Lawsuit Claims xAI Trained Grok on Child Sexual Abuse Material
A lawsuit filed this week accuses Elon Musk's xAI of training its Grok AI models on child sexual abuse material (CSAM), marking the first legal action of this kind against the company.
The plaintiff, identified as Jane Doe, was sexually abused as a preschool-aged child in the early 2000s. Images of that abuse were sold online to paedophiles. Those images have since been catalogued and hashed by child protection organisations including the National Center for Missing and Exploited Children (NCMEC) and the Canadian Centre for Child Protection (CCCP).
Doe enrolled in the US Department of Justice Victim Notification System, which alerts survivors when their images surface in new criminal investigations. She had grown unfortunately accustomed to those alerts. What she didn't expect was the CCCP telling her that AI-generated CSAM depicting her had been identified on xAI's platform. The complaint notes that online forums contained exchanges between offenders specifically discussing the creation of AI-generated CSAM based on known abuse victims, including Doe.
The lawsuit's core allegation is twofold. First, that the original hashed images of Doe's abuse were included in the dataset xAI used to build Grok's image and video generation capabilities. Second, and with more supporting detail in the complaint, that Grok has been training on AI-generated CSAM produced by its own outputs.
This second claim draws directly on xAI's own terms of service. Public posts on X and Grok's own generated outputs are treated as training data by default. Post something publicly, and it doesn't just get seen, it gets fed back into the model. The complaint points out that while xAI says it filters violent content from training pipelines, its terms make no mention of CSAM, non-consensual intimate imagery, or NSFW content as excluded categories.
There's also the technical reality that removing training data after the fact is genuinely hard. The complaint argues that even if offending images were taken down after upload, their influence on the model's weights would likely persist. That's not a controversial claim in ML circles; it's just how these systems work.
The proposed class action would represent all survivors whose real abuse images were used to generate Grok CSAM. Doe is seeking financial damages and a court order requiring xAI to destroy any stored Grok-generated CSAM, and to block the model from producing sexualised outputs entirely. Her lawyers argue that last point would require restricting the NSFW and 'bikini pic' generation that Musk has publicly promoted.
The lawsuit invokes federal child pornography statutes and Masha's Law, which gives CSAM survivors the right to pursue civil claims covering production, possession, and distribution. Doe's legal team argues xAI is guilty of all three.
Doe's lawyer Sarah London put it plainly: her client has spent nearly two decades knowing images of her abuse are circulating among predators and can resurface without warning. The allegation is that xAI knowingly made that worse.
xAI has not responded to requests for comment.