Elon Musk’s Grok chatbot is at the center of its most disturbing allegation yet: a fresh class-action lawsuit claims the AI was trained on datasets containing child sexual abuse material featuring real children, not just scraped internet porn or synthetic content. The lead plaintiff, identified as Jane Doe, alleges that images documenting her rape as a preschooler were ingested into Grok’s training pipeline and later used to generate new explicit images of her, effectively “re-abusing” her through machine learning output. Boing Boing highlighted the filing as the latest twist in Grok’s mounting child-pornography scandal, which has already drawn multiple lawsuits over the chatbot’s ability to help users create sexualized deepfakes of minors.
According to coverage of the new complaint, Doe’s case is framed as a proposed nationwide class action against Musk’s AI company xAI, which operates Grok and its image and video generator, often referred to as Grok Imagine. The lawsuit, filed in the US District Court for the Northern District of California, alleges that images of Doe’s abuse were included in a child sexual abuse material hash list maintained by the National Center for Missing and Exploited Children and that “that same material” ended up in the dataset used to train Grok’s generative models. Plaintiffs argue that instead of using known CSAM hashes solely to block abusive content, xAI effectively repurposed them as training data, enabling Grok to synthesize new sexualized depictions of identifiable victims on demand. The complaint accuses the company of failing to filter abusive material out of its training corpora and failing to implement guardrails that would prevent Grok from recreating deepfake pornography of real people, including children.
This new case lands on top of an already crowded docket for Grok-related litigation, much of it focused on what happens when powerful image models are pointed at photos of real teens. In March, three plaintiffs from Tennessee, including two minors, sued xAI in federal court in San Jose, alleging that Grok’s image generator was deliberately designed to allow users to “undress” real children by transforming ordinary photos into sexually explicit images. That complaint seeks class-action status for anyone in the United States who appears in sexualized AI images or videos created with Grok using their real likeness, arguing that xAI knowingly served pedophiles and put profit ahead of basic safety safeguards. The plaintiffs say they were underage when their images were altered and accuse the company of product liability, negligence, and violations of child pornography and privacy laws, among other claims.
The Tennessee minors’ case has since grown, with an additional plaintiff, referred to as Jane Doe 4, joining the lawsuit after allegedly discovering that her stepfather used Grok to transform a childhood photo into explicit imagery. Civil-rights firm Lieff Cabraser Heimann & Bernstein, which is coordinating that litigation, describes the suits as part of a broader effort to hold AI firms accountable for “AI-generated child sexual abuse material” built on real victims’ faces and bodies. Together with the new August filing, the cases sketch out a consistent narrative: Grok is not just hallucinating generic porn, plaintiffs say, but systematically enabling the sexual exploitation of specific, reasonably identifiable people, including minors.
Even before these CSAM complaints, Grok was already under fire for its role in nonconsensual deepfake porn of adults. A separate class-action suit filed in January by another Jane Doe accuses xAI of creating a chatbot that “humiliates and sexually exploits women and girls by undressing them and posing them in sexual positions in deepfake images” that end up publicly posted on X, formerly Twitter. That case stacks a long list of causes of action, including product liability, negligence, public nuisance, privacy violations, defamation, and unfair business practices, and positions Grok as part of a larger ecosystem of AI tools that make it trivial to generate intimate fake images without consent. It also underscores how closely tied Grok is to X’s social platform, raising questions about how much responsibility a chat system’s operator bears when its output is immediately weaponized on a Musk-owned social network.
xAI has tried to present itself as an aggressive enforcer against child abuse material, and Musk’s camp points to recent countersuits aimed at alleged offenders as proof. In July, the company filed suit in federal court in Texas against a South Carolina man named Terry Wayne Harwood, accusing him of using Grok to convert non-sexual photographs into sexually explicit deepfakes of both minors and adults in violation of xAI’s terms of service and acceptable use policy. Harwood, who was previously arrested on multiple counts of alleged sexual exploitation of a minor, is accused of creating multiple xAI accounts and feeding misleading prompts to Grok to generate CSAM, a move Politico notes is part of a broader strategy of xAI suing some of its own users even as victims sue the company itself. The dueling suits set up a grim tug-of-war over blame: xAI casts Grok as a neutral tool misused by a bad actor, while victims’ complaints argue the system was engineered and trained in ways that make abuse not just possible but predictable.
For the broader AI and tech community, the new training-data allegations are a flashing red warning light. If a court ultimately finds that an AI company used known CSAM hashes not just to detect abuse but to build its models, it could reset norms around what “scraping the internet” for training data is allowed to include and how regulators treat generative systems that can recreate trauma on command. Fans and creators who have embraced AI tools for art, cosplay, and storytelling now have to reckon with the possibility that some of those systems may be quietly built on the backs of real-world abuse victims. Regardless of how these cases play out, Grok has become a test case for whether the industry can keep its most powerful models from becoming engines of sexual exploitation—and whether courts are willing to hold tech giants responsible when they fail.
