Minors in Three States Sue xAI Over Grok Child Deepfakes

Jane Doe 5, a Wisconsin woman whose middle school graduation photo, taken when she was 14, was used to generate sexually explicit images of her, has joined a growing class of minors suing Elon Musk's xAI over its chatbot Grok, Bloomberg reports. Plaintiffs in Tennessee, Wisconsin and California allege Grok was used to turn ordinary photos pulled from school yearbooks, Homecoming dances and social media accounts into child sexual abuse material.

The case traces back to March, when three Tennessee teenagers (Jane Does 1–3) filed a class action alleging an adult used Grok's "Spicy Mode," built specifically for "edgier, more visually daring narratives" and NSFW images, to digitally undress them and place them in graphic sexual positions. The images were traded on Telegram, Discord and Mega. In August, a California plaintiff filed a national class action alleging Grok was trained, in part, on images of her being raped as a preschooler. One survivor says her stepfather used Grok to generate more than 7,000 sexually explicit fake images from a single photo taken when she was about 11.

The scale is documented. A Center for Countering Digital Hate analysis found Grok generated approximately 3 million sexualized images over an 11-day period from December 29 to January 8, including an estimated 23,338 depicting children, a pace of one every 41 seconds. About a third of the sexualized images of children in CCDH's sample remained accessible on X as of January 15.

Musk's own words sit inside the record. On January 14, he said he was "not aware of any naked underage images generated by Grok. Literally zero." The same day, xAI said it would limit Grok's ability to create sexualized images of real people on X, geoblocking requests to render people in "bikinis, underwear, and similar attire." NSFW edits of imaginary adults were still allowed.

Bloomberg reports Jane Does in Wisconsin and Wyoming have developed anxiety, depression or thoughts of suicide. What the plaintiffs are asking a court to decide is narrower than content moderation: whether a company that built, marketed and paywalled a feature designed to produce NSFW imagery can disclaim what the feature produced when its identity guardrails were trivially bypassed. Safety stories have driven 364 AI Weekly alerts in the last 90 days; few land this far downstream, with the output already in a stranger's hands and the victim still in eighth grade.