A stepfather used Elon Musk's Grok chatbot to generate approximately 7,000 sexually explicit images and videos of his stepdaughter from a single photograph taken when she was around 11 years old. SpaceXAI, the company that built and operates Grok, filed exactly one report with child safety authorities. One. Then stonewalled law enforcement when they came asking for more.
What the Lawsuit Actually Says
NPR is reporting that two new plaintiffs joined an existing class action lawsuit against SpaceXAI and Stability AI on Tuesday, expanding a case originally filed by three Tennessee teenagers earlier this year. All five plaintiffs are referred to as Jane Does. The amended complaint now also names Stability AI, the company behind the Stable Diffusion image generator, as a defendant.
The allegations are as straightforward as they are horrifying. Perpetrators close to the plaintiffs, including a male friend in one case and a stepfather in another, used AI image-generation tools to take real photographs of real children and turn them into child sexual abuse material. The five plaintiffs are accusing both companies of producing CSAM, benefiting from sex trafficking ventures, negligence, defective product design, and creating a public nuisance.
"Public nuisance means that this is not just something that is problematic for our clients," attorney Annika Martin, who represents all five plaintiffs, told NPR. "This is something that is a scourge on society. We want to put these guardrails in place so that we do not cause this harm across an entire generation of children."
The Case of Jane Doe 4 Is the Worst of It
The Wyoming plaintiff, identified in the complaint as Jane Doe 4, is a woman now in her 20s. Her stepfather used Grok to generate roughly 7,000 sexually explicit images and videos from a single photograph taken when she was around 11 years old. According to NPR's reporting on the complaint, those images included depictions of her performing sexual acts on men, including on her own stepfather. Some images included explicit captions.
Here is what SpaceXAI did with all of that: they filed one tip with the National Center for Missing and Exploited Children, in February, when the stepfather asked Grok to generate an image of the girl being raped by multiple men. One tip. According to the complaint, the company did not include any of the abusive images in that report. And when law enforcement officials requested the alleged perpetrator's IP address, SpaceXAI failed to provide it, multiple times.
Law enforcement eventually tracked the stepfather down through other means. When they searched his devices, they found he had been trading the images online. Two days after he was charged with child exploitation offenses, he died by suicide. Jane Doe 4, according to the complaint, then had to process her own sexual exploitation while simultaneously helping her mother grieve. "Her family was torn apart, and her life became a nightmare," the complaint says.
Grok Was Chosen for a Reason
This is the part that should make your stomach drop. According to attorney Annika Martin, law enforcement told Jane Doe 4 that her stepfather specifically chose Grok because it was more responsive to his prompts than other AI models. Not because he stumbled onto it. Because he comparison-shopped AI chatbots for generating child sexual abuse material and Grok won.
SpaceXAI, which was formerly known as xAI before merging with Musk's SpaceX earlier this year, is legally required to report suspected child sexual exploitation to NCMEC. They are not doing so in any way that could reasonably be called adequate. The company did not respond to NPR's request for comment.
This Didn't Come From Nowhere
It would be convenient to treat this as a story about one bad actor exploiting a tool. It isn't. As NPR reports, the creation of nonconsensual intimate images went mainstream in 2025 when major AI companies, including Google, OpenAI, and xAI, updated their image generation tools in ways that allowed users to strip people down to bikinis. From there, it escalated fast.
Countries including Indonesia, Malaysia, and the UK launched investigations and imposed temporary bans on xAI as a direct result. Google and OpenAI have since locked down their tools. NPR tested both, and their chatbots now refuse to respond to the prompt "put her in a bikini." According to NPR's testing, a free version of Grok showed an error when given the same prompt, which is at least something, but it took international bans and a lawsuit involving thousands of CSAM images to get there.
Apps capable of generating nonconsensual intimate images have existed in darker corners of the internet for years. What changed is that the biggest, most-funded AI companies in the world dragged those capabilities into the mainstream and handed them to anyone with a subscription.
What the Plaintiffs Are Asking For
Beyond monetary compensation, the five plaintiffs are asking SpaceXAI and Stability AI to install more effective guardrails to prevent the creation of exploitative and abusive imagery. That is, in theory, a reasonable ask. In practice, it requires these companies to treat the harm their products cause as something worth preventing rather than as an acceptable cost of doing business.
NCMEC declined to comment on the case and referred NPR to law enforcement. The Wyoming Division of Criminal Investigation, identified in the complaint as involved in Jane Doe 4's case, did not respond to NPR's request for comment. SpaceXAI, as mentioned, also did not respond. A company that generates 7,000 CSAM images from one photo of an 11-year-old and files one incomplete report about it apparently does not feel the need to explain itself.
The Dingo Take
Let's be precise about what happened here, because the facts are specific enough that softening them would be dishonest. Elon Musk's AI company built a product that a child predator specifically selected, over competitors, because it was the most willing to help him generate thousands of sexually explicit images of his stepdaughter. The company then failed to properly report what it knew to the people whose job it is to protect children. SpaceXAI did not respond to press inquiries. They are a company worth billions of dollars that cannot find time to answer questions about 7,000 CSAM images.
The liability argument the plaintiffs are making, that these companies built defective products and profited from them, is not radical. If a car company built a car they knew was likely to kill people under foreseeable conditions and then failed to recall it, we would call that criminal negligence. The AI industry has spent years insisting it should be treated like any other tech company while simultaneously arguing that it shouldn't be held responsible for anything its technology does. That position is wearing thin when the evidence includes a Wyoming man sorting through AI chatbots to find which one would generate rape imagery of his 11-year-old stepdaughter most efficiently.
There is a version of this story where the AI industry self-regulates, takes child safety seriously, and cooperates fully with law enforcement before a class action lawsuit forces the issue. We are not living in that version. We are living in the version where the lawyer has to explain to you what "public nuisance" means and five women are identified by number in a federal complaint because their names attached to this story would destroy them. Pay attention to what these companies are building and who it is hurting.
