xAI Accused of Training Grok AI Models on Child Sex Abuse Material in Lawsuit

xAI, the artificial intelligence company, faces a lawsuit alleging its Grok models were trained using child sex abuse material (CSAM), including both real and AI-generated content. A complaint filed by a plaintiff identified as Jane Doe claims her childhood images were part of the dataset used to develop Grok's image and video generation capabilities.

Jane Doe, a survivor of childhood CSAM from the early 2000s, was notified by the Canadian Centre for Child Protection (CCCP) that AI-generated CSAM depicting her had been identified on xAI. Her original images had been hashed by groups like the National Center for Missing and Exploited Children (NCMEC) for identification. The discovery of new AI-generated images re-traumatized Doe, whose complaint alleges online forum discussions among offenders about creating such content using her and other known victims. She fears xAI has facilitated the creation of more violative images and that the company stores these generated outputs, using them to further train Grok. Doe believes Grok has been trained on both her original CSAM and the more recent AI-generated versions.

Allegations of Training Data Misuse

This lawsuit marks the first time xAI has been accused of training its models on CSAM. Lawyers representing Doe stated that her images were part of an NCMEC CSAM Hash List, and this material was allegedly included in the dataset xAI used for Grok's image and video generation. The complaint specifically alleges that CSAM depicting Doe, with its known hash values, was part of xAI's dataset. The lawsuit also details how Grok's training process might incorporate harmful content. xAI's terms of service treat public X posts and Grok's own outputs as default training data. While xAI filters violent content from Grok's outputs to exclude it from training, the terms do not explicitly state that CSAM, non-consensual intimate imagery (NCII), or NSFW material are excluded categories. The complaint suggests that any CSAM ingested into training before removal would likely continue to influence the model's outputs, even if the original images were later taken down.

Legal Action and Demands

Doe's proposed class action seeks to end Grok's harmful outputs. If approved, the class would include all victims whose childhood images were used to generate Grok CSAM. The lawsuit accuses X, xAI's parent company, of violating federal child pornography laws and Masha's Law, which grant CSAM survivors the right to sue for production, possession, and distribution. Margaret E. Mabie, a lawyer for Doe, indicated that xAI allegedly committed all three of these actions. A successful outcome for Doe could result in xAI owing money damages to victims who can prove Grok generated CSAM.