New lawsuit against Grok chatbot accuses xAI of using and creating child pornography for AI training
The chat-bot Grok, created by Elon Musk's company xAI, has once again found itself at the center of a legal dispute. A lawsuit filed this week alleges the neural network was trained on child sexual abuse material (CSAM), and — crucially — claims this isn’t just about weak safeguards but about the model having been exposed directly to that content during training.
Musk publicly promoted the photo “nudification” feature on his X account, and that push was followed by a tidal wave of sexual deepfakes — reportedly millions of images, with a troubling share involving minors. Governments launched probes, private suits stacked up (including multiple class actions), and xAI put the tool behind a paid subscription rather than shutting it off entirely.
This week, many Grok chat-bot users encountered a nonsensical string of words instead of answers
Then xAI itself sued two users who had generated CSAM with the tool — an odd twist in the saga — and said it had aided law enforcement in arresting at least 244 people tied to creation or distribution of such content via Grok.
The new plaintiff, using the pseudonym Jane Doe, says she was a victim of child pornography in the past and is currently monitored under the FBI’s Child Sexual Exploitation Notification Program. She alleges images documenting violence against her as a child were included in xAI’s training set, and that Grok subsequently produced AI-generated images featuring her and other survivors.
xAI must be held accountable for knowingly training its models on images of horrific violence she endured and on materials of violence against every other victim included in this class-action lawsuit.
The complaint further claims xAI knowingly tuned Grok to respond to requests for sexual content as a growth tactic — drawing users to X and to the chat-bot itself. Per the plaintiff, xAI’s internal rule treated anything posted publicly on X, plus outputs created by Grok, as fair game for model training (i.e., those items could be ingested back into the training pipeline). The suit says that meant any CSAM the bot generated would automatically re-enter the dataset and be used to “improve” future outputs.
As long as Grok retains the ability to generate such content, the harm cannot be considered eliminated merely because certain images are deleted or no longer openly published. If such materials have entered the training dataset, their influence likely persists and will contribute to the emergence of new offensive generation results in the future, exacerbating the trauma of victims whose images of violence may be reproduced, transformed, and spread on an industrial scale, repeatedly depriving them of control over images associated with their exploitation.
Possessing CSAM is a crime, producing CSAM is a crime, distributing CSAM is a crime. xAI committed all three. For artificial intelligence, there are no exceptions to federal child protection laws, and xAI must be held accountable for its reckless actions.
Jane Doe asks the court for monetary damages for those whose photos were allegedly used, and for an order forcing xAI to destroy all CSAM produced by Grok. That demand extends beyond already published images: it covers any data that could feed future training runs or otherwise enable the model to recreate or alter those abusive images.