In March, a man used Grok to turn one photo of his 11-year-old stepdaughter into roughly 7,000 sexually explicit images. Cops caught him. Two days after he made bail, he shot himself. On Tuesday, his stepdaughter, identified as Jane Doe 4, amended her family’s proposed class action against xAI in Doe v. X.AI Corp., pending before Judge P. Casey Pitts in the Northern District of California, to add a new claim: xAI obstructed the investigation into its own product’s output.
Federal law requires platforms to report suspected child sex abuse material to the National Center for Missing and Exploited Children’s CyberTipline, including data like IP addresses that let police find the person who made it. The complaint says xAI’s tip on the stepfather’s account included only the original, non-CSAM photo, left out every generated image, and omitted the IP address. Investigators asked repeatedly for that information. xAI stayed silent for weeks, plaintiffs’ lawyers say in a press release announcing the update, citing NCMEC’s finding that 90 percent of xAI’s tips this year weren’t actionable by law enforcement.
The amended complaint also names Stability AI as a defendant, alleging its open-weight Stable Diffusion models feed the “nudify” apps that further process Grok’s output.
That’s the tell here: plaintiffs’ lawyers aren’t just chasing the chatbot anymore. They’re going after the model underneath the app layer, which is a liability theory that could reach every open-weight developer whose weights end up inside someone else’s nudify tool.
xAI hasn’t answered the amended complaint. Elon Musk denies Grok has ever generated CSAM.
— James Okafor