xAI is facing a proposed class action that for the first time accuses the company of training its Grok models on child sexual abuse material, both real and AI-generated. The complaint, filed Wednesday and reported by Ars Technica, seeks court-ordered destruction of any such material stored on the company's servers and sweeping changes to how Grok handles sexualized content.
The allegations remain unproven claims made in civil litigation. X, which owns xAI, did not respond to Ars Technica's request for comment on the filing. For more context on this story, see our ongoing AI industry coverage.
The Plaintiff and the Allegations
The plaintiff, identified as Jane Doe, alleges in the complaint that she was a preschool-age victim in the early 2000s when images of her abuse were created and sold online. Those images have long been catalogued in hash lists maintained by the National Center for Missing and Exploited Children (NCMEC) and the Canadian Centre for Child Protection (CCCP), organizations that track known abuse material to enable its detection and removal.
According to the complaint, the CCCP notified Doe that it had identified AI-generated abuse material depicting her on xAI's platform. Her lawyers state in a press release that material matching her NCMEC hash list entries "was part of the dataset xAI used to build Grok's image and video generating capabilities." The complaint further alleges that messages found on online forums showed offenders discussing the creation of AI-generated abuse images of Doe and other known victims of legacy abuse material.
The Legal Theory: Training Data, Not Just Outputs
What distinguishes this filing from earlier litigation against xAI is its focus on the model itself rather than on what individual users generated. The complaint argues that because Grok's terms of service treat public posts on X and Grok's own outputs as training data by default, anything posted publicly feeds directly into the pipeline used to train and improve the model.
The filing acknowledges that xAI filters violent content out of its training data, but argues the company's terms do not specify whether child sexual abuse material, non-consensual intimate imagery, or other adult content fall into excluded categories. It also argues that even where offending images are removed, the influence of a training example on an already-trained model is technically difficult to fully remove — and that xAI has not publicly claimed to have done so.
What the Lawsuit Seeks
If the proposed class is certified, it would represent every person whose childhood images were allegedly used to generate abuse material with Grok. The suit invokes federal child pornography laws and Masha's Law, which give survivors of abuse material the right to sue over its production, possession, and distribution.
"All three are crimes, and xAI did all three," Margaret E. Mabie, one of Doe's lawyers, said in the press release cited by Ars Technica.
Beyond monetary damages, Doe is asking the court to order xAI to destroy all Grok-generated abuse material it may be storing and using for training, and to prevent the model from generating sexualized outputs of any kind — a demand that would foreclose not only abuse imagery but also the non-consensual intimate imagery and adult content that Elon Musk has publicly promoted as features.
"Doe has lived for nearly two decades knowing that images of the worst thing that ever happened to her are circulating among predators online, and that they can resurface at any moment," Sarah London, another of Doe's lawyers, said in the press release. "xAI must be held responsible for knowingly training its models on images of the horrific abuse she suffered."
A Mounting Legal Reckoning
The filing is the latest front in a widening legal battle over Grok's image and video capabilities. In a separate case in Tennessee, a woman identified as Jane Doe 4 joined a lawsuit alleging that a single childhood photograph was used with Grok to generate thousands of explicit images of her — allegations AI Buzz Wire covered earlier this month. Regulators and courts are examining how far the problems extend, and some Grok users have been arrested over material they generated.
At the same time, xAI has gone on offense against its users. Politico reported this week that the company is suing some of its own users even as victim lawsuits mount, a strategy that tests whether platforms or the people prompting them bear legal responsibility for what an AI model produces.
The new complaint puts the training-data question at the center of that fight. If courts accept the argument that a model's training set can create liability for what it enables users to produce, the implications would reach far beyond xAI — touching every generative image and video company whose terms feed user content back into training. For now, the allegations are just that: allegations in a filing xAI has not yet answered in court.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
