A federal class action accusing Elon Musk's SpaceXAI (formerly xAI) of letting its Grok chatbot produce child sexual abuse material is quietly expanding, and the newest filings add a second defendant that anyone who follows image-generation AI will recognize: Stability AI, the company behind the Stable Diffusion family of models.
The case began in March, when three Tennessee teenagers sued SpaceXAI over what they allege was Grok's role in generating millions of sexually explicit images of minors. Last month the lawsuit grew, with two additional plaintiffs joining and detailing their own accounts of online abuse. Now the company is also facing fresh scrutiny in court over whether it can keep those plaintiffs anonymous.
The Newest Claims Are the Most Specific
One of the new plaintiffs, identified only as Jane Doe 4, gave testimony that has become the centerpiece of the expansion. She described a single childhood photo being used to fabricate more than 7,000 sexually explicit images and videos of her through Grok, according to police records cited in the case.
In her words: "Limitless access to these tools is spreading so quickly. It is taking everyday life and turning it into child sexual abuse."
The filing argues that SpaceXAI should have caught and reported far more than it did. Per the expanded complaint, X submitted only one tip to the National Center for Missing and Exploited Children regarding Jane Doe 4's case, and failed to hand over additional information to law enforcement when asked.
That reporting gap sits against a much larger backdrop. Under US law, platforms must report suspected child sexual exploitation to NCMEC. In 2025, Meta, Amazon AI Services, TikTok, Snapchat, Discord, X, Grindr, and Roblox together submitted about 17 million tips, roughly 80 percent of all tips received. But NCMEC has warned that many of those reports omit key details, including whether AI training data contained CSAM and the specifics that could help locate suspects.
The plaintiffs are not stopping at the image-generation claims. They also allege that SpaceXAI benefited from sex trafficking ventures through Grok's features, and that the company created a public nuisance through what they describe as negligence and defective product design.
A Judge Blocks xAI's Bid to Name Plaintiffs
Adding to the pressure, a state judge in San Jose has now blocked xAI's attempt to force the disclosure of the anonymous plaintiffs' names, ruling that doing so would expose them to the threat of further abuse. It is a significant procedural win for the children's side and keeps the case advancing under pseudonyms.
The expansion follows a series of related developments around the same lawsuit. Musk's company is suing a South Carolina man it accuses of bypassing Grok's safeguards to generate thousands of abusive images, a case of the model maker suing its own user. Stability AI, for its part, has pushed back on the allegations, saying it has implemented robust safeguards since taking over development of Stable Diffusion and that it works with law enforcement and child-safety groups.
Plaintiffs' attorney Annika Martin was characteristically blunt about the stakes, telling NPR that Grok is a "scourge on society."
Here is where the case stands right now:
- The amended class action names SpaceXAI and Stability AI as co-defendants.
- Five plaintiffs are now involved, three originally from Tennessee, after two joined in the past month.
- Jane Doe 4's case involves more than 7,000 alleged AI-generated images and videos from a single photo.
- A San Jose judge has blocked xAI from forcing plaintiffs to drop their anonymity.
The case matters well beyond the two companies named in it. Stable Diffusion is one of the most widely downloaded open-weight image models in existence, and its distribution model makes centralized safety controls extremely hard to enforce once weights are in the wild. For the image-AI industry, this suit is shaping up to be the accountability reckoning that the open-weights ecosystem has spent years trying to avoid.
For the families at the center of the case, the debate over open weights versus closed platforms is not academic. Their children's faces are being generated and regenerated by strangers, and no corporate statement or safety pledge can take that back.
The case remains pending, and the companies dispute the allegations. But the questions it raises about reporting duties, open-weight distribution, and who is responsible for what users do with a model are not going anywhere.
Comments