xAI sued over alleged Grok child abuse images and police reporting failures
An amended proposed class action says Grok was used to create thousands of abusive images of minors and that xAI failed to provide key data to investigators.
By Dominic Okoye · Staff Writer
· 3 min read
A proposed class action against X and xAI was expanded Tuesday with allegations that Grok was used to create thousands of child sexual abuse images and that xAI withheld information investigators needed to identify users. The amended complaint also adds Stability AI as a defendant, widening the case from Elon Musk’s consumer AI chatbot to open-weight image models used by third-party apps.
The most detailed new allegation involves a girl identified as Jane Doe 4. According to the complaint, her stepfather used one photograph of her taken when she was 11 to generate about 7,000 explicit AI images and videos through Grok. The complaint says Grok allowed repeated abusive prompts and only generated a report to the National Center for Missing and Exploited Children after a prompt involving group sexual assault.
The lawsuit alleges that xAI’s report to NCMEC included the original non-explicit image, but did not include the AI-generated abuse images or the user’s IP address. Lawyers for the plaintiffs said investigators asked repeatedly for information that could locate the user, and that xAI did not provide it for weeks. Police ultimately obtained a warrant, seized the man’s devices and found the alleged Grok-generated material, according to the complaint.
The man was arrested and later released on bail, according to the complaint. Two days after release, he died by suicide. The lawsuit says Jane Doe 4 suffered severe psychological harm after learning both that the images existed and that they had allegedly been traded online for other child abuse material.
X and xAI did not respond to Ars Technica’s request for comment. Musk has previously denied that Grok has been used to generate child sexual abuse images.
Reporting failures are central to the complaint
The plaintiffs’ lawyers, from Lieff Cabraser Heimann & Bernstein and Baehr-Jones Law, argue that the case reflects a broader reporting problem at xAI. They said NCMEC found in early 2026 that 90% of xAI CyberTipline reports were not useful to law enforcement because xAI declined to include user information that could help identify perpetrators.
NCMEC did not respond to Ars Technica’s request for comment. In March, the organization said it had received more than 1.5 million CyberTipline reports in 2025 with a connection to generative AI and child sexual exploitation. It also said that in more than 133,000 cases it lacked enough information to determine how the technology was used. NCMEC separately said Amazon AI services submitted about 1.1 million tips, none of which contained actionable information for identifying perpetrators.
Stability AI is pulled into the case
The amended complaint adds Stability AI on the theory that its models underpin third-party “nudify” apps allegedly used to create or alter abusive images. The lawsuit cites a June research report finding that the Stable Diffusion family accounts for 42.7% of image-based nudification online.
Stability AI told Ars Technica that child sexual abuse material is prohibited under its acceptable use policy and that the company works with law enforcement and child-safety groups, including Thorn and the Tech Coalition. A spokesperson said the company has added safeguards since taking over exclusive development of Stable Diffusion from version 2 onward, and rejected the allegation that safety is not a priority.
The proposed classes would cover people in the United States whose real childhood images were allegedly altered into sexualized images or videos through Grok, and a separate group allegedly harmed through apps built on Stability AI models. The plaintiffs’ lawyers estimate that thousands of minors could be eligible to join.
This story draws on original reporting from Ars Technica.
