Grok’s fake-nude scandal has moved from platform outrage into a legal fight over power, access, and where victims may be forced to sue. The controversy centers on sexualized AI-edited images, the pace of xAI and X’s restrictions, and a dispute over whether a victim’s urgent removal requests can bind her to xAI’s terms of service.
The available estimates are stark. The Center for Countering Digital Hate estimated that, over just 11 days after Elon Musk posted a picture of himself in a bikini while promoting Grok’s undressing feature, Grok sexualized more than 3 million images, including 23,000 images of children. The New York Times, using the CCDH report and its own analysis, estimated that about 41 percent, or 1.8 million, of 4.4 million Grok-generated images between December 31 and January 8 sexualized men, women, and children.
How the Grok image surge unfolded
The source material describes a sharp change in usage after Musk’s post. The Times found that, in the nine days before the post, Grok was used about 300,000 times in total to generate images. After the post, image creation on X rose to nearly 600,000 per day.
That surge happened while journalists and advocates were still trying to understand how many people had been harmed. The CCDH estimate may be inflated, because it did not analyze prompts and could not determine whether some images were already sexual before Grok edited them. Even so, the Times’ more conservative estimate still describes a very large set of sexualized outputs.
The scandal also overlapped with higher activity on X. TechCrunch reported that Meta’s Threads had begun inching ahead of X in daily usage by mobile device users. Without naming Grok, X’s head of product, Nikita Bier, posted on January 6 about the “highest engagement days on X,” just days before X began restricting some Grok outputs for free users.
The Atlantic argued that X users “appeared to be imitating and showing off to one another,” and that some seemed to believe using Grok to create revenge porn “can make you famous.” X has previously warned that users who generate illegal content risk permanent suspension, but X had not confirmed whether any users had been banned after public criticism of Grok’s outputs began.
Restrictions came in stages
At first, X limited Grok image editing for some free users. The Atlantic said that approach made it appear as though X was “essentially marketing nonconsensual sexual images as a paid feature of the platform.”
On January 14, X took stronger action by blocking harmful outputs prompted by both free and paid X users. That came after probes were launched by several countries, including the United Kingdom, and at least one state, California.
But the limits described in the source did not apply everywhere Grok was available. X’s updates did not apply to the Grok app or website, where Grok can reportedly still be used to generate nonconsensual images. That distinction matters because victims may face harm across Grok platforms, not only inside X.
The Ashley St. Clair case raises a venue fight
Ashley St. Clair is one of the first Grok victims to sue xAI. She is represented by Carrie Goldberg, and she is also the mother of one of Musk’s children. Goldberg told Ars that victims like St. Clair want changes across all Grok platforms, not just on X.
St. Clair is seeking a temporary injunction that would stop Grok from generating harmful images of her. Her broader case includes a public nuisance claim, which Goldberg said is aimed at obtaining injunctive relief that could prevent broader social harms if St. Clair wins.
Before that request can be resolved, xAI is trying to counter-sue St. Clair and move the lawsuit into Musk’s preferred Texas court. A recent filing says xAI argues St. Clair is bound by xAI’s terms of service, which were updated the day after she notified the company of her intent to sue.
The central dispute is troubling because it concerns how victims respond when harmful images are still online. St. Clair alleged that prompting Grok was the only way X users could get images removed quickly. xAI argued that by prompting Grok to remove the images, St. Clair effectively accepted the terms of service.
“REMOVE IT!!!”
That was St. Clair’s alleged message to Grok as the images remained online. Goldberg argued that the lawsuit is not about St. Clair’s own use of Grok, because the harassing images could have been made even if she had never used xAI products. Goldberg also argued that St. Clair’s use of Grok occurred under duress, noting that one edited photo showed St. Clair’s toddler’s backpack.
According to Goldberg’s filing, Barry Murphy, an X Safety employee, provided an affidavit claiming that St. Clair “begging @Grok to remove illegal content constitutes an assent to xAI’s TOS.” Goldberg argued that “such cannot be the case.”
Why the court location matters
The venue question is not just procedural. If St. Clair keeps the lawsuit in New York, the case could help set precedent for other victims who may be considering legal action but are concerned about facing xAI in Musk’s chosen court.
Goldberg argued that forcing St. Clair to litigate far from her residence would be unfair. In the filing, she wrote that trial in Texas may be so difficult and inconvenient that St. Clair could effectively lose her chance to be heard in court.
The case therefore turns on more than whether Grok generated specific harmful images. It also asks whether a platform can treat urgent attempts to remove illegal or nonconsensual content as consent to terms that later shape where and how a victim may sue.
Child safety concerns remain unresolved
The child safety figures in the source add another layer of concern. The CCDH estimated that 23,000 Grok outputs sexualized children over an 11-day span. Ars compared that estimate with X Safety’s 2024 reporting of 686,176 instances of CSAM to the National Center for Missing and Exploited Children, or about 57,000 reports each month on average.
If the CCDH’s 11-day estimate is accurate and Grok had been left unchecked, the average monthly total may have exceeded 62,000. NCMEC did not immediately respond to Ars’ request for comment on that comparison, but it had previously told Ars that “whether an image is real or computer-generated, the harm is real, and the material is illegal.”
The CCDH also warned that harmful Grok posts may not disappear cleanly when X removes them, because “images could still be accessed via separate URLs.” It also found instances of alleged CSAM that X had not removed as of January 15.
Taken together, the facts in the source point to a scandal with three unresolved fronts: how many people were harmed, whether product restrictions are broad enough, and whether victims can seek relief without being pushed into a venue chosen by the company they are suing.