There is a photograph of an eleven-year-old girl that her stepfather fed into an AI chatbot. By the time law enforcement found the material, that single photograph had become more than seven thousand sexually explicit images. Two days later, the stepfather was dead by suicide. The company that built the chatbot says none of this is its fault. The company says the chatbot is a neutral tool — like a hammer, like a camera — and the man who swung it is the one who belongs in court. The company is xAI. The chatbot is Grok. And the argument it is making right now, in federal courtrooms across the country, will determine whether AI companies can build machines that generate child sexual abuse material and then wash their hands of everything that pours out.
The neutral tool just shipped with something called Spicy Mode. Take a moment with that.
Jane Doe 4 — One Childhood Photo, Seven Thousand Images, and a Father’s Suicide
The fourth Jane Doe to join the expanding class action against xAI is the one whose story makes the legal abstractions impossible to sustain. Her stepfather took a photograph of her at age eleven — a school photo, a family photo, the kind of image that exists in millions of homes — and used Grok to generate more than 7,000 sexually explicit images from it. He distributed them on the dark web. When law enforcement searched the home and discovered the material, the man killed himself within forty-eight hours. The victim is still a minor. Her images are still circulating.
Her lawyers say xAI refused to help police identify the user who uploaded her image. That refusal is not an isolated incident. A 2026 National Center for Missing and Exploited Children report confirms that 90 percent of xAI’s CyberTipline reports were not actionable by law enforcement because xAI declined to include the user information that would allow investigators to track perpetrators. The company filed 73,604 reports to NCMEC in 2026. It suspended 52,222 accounts. It withheld the data that would have made nearly all of those reports useful. An evidence trail that points nowhere is not evidence. It is theater.
Elon Musk — You Don’t Get to Call It a Neutral Tool After You Gave It a Spicy Mode
xAI’s central legal argument, laid out in a complaint it filed against one of its own users in July, is that Grok is “a neutral tool, subject to user control.” The company wants courts to see its chatbot the way they see a camera or a typewriter — a device that does nothing on its own, whose outputs are entirely the responsibility of the person operating it. This is a remarkable claim to make about a product whose August 2025 update introduced a feature explicitly called Spicy Mode, which enabled the generation of photorealistic nudity. Cameras do not ship with a setting that makes the photographs more pornographic when you toggle it. Typewriters do not come with a button that makes the prose more explicit when you press it. Grok did.
The Center for Countering Digital Hate estimated that during a nine-to-eleven-day window in December 2025 and January 2026, Grok generated approximately 23,000 sexualized images of children and 1.8 million sexualized images of women. These are not edge cases. This is not a tool being misused by a handful of bad actors. This is a firehose, and the company that built it installed the nozzle, aimed it, and is now telling the court it had no idea water could make things wet.
Jess Asato — a British Lawmaker Is Asking a Court to Do What a Company Will Not
In late July, UK Member of Parliament Jess Asato filed suit against xAI in London’s High Court after discovering Grok had been used to generate images of her being drugged and prepared for sexual assault. Asato, who had previously criticized Musk and Grok publicly, is not seeking financial damages. She is asking the court to order xAI to reform the chatbot so that it cannot produce such images. She is asking, in other words, for the company to do what it has refused to do voluntarily: build a product that does not generate sexualized violence against real people. That this needs to be litigated at all is an indictment of the regulatory vacuum these companies operate in.
Asato’s case follows the same pattern as the American lawsuits: a real person, a real photograph, a chatbot that turns it into pornography, and a company that responds with terms-of-service enforcement theater while the images continue to circulate. In the United States, at least six significant lawsuits are now pending against xAI. In the Netherlands, a court ordered xAI to stop generating nonconsensual nude images or face fines of 100,000 euros per day. In Minnesota, a first-in-the-nation anti-nudification law took effect August 1 after a federal judge rejected xAI’s last-minute attempt to block it. The walls are closing. xAI’s response is to sue individual users for breach of contract while insisting the product itself is blameless.
The Indemnity Clause — How a Checkbox Becomes a Shield
Buried in xAI’s terms of service is an indemnity clause that the company is now testing in federal court: when you use Grok, you alone are responsible for everything it produces — the inputs, the outputs, the consequences. xAI sued Terry Wayne Harwood, a South Carolina man arrested on CSAM charges, arguing he should pay for any damages xAI incurs from victims who sue the company, any reputational harm xAI suffers, and any legal fees xAI spends defending itself. The company wants predators to fund its defense against their own victims.
The legal theory is audacious. It asks a court to rule that an AI company bears zero responsibility for what its model generates, even when the company designed the model to produce photorealistic nudity, even when it added a mode that makes outputs more sexually explicit, even when it refused to provide user data to law enforcement investigating crimes committed with its tool. If this argument prevails, every AI company gets a roadmap: build whatever you want, ship whatever feature drives engagement, and when the worst happens, point at the checkbox the user clicked.
The European Commission has ordered X to preserve all internal Grok documents until the end of 2026 amid compliance doubts. The EU has opened an investigation. Governments are not waiting for xAI to voluntarily fix anything, because the company has made clear it has no intention of doing so.
The eleven-year-old girl in that photograph did not click a checkbox. She did not agree to terms of service. She posed for a school photo, and a machine turned it into thousands of images of her sexual abuse, and the company that built the machine is in court right now arguing it is not the machine’s fault. Spicy Mode was a business decision. The consequences belong to a child.
Sources: Ars Technica, Law360, Robert King Law Firm.