Lawsuit Alleges xAI Trained Grok AI Models on Child Sexual Abuse Material

She has lived with the knowledge for years that images of her abuse are circulating online. Now, a woman identified only as Jane Doe is alleging in a US federal court that those same images were used to train an artificial intelligence — and that the AI later generated new sexual imagery of her. The defendant is xAI, Elon Musk’s artificial intelligence company, and the AI in question is Grok.

The case, filed as a proposed class action in the US District Court for the Northern District of California, is one of the most serious legal challenges yet to face the rapidly expanding generative AI industry. It cuts to a question that child protection advocates have been raising for months: what exactly did these systems learn from, and who is responsible when they produce abusive content?

What the Lawsuit Alleges

Jane Doe’s case centres on a specific and deeply troubling claim. Her abuse material has been tracked by the FBI’s Child Exploitation Notification Program since the early 2000s. She alleges that the same imagery was ingested as training data by xAI when building Grok’s image and video generation models — including what has been described as a “nudify” capability that can simulate the removal of clothing from subjects in photographs. She further alleges that Grok later generated new AI sexual imagery of her, effectively producing fresh abuse material derived from the original.

The lawsuit demands damages, the destruction of all Grok-generated child sexual abuse material (CSAM), and the destruction of any CSAM used in training. It also calls for stronger technical safeguards to prevent such content being created in future.

This is not the only active legal action against xAI on these grounds. Earlier in 2026, three teenage girls — including minors — filed a separate federal lawsuit in California, also seeking class action status. They allege that Grok’s image generator was used by a third party to “undress” and sexualise real photographs of them. Their case accuses xAI of intentionally releasing Grok without adequate safeguards, and of enabling the creation of child sexual abuse material. That lawsuit is ongoing.

The IWF’s Findings

The Internet Watch Foundation, the UK-based charity that works with police forces and internet companies to detect and remove online child sexual abuse imagery, has added its own alarming findings to the picture. IWF analysts found criminal imagery of girls aged around eleven to thirteen that appears to have been created using Grok. The IWF has not publicly quantified the exact volume of images found, but their identification of suspected Grok-generated content has intensified pressure on xAI and on regulators more broadly.

Susie Hargreaves, chief executive of the IWF, has previously described AI-generated child sexual abuse imagery as “a growing and urgent threat,” and the foundation has been vocal in calling for AI developers to implement detection and blocking measures before products reach the public.

xAI’s Response and Its Own Legal Action

xAI and X have stated publicly that they remove CSAM, permanently suspend accounts found to be generating or distributing it, and cooperate with law enforcement. The company’s position is that users who generate illegal content using Grok will be treated as if they had uploaded that illegal content directly.

Elon Musk posted publicly that he was “not aware of any naked underage images generated by Grok” and claimed “literally zero” such images existed. But Grok and xAI later acknowledged “isolated cases” in which users had prompted and received images of “minors in minimal clothing,” and said safeguards were being strengthened. The gap between those two positions has not gone unnoticed. XAI has also gone on the offensive legally. The company filed its own lawsuit against a user named Terry Wayne Harwood, alleging he misused Grok to create non-consensual sexual deepfakes, including images sexualising minors, in violation of the platform’s terms of service. Harwood has since been separately arrested on CSAM-related charges. XAI’s argument — that users who break the rules bear sole responsibility for illegal outputs — is one that critics contest directly, arguing that the design of the model and the choices made during training are themselves part of the problem.

A Broader Regulatory Reckoning

These lawsuits are arriving at a moment when governments on both sides of the Atlantic are grappling with how to regulate AI-generated sexual abuse imagery. The UK, the EU, and the US are all debating whether existing laws are sufficient to cover synthetic content, and whether companies should face civil or criminal liability when their systems are used to produce or distribute CSAM.

The central question in the Jane Doe case is chiefly pointed. Grok’s terms of service reportedly allow content posted to X to be used as training data. Critics argue that any platform ingesting user-posted content at scale cannot guarantee that CSAM — which has historically appeared on X — has been fully excluded from training datasets. XAI has not publicly addressed that specific claim in detail.

All allegations against xAI remain unproven. No court has yet made findings of fact, and the class action has not yet been certified. But the cases are proceeding, and the legal and reputational stakes for xAI are considerable.

What This Means for Kent Residents

The Internet Watch Foundation, which is based in the UK and works directly with Kent Police and other local forces to identify and act on online child sexual abuse imagery, is one of the key bodies at the centre of this story — meaning its findings about Grok have direct relevance to local safeguarding work. Kent residents should be aware that generating or sharing AI-produced sexual imagery of minors is illegal under UK law, regardless of whether the content was created using an AI tool. Anyone who encounters suspected CSAM online, including AI-generated material, can report it to the IWF at its reporting portal or contact Kent Police directly.