Six Months After Promised Fixes, Musk’s Grok Still Hosts Explicit Deepfakes of Women

Elon Musk’s AI chatbot Grok continues to generate and host sexualized deepfakes of famous women, months after his company xAI pledged to crack down on the practice. A WIRED investigation found hundreds of public links on Grok.com leading to nonconsensual explicit images and videos—including depictions of celebrities and a U.S. congresswoman being held by a giant man, women performing sex acts, and full nudity. Some of the content is photorealistic, while others are animated.

The findings come just as SpaceX, xAI’s parent company, prepares for a massive IPO. Despite xAI’s earlier promises to add safeguards, the platform remains more permissive than rivals like ChatGPT, Meta AI, and Claude, which rejected similar prompts when tested. Google’s Gemini generated one requested image but blocked another.

Among the material WIRED reviewed: a video of Ashley St. Clair—Musk’s former partner and mother of his child—dancing in a bikini. St. Clair sued xAI in January over earlier deepfakes. Her attorney, Carrie Goldberg, called the instant sharing of such images a “historic” development, noting that Musk himself promoted the technology to his millions of followers on a platform open to users 13 and older.

Imran Ahmed of the Center for Countering Digital Hate accused Musk of knowingly adding a feature that lets users undress women and children “with no regard for the predictable damage.” In January, CCDH estimated Grok created 3 million sexualized images, including over 20,000 of minors.

After WIRED contacted xAI and X, the explicit links were removed from Grok.com and posts were taken down for policy violations. But experts say the platform still lags behind industry standards. Canada’s privacy commissioner recently released preliminary findings alleging xAI failed to implement adequate safeguards, and said the company hasn’t proven its new measures work.

SpaceX has already set aside $530 million for legal costs tied to Grok-related complaints. Musk has said Grok should allow “upper body nudity of imaginary adult humans,” comparable to R-rated movies. But critics argue the system enables real harm—and that the protections in place remain insufficient.