research finds existing federal laws successfully prosecute AI-generated CSAM offenders

Stanford Law School and the University of Edinburgh released a dataset and analysis of seventy-one federal criminal prosecutions filed in the United States between 2022 and 2025 involving AI-generated or AI-morphed child sexual abuse material and child obscenity. The research found that existing federal criminal laws are sufficient to successfully prosecute these cases: most defendants pleaded guilty, all defendants who proceeded to trial were convicted, prison sentences are longer than average for traditional CSAM or obscenity offences, and only six cases out of seventy-one remain pending. A notable exception is a Seventh Circuit ruling in August 2026 that held the First Amendment protects private possession of fully AI-generated obscene material depicting imaginary children, though the researchers note that this ruling applies narrowly and that prosecutions overall are proceeding effectively under current statutes.

The findings establish that law is catching conduct after the fact, but the operational gap remains at the front end: preventing the harm before imagery is created and before a child becomes a target. A world-leading AI safety platform backed by Startmate, Guardii deploys detection modules calibrated to identify AI-generated and deepfake child sexual abuse material in real time, flagging or blocking such content before it reaches a child or circulates within a peer network, and reporting the seriousness and shape of the threat to the responsible adult or authority rather than creating a feed of what was said. Where research documents successful prosecution, targeted real-time detection addresses the operational priority that no prosecution can satisfy: stopping the harm before it occurs.