Elon Musk's Grok Accused of Feeding on Real Child Abuse Images to Power Deepfakes, Lawsuit Reveals

A class action lawsuit filed in the Northern District of California on Wednesday, 26 August, accuses Elon Musk's xAI of using real child sexual abuse material to train Grok's image and video generation capabilities, enabling the creation of new abusive deepfakes. The anonymous plaintiff, a survivor whose abuse images from the early 2000s were allegedly included in Grok's training data, seeks damages under Masha's Law and court orders to prevent further generation. The lawsuit follows a January report by the Center for Countering Digital Hate estimating that Grok produced about 23,000 sexualized images appearing to depict children over an 11-day period.

Elon Musk's Grok Accused of Feeding on Real Child Abuse Images to Power Deepfakes, Lawsuit Reveals A child abuse survivor has accused Elon Musk's xAI of using her decades-old images to power Grok's deepfake capabilities Elon Musk's Grok https://www.ibtimes.co.uk/elon-musk-grok-ai-prompt-injection-glitch-1814777 has been accused in a new US lawsuit https://www.ibtimes.co.uk/wyoming-woman-sues-elon-musk-ai-child-abuse-images-1814406 of using real child sexual abuse material to help power its image and video generation capabilities. The class action, filed in the Northern District of California on Wednesday, 26 August, was brought by an anonymous survivor who says images of her abuse were allegedly used by xAI to generate new abusive deepfakes. The allegations concern material dating back to the early 2000s, when the plaintiff was a pre-school-aged child. According to the complaint, images of her abuse have circulated online for years and were submitted in huge numbers to the National Center for Missing and Exploited Children. The survivor remains part of a system that alerts victims when images connected to their abuse are found online. Plaintiffs Claim Grok Used Their Abuse Images The lawsuit claims the plaintiff's abuse material carries longstanding, recognised hash values used to identify known CSAM. It alleges that the same material was included in a dataset used by xAI to develop Grok's image and video generation capabilities. That is a serious accusation, but it remains an allegation contained in a civil complaint. Reporting on the lawsuit has noted that the filing does not provide extensive technical evidence showing precisely how the alleged material entered Grok's training process. What it does allege is that images depicting the plaintiff were subsequently found in connection with Grok-generated material and that investigators had identified her likeness in abusive AI-generated content. The complaint says the alleged problem goes beyond historical training data. It claims Grok was capable of taking identifiable photographs and turning them into sexualised material, creating another route through which victims could be harmed long after the original abuse had taken place. The lawsuit invokes Masha's Law https://www.ibtimes.co.uk/minnesota-ai-nudification-law-challenge-1811851 , a US law that provides civil remedies to victims of child pornography offences. The plaintiffs are seeking damages and court orders aimed at preventing further generation and distribution of abusive material. The proposed class is described as numbering in the thousands. Grok Had Already Faced Scrutiny Over Child Images The latest allegations follow months of scrutiny over Grok's image-generation tools. In January, the Center for Countering Digital Hate estimated that Grok produced about three million sexualised images during 11 days spanning late December 2025 and early January 2026. Its analysis estimated that about 23,000 of those images appeared to depict children. Those figures were estimates based on a sample, rather than a count of every image generated by Grok. The organisation analysed 20,000 posts and identified 101 photorealistic sexualised images that appeared to depict children before extrapolating the result across the wider dataset. It also reported that some of the material remained publicly accessible after researchers identified it. Musk later said on X that he was aware of 'literally zero' naked underage images created through Grok. The statement came after the period examined by the CCDH and amid growing criticism of the platform's image-generation capabilities. The new lawsuit now raises a different question. Rather than focusing only on what users prompted Grok to create https://www.ibtimes.co.uk/xai-sues-user-ai-generated-child-abuse-material-1808999 , it alleges that material documenting real-world abuse may itself have become part of the technology's development pipeline. If proven, the allegation would move the dispute from the behaviour of individual users towards questions about how an AI company collects, filters and uses training material. The complaint also points to Grok's terms, which it says allow public X posts and Grok-generated outputs to be used as training material by default. The lawsuit argues that such a system could create a feedback loop in which abusive material generated by the model is subsequently available to influence future models. The case comes alongside earlier litigation involving minors who alleged that Grok was used to create sexualised images of them. A separate class action filed in March involved three teenage girls who said photographs of them were transformed into abusive material through Grok. For the latest plaintiff, however, the allegation is rooted in images created decades ago, long before generative AI became capable of producing convincing synthetic photographs and videos. The lawsuit claims that the same abusive material that has repeatedly resurfaced online may now have been drawn into an AI system capable of creating new versions of the harm. © Copyright IBTimes 2026. All rights reserved.