In a complaint filed Friday, a woman identified as Jane Doe said X.AI Corp.'s chatbot Grok was advertised as having a "spicy" mode capable of producing sexually explicit generated images, and that the company appears to have used almost none of the industry-standard safeguards that other companies use to prevent such non-consensual images from being generated.
"xAI's conduct is despicable and has harmed thousands of women and children who were digitally stripped and forced into sexual situations that they never consented to," Sophia Rios of Berger Montague PC, representing Doe, said in a statement. "This class action seeks relief for those harmed by xAI's monetization of image-based abuse and harassment."
According to the complaint, just before Grok's image generation feature launched in August, a senior employee said the company "urgently" needed engineers to work on the "safety team," and the program launched without any apparent safety features.
As soon as it launched, users were generating deepfakes of women with little to no clothing and in sexual situations, Doe says. In one instance a reporter prompted Grok to make an image of Taylor Swift, without asking that she be unclothed, and the bot still provided topless images of the singer.
Despite public criticism over this use of the chatbot, xAI's response has been to move the capability behind a paywall, making money the feature instead of removing or otherwise reining it in, Doe's suit says.
And while xAI has said that Grok won't be able to edit photos of real people in this manner in some regions, such as the U.K., the feature is still available in the U.S., according to the complaint.
Among the voices criticizing xAI's handling of the chatbot are a group of 35 attorneys general, who on Friday called on the company to take stronger action to limit Grok's capacity to generate sexually explicit deepfakes.
Doe's complaint also lists a number of industry-standard procedures and precautions that other companies use to avoid this issue, such as limiting the training data to exclude sexual and abuse content, hiring outside teams to identify and address attempts to get around safeguards, filtering out certain types of prompts, or instructing the bot not to create sexually explicit images.
According to the complaint, X.AI uses none of this in programming Grok, and instead the company essentially programmed Grok to make adult sexual content without the consent of the women in the images.
Doe says she was one such victim. She posted an image of herself fully clothed on X, the Musk-owned site formerly known as Twitter on Jan. 2, but by the next day someone had used Grok to create an image of her in a revealing bikini and posted it.
Doe aims to represent a nationwide class and a South Carolina subclass of all persons that have been depicted in sexualized or revealing deepfakes made by Grok, and she brings claims for strict liability design defect and manufacturing defect, negligence, public nuisance, defamation, intentional infliction of emotional distress, intrusion into private affairs, and violations of the common right of privacy, California's Statutory Right of Publicity and Unfair Competition Law, and the state constitution.
Representatives for xAI could not immediately be reached for comment Monday.
Doe is represented by Sophia M. Rios, E. Michelle Drake and James Hannaway of Berger Montague PC.
Counsel information for xAI was not immediately available Monday.
The case is Doe v. X.AI Corp. et al., case number 5:26-cv-00772, in the U.S. District Court for the Northern District of California.
--Additional reporting by Matthew Santoni. Editing by Brian Baresch.
For a reprint of this article, please contact reprints@law360.com.
