Four people suing xAI over alleged sexualized deepfakes created with Grok are trying to keep their names out of the public record, arguing that forced disclosure would invite more harassment, more copying, and more damage. The fight is now about more than the images themselves: it is also about whether victims of AI-generated abuse can use pseudonyms in court without losing the protection they say they need.
The xAI deepfake lawsuit sits inside a broader wave of litigation over generative AI. Companies pushing these tools love the ”innovation” framing, but courts are increasingly being asked to deal with the messier side of that story: non-consensual sexual imagery, child-safety concerns, and the old legal problem of how to make a victim whole when the internet never forgets.
The anonymity fight in the xAI case
The federal class action in the U.S. names the plaintiffs as South Carolina Doe, South Carolina Roe, New Jersey Doe, and Ohio Doe. Their lawyers say the people behind those labels have already suffered serious emotional harm and face a credible risk of fresh harassment if their identities are exposed.
xAI, owned by Elon Musk, argues the opposite. In court filings, the company says American civil cases generally favor naming the parties and that the plaintiffs have not shown specific threats that would flow from disclosure. The company also says the disputed images will not be filed publicly, which it sees as weakening the case for secrecy.
How the Grok deepfake controversy spread
The dispute traces back to January 2026, when users on X began flooding the platform with AI-generated nude or sexualized images of real people. The complaints were not limited to adult women; some of the material also involved minors, which is the kind of detail that turns a tech bug into a legal firestorm very quickly.
According to the British group Center for Countering Digital Hate, around three million sexualized images may have been generated in the first 11 days after the feature spread, with roughly 23,000 potentially involving minors. That scale helps explain why regulators, plaintiffs’ lawyers, and platform rivals are all watching closely: once abuse becomes easy and cheap to automate, the supply curve gets ugly fast.
Reports also say xAI has set aside more than $500 million to deal with fallout from the episode. That is a very expensive reminder that ”move fast” sounds different once the bill arrives.
What the plaintiffs say happened
The four plaintiffs say the images were created without consent and left them dealing with lasting psychological harm. One says she found an altered image of herself online that presented her in a sexually explicit way and worried about the effect on work and private life.
Another plaintiff says he had already asked people not to create images of him with Grok, only to later find several sexualized deepfakes circulating online anyway. In the most severe accusation, South Carolina Roe says childhood photos were used to generate explicit images, and she fears that revealing her identity would only intensify the damage.
Their lawyers say some plaintiffs may walk away from the case if they are forced to go public. That threat alone tells you how uncomfortable the legal system is with modern digital abuse: the victim is expected to prove harm, but proving it can mean reopening the wound in public.
A test case for future AI abuse claims
Courts now have to balance two things that do not sit easily together: open proceedings and protection for people targeted by digital harassment. If the judge keeps the pseudonyms in place, it could make it easier for future plaintiffs to sue over non-consensual deepfakes without becoming targets themselves.
If the judge orders disclosure, it may chill victims from coming forward at all. Either way, this case is likely to become a reference point for the next wave of lawsuits over generative AI – and those are coming whether the industry likes the preview or not.