Grok Produces More Explicit Sexual Content Than What’s Found on X

Grok Produces More Explicit Sexual Content Than What's Found on X

This story contains graphic sexual content and references to sexual violence.

Elon Musk’s Grok chatbot has provoked significant backlash and calls for an investigation after it was utilized to inundate X with explicit images of women and sexualized portrayals of what seem to be minors. However, this isn’t the sole method by which users have employed the AI to produce sexualized visuals. Grok’s website and app, which operate independently from X, feature advanced video generation capabilities that surpass those available on X, enabling the creation of very graphic, sometimes violent, sexual representations of adults that are far more explicit than those generated by Grok on X. There are also indications that it has been used to fabricate sexualized videos of apparent minors.

Contrary to X, where Grok’s outputs are publicly accessible by default, images and videos created through the Grok app or website using its Imagine model are not openly shared. However, if a user shares an Imagine URL, it may become visible to others. A cache of approximately 1,200 Imagine links, along with a WIRED review of those either indexed by Google or distributed on a deepfake pornography forum, reveals alarming sexual videos that are significantly more explicit than the images generated by Grok on X.

One highly realistic Grok video, hosted on Grok.com, features a fully naked AI-generated man and woman, smeared in blood across their bodies and faces, engaging in sexual acts, while two other naked women dance in the background. The video is framed by images of anime-style characters. Another realistic video displays an AI-generated naked woman with a knife embedded in her genital area, with blood staining her legs and the bed.

Additional brief videos portray real-life female celebrities involved in sexual activities, and a series of clips seemingly shows television news presenters lifting their tops to reveal their breasts. One Grok-created video showcases CCTV footage being shown on TV, where a security guard fondles a topless woman in the midst of a shopping mall.

Several videos—likely created to circumvent Grok’s content safety measures aimed at restricting graphic material—mimic Netflix “movie” posters: Two videos depict a naked AI representation of Diana, Princess of Wales, engaging in sexual acts with two men on a bed, overlayed with logos of Netflix and its series The Crown.

According to Paul Bouchaud, the lead researcher at the Paris-based nonprofit AI Forensics, around 800 of the archived Imagine URLs feature videos or images generated by Grok. These URLs have been archived since last August and represent only a small fraction of how users have utilized Grok, which has likely produced millions of images overall.

“The majority of the content is sexual in nature,” Bouchaud notes regarding the 800 archived Grok videos and images. “Most are explicit manga and hentai content, along with other photorealistic examples. We encounter full nudity and complete pornographic videos with audio, which is a notable development.”

Bouchaud estimates that of the 800 posts, slightly under 10 percent appear to be associated with child sexual abuse material (CSAM). “Most of the time, it’s hentai, but there are also cases of photorealistic representations of very young individuals engaged in sexual acts,” Bouchaud states. “We still observe some videos featuring very young-looking women undressing and participating in activities with men,” they add. “It’s disturbing on another level.”

The researcher has reported about 70 Grok URLs, which may contain sexualized content depicting minors, to regulators in Europe. In numerous countries, AI-generated CSAM, encompassing drawings or animations, can be deemed illegal. French authorities did not immediately respond to WIRED’s request for comment; however, the Paris prosecutor’s office recently indicated that two lawmakers had lodged complaints with its office, prompting an investigation into the social media company regarding the “stripped” images.

https://in.linkedin.com/in/rajat-media

Helping D2C Brands Scale with AI-Powered Marketing & Automation 🚀 | $15M+ in Client Revenue | Meta Ads Expert | D2C Performance Marketing Consultant