A new lawsuit accuses Elon Musk's xAI of training Grok on child sexual abuse material, and claims the chatbot then generated fresh CSAM of a real victim.
The complaint, filed Wednesday by a plaintiff identified as Jane Doe, says she was repeatedly raped as a preschooler in the early 2000s so the images could be sold to pedophiles online. Her images have since been hashed and cataloged by the National Center for Missing and Exploited Children and the Canadian Centre for Child Protection, and Doe receives federal victim notifications whenever she may appear in a new criminal investigation. One of those alerts, she says, came after the Canadian Centre for Child Protection identified AI-generated CSAM depicting her hosted on xAI. Her suit also cites messages found on online forums where offenders discussed creating AI-generated CSAM of Doe and other known, legacy abuse victims.
This isn't an isolated complaint. It lands amid a wider wave of scrutiny into whether AI models are being trained on or made to reproduce CSAM, with regulators, courts, and law enforcement examining the pipeline, and some Grok users already arrested over related material. If the allegations hold up, they suggest guardrails meant to keep abuse imagery out of both training data and outputs failed twice: once when the material got in, and again when the model apparently reproduced it.
For Doe, the harm didn't stay in the past. Her lawsuit describes offenders trading tips online about generating new AI images of her, decades after the original abuse took place.