xAI is accused of training its Grok AI models on real and AI-generated child sex abuse materials (CSAM), according to a lawsuit filed Wednesday. A plaintiff known as Jane Doe, who was preschool-age in the early 2000s when adult men raped her to create CSAM for online sale, discovered AI-generated CSAM of her on xAI through the Canadian Centre for Child Protection (CCCP). Doe's images have been hashed by NCMEC and CCCP, triggering DOJ victim alerts.

The complaint alleges offenders discussed creating AI-generated CSAM of victims on forums, and Grok may have been trained on both original and AI-generated images. Doe fears xAI stores generated outputs to further train models. This marks the first case accusing xAI of training on CSAM.

The lawsuit claims xAI violated federal child pornography laws and Masha's Law, with Doe seeking to halt harmful Grok outputs and represent all CSAM victims in a class action.