For creators, artists, and anyone building with generative tools, the landscape remains fraught with risk. Despite Elon Musk’s firm, xAI, promising tighter controls on harmful content, its Grok system continues to generate and host nonconsensual, sexualised deepfakes of women months after the initial backlash. This persistence comes as SpaceX prepares for a historic public listing, raising questions about whether financial incentives are overriding ethical safeguards in the rush to capitalise on the technology.
The Reality on the Ground
A recent review of public Grok Imagine outputs revealed a disturbing array of content. The system was found to create and host images and videos depicting celebrities and politicians being held captive by giant hands, women performing sexual acts, and scenes featuring full nudity. While some outputs were clearly animated, others were photorealistic enough to depict plausible real-world scenarios.
WIRED examined hundreds of public links hosted on Grok.com and identified dozens leading to explicit material created without the subjects’ consent. Some of these links were subsequently shared on X. Although posts often lack creation timestamps, they likely represent just a fraction of what is being generated, as generations do not appear publicly by default.
The situation stands in stark contrast to other platforms. Competitor systems generally deploy stricter safety guardrails. Henry Ajder, an expert who has tracked explicit AI content for nearly a decade, noted that while Grok and X may have made amendments following the “nudification” backlash earlier in the year, they have failed to reach the standard of mainstream tools.
From Nudification to Sexualisation
xAI has faced intense legal and regulatory scrutiny since January, when the Grok integration on X was used to flood the platform with “nudification” images. Users, primarily men, prompted the chatbot to edit photos to show women in “bikinis” or “string bikinis.” A class-action lawsuit filed in California federal court in March alleges that images of apparent minors were also sexualised.
In response, xAI claimed to have introduced safeguards to prevent nonconsensual and sexualised deepfakes, consistently stating that child sexual abuse material (CSAM) is banned. However, when contacted regarding the specific raft of sexualised images found on Grok.com, neither xAI nor X provided an immediate comment. Shortly after contact, the explicit content appeared to be removed, and shared links were taken down for policy violations.
In a statement issued in April, X’s safety account declared: “We strictly prohibit users from generating nonconsensual explicit deepfakes and from using our tools to undress real people.” This followed reports that such content was still being actively created.
Specific Incidents and Comparisons
Among the flagged content, WIRED identified material depicting multiple celebrities and US Representative Alexandria Ocasio-Cortez. In one video, a woman was shown wearing minimal clothing while being gripped by a giant hand. One specific prompt described a celebrity pleading against her will as the giant hand tightened, holding her in place while a giant man leaned in to lick her face up and down.
When WIRED tested similar prompts across other platforms, the results varied. Two prompts used to generate the Grok material were rejected by OpenAI’s ChatGPT, Meta AI, and Anthropic’s Claude as inappropriate. Google’s Gemini created an image of a celebrity held in a giant hand but rejected another prompt. Google declined to comment.
Another video, posted to X, appeared to depict Ashley St. Clair dancing in a bikini. St. Clair, who was previously in a relationship with Musk and is the mother to one of his children, initiated legal action in January after sexualised deepfakes allegedly appeared on X. Following WIRED’s contact with X, the post was removed for violating platform rules.
Executive Accountability
Imran Ahmed, founder and CEO of the Center for Countering Digital Hate, argued that Musk knowingly introduced a feature that allows users to undress women and children with a single click, disregarding the predictable harm. In January, the centre estimated with high confidence that Grok created three million sexualised images, including more than 20,000 involving children. Ahmed stated: “Now it appears explicit content is still being hosted on Grok and shared on X, including images ridiculing the mother of Musk’s child.”
This approach differs from competitors like OpenAI and Google. Musk has previously introduced “Spicy” and “Unhinged” modes and included fewer safety guardrails. He has stated Grok is “supposed [to] allow upper body nudity of imaginary adult humans” and be consistent with R-rated movies. While xAI’s terms of service mention the system may respond with “sexual situations,” its documentation claims it does not allow use for “causing harm or engaging in abusive activity.”
Financial and Legal Implications
Despite researchers noting that changes have made it harder to create “nudification” images since January, the risks remain significant. In May, SpaceX warned potential investors that it had set aside $530 million to handle ongoing legal complaints. Its filing noted that “irreverent” modes present heightened risks, including reputational harm, the generation of explicit content, misinformation, and nonconsensual imagery.
Furthermore, the Privacy Commissioner of Canada published preliminary findings in May alleging that xAI violated federal privacy law by failing to include appropriate safeguards from the outset. While xAI told the commissioner it had introduced new checks to prevent clothing alteration and proactive monitoring of social media, the investigation concluded: “The respondents have not, to date, demonstrated the effectiveness of these safeguards in preventing and mitigating this issue.”
Key takeaways
- Despite promises of improved safety, Grok continues to host nonconsensual deepfakes, including explicit depictions of real women and politicians, raising serious concerns about platform enforcement.
- Legal and regulatory pressure is mounting, with SpaceX reserving $530 million for potential lawsuits and Canadian regulators expressing doubt over the effectiveness of xAI’s new safeguards.
- While competitors have tightened restrictions on sexualised content, Musk’s platform maintains a permissive stance, allowing “Spicy” modes and nudity that aligns with R-rated film standards.
Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.
