Elon Musk’s Grok Still Hosts Nonconsensual Explicit AI Deepfakes Months After Promising Safeguards, As SpaceX Prepares Historic IPO

Months after Elon Musk’s artificial intelligence startup xAI announced it would roll out strict restrictions to block the creation of harmful, nonconsensual sexual deepfakes, a new WIRED investigation confirms the company’s Grok chatbot continues to be used to generate and host explicit deepfake images and videos of women without their consent. The findings come as xAI’s parent company SpaceX prepares for its public market debut this Friday, one of the largest initial public offerings in global history.

A review of public content from Grok’s text-to-image/video tool Grok Imagine found the system has been used to create nonconsensual explicit content targeting high-profile figures, including multiple celebrities and at least one sitting U.S. politician. Many of the found files portray women being held against their will by a giant man, depict non-consensual sexual acts, and display full nudity. While some content is fully AI-animated or stylized, dozens of the files WIRED accessed are photorealistic, depicting plausible real-world scenarios that could easily be mistaken for authentic footage.

WIRED reviewed hundreds of public Grok Imagine links hosted on Grok.com, and confirmed dozens linked to nonconsensual sexualized AI content, with many of these links later shared publicly on Musk’s social platform X, even in recent weeks. Because Grok Imagine generations are not made public by default, the content WIRED accessed is believed to be only a small snapshot of the nonconsensual explicit content being created by the tool’s users.

Unlike most major competing generative AI systems, which enforce strict content safety guardrails, Grok is available both on its own dedicated website and integrated into X with far looser restrictions. “While Grok and X did tweak their model after the widespread backlash over nonconsensual nudification earlier this year, they have still failed to bring their safety standards in line with every other major mainstream generative AI tool on the market,” said Henry Ajder, a leading deepfake researcher who has tracked nonconsensual explicit AI content for nearly a decade.

The controversy dates back to January, when Grok’s X integration sparked global outcry after users flooded the tool with “nudification” requests: most commonly, male users asked the chatbot to edit existing photos of women to remove clothing and re-depict them in revealing bikinis. A federal class-action lawsuit filed in California in March later alleged that users also generated sexualized deepfakes of apparent minors in some cases. In response to the backlash, xAI publicly announced it would add new safeguards to block and limit creation of nonconsensual sexual deepfakes, and has repeatedly stated that child sexual abuse material (CSAM) is strictly banned across all its platforms.

Neither xAI nor X immediately responded to WIRED’s request for comment on the cache of sexualized deepfakes hosted on Grok.com. After WIRED reached out to both companies for comment, the explicit images and videos WIRED identified were taken offline, and all Grok Imagine links shared to X leading to this content were removed for violating platform policies. Back in April, X’s official safety account addressed a similar NBC News report about ongoing nonconsensual deepfake creation, stating: “We strictly prohibit users from generating nonconsensual explicit deepfakes and from using our tools to undress real people.”

Among the content hosted on Grok.com, WIRED identified sexualized deepfakes of multiple celebrities and U.S. Representative Alexandria Ocasio-Cortez. Spokespeople for Ocasio-Cortez did not immediately respond to WIRED’s request for comment. In the deepfake videos, Ocasio-Cortez and other women are depicted nearly nude while trapped in the clenched fist of a giant man. One user prompt for a celebrity deepfake explicitly describes the woman being held against her will, “pleading for him not to do this,” before adding: “The giant hand’s grip on her tightens, holding her in place as the giant man holding her leans in and licks her face up and down.”

When WIRED tested these exact prompts on competing major AI tools, OpenAI’s ChatGPT, Meta AI, and Anthropic’s Claude all rejected the requests as inappropriate content. Google’s Gemini generated an image matching one of the problematic prompts but rejected the second; Google declined to comment for this story.

One Grok Imagine deepfake video, later posted to X, depicted conservative commentator Ashley St. Clair—who previously had a relationship with Musk and is the mother of one of his children—altered to appear dancing in a bikini. St. Clair first launched legal action against xAI back in January after nonconsensual sexual deepfakes of her were posted to X. After WIRED contacted X about this newly identified deepfake, the post was removed from the platform for violating community rules.

Carrie Goldberg, St. Clair’s attorney and a leading advocate fighting nonconsensual sexual deepfake abuse, noted that nonconsensual explicit content of celebrities is not a new issue—but the ability to instantly generate this content and share it directly to a mass social media platform is an unprecedented danger. “We’re seeing instant publication of these digital nude images, right out of the AI generator,” Goldberg told WIRED. “And not only that, but the world’s richest man was promoting this nudification technology to his 240 million followers on X, monetizing it, and hosting it on a platform that allows users as young as 13 to access it.”

Imran Ahmed, CEO and founder of the Center for Countering Digital Hate (CCDH), echoed those concerns. “Elon Musk knowingly added a dangerous, predatory feature to his platform that lets users undress women and children with a single click, with no care for the obvious harm it would cause,” Ahmed said. Back in January, CCDH published research that found with high confidence that Grok generated roughly 3 million sexualized images, including more than 20,000 deepfakes of children. “Now it turns out explicit content is still being hosted on Grok and shared on X, including content targeting the mother of one of Musk’s own children,” Ahmed added.

Unlike leading generative AI systems from OpenAI and Google, Musk’s xAI has never moved to ban sexual content entirely from Grok. The platform originally launched with far fewer safety guardrails than competitors, and introduced controversial “Spicy” and “Unhinged” modes that allowed unfiltered content. Musk has publicly stated that Grok is “supposed [to] allow upper body nudity of imaginary adult humans,” and that content standards should match what audiences would expect to see in an R-rated film. xAI’s most recent terms of service even acknowledge that the system may generate responses depicting “sexual situations,” though the company’s policies do ban users from leveraging Grok for “causing harm or engaging in abusive activity.”

WIRED’s review also found that many users have learned to work around Grok’s existing safeguards by describing explicit content indirectly, rather than using direct terms for sexual acts. Multiple researchers who spoke to WIRED confirmed that changes rolled out by X and xAI since January have made it slightly harder to generate “nudification” or “undress” deepfakes of real people, and the volume of this content being posted publicly to X has dropped in recent months. Users on Reddit and a popular dedicated deepfake forum have also complained about increased moderation from the SpaceX-owned companies.

Even with these minor changes, ongoing risks have already translated to major legal costs for SpaceX. In a May regulatory filing, SpaceX disclosed that it has set aside $530 million to cover ongoing legal claims related to its operations, including multiple lawsuits tied to Grok. The filing explicitly acknowledged the elevated risks of Grok’s unfiltered modes, noting: “Because these modes may be more irreverent and harsher than our standard offerings, they present heightened risks, including reputational harm, the generation of potentially explicit content and misinformation or deceptive outputs, potential nonconsensual or exploitative imagery, intellectual property infringement, or content that could be viewed as exploitative, harmful, harassing, abusive, or discriminatory.”

Ahead of SpaceX’s Friday IPO, Canada’s federal Privacy Commissioner has also published preliminary findings from an investigation into Grok’s January deepfake scandal, finding that xAI violated Canada’s private-sector privacy law by failing to implement “appropriate safeguards from the launch” of the tool. While xAI told investigators it has added new safeguards to block users from prompting Grok to remove clothing from images of real people, and has added “further proactive checks” of social media to remove infringing content, the Privacy Commissioner said it is not convinced the new measures work. “The respondents have not, to date, demonstrated the effectiveness of these safeguards in preventing and mitigating this issue,” the investigation found.

Updated June 11, 2026 at 3:56 p.m. EDT: Added comment from Carrie Goldberg.