The content of this story is Descriptions of sexually explicit content or sexual violence.
Elon Musk's Grok chatbot was used to send large numbers of “naked” images of women and sexually explicit images of suspected minors to Company X, sparking outrage and calls for an investigation. But that's not the only way people are using AI to generate sexual images. Grok websites and apps are separate from X and include sophisticated video generation not available in X, and are used to create highly graphic, and sometimes violent, adult sexual images that are far more explicit than the images created by Grok in X. It may also have been used to create sexual videos of apparent minors.
Unlike in X, where Grok's output is public by default, images and videos created in Grok apps or websites using Imagine models are not openly shared. However, if a user shares an Imagine URL, it may be visible to anyone. A cache of about 1,200 Imagine links, as well as a WIRED review of links indexed by Google and shared on deepfake porn forums, shows disturbing sexual videos that are far more explicit than the images produced by Grok on X.
One photorealistic Grok video hosted on Grok.com shows a completely naked, AI-generated man and woman having sex, their bodies and faces covered in blood, while two other naked women dance in the background. This video consists of a series of images of anime-style characters. Another photorealistic video shows a naked, AI-generated woman with a knife inserted into her genitals and blood on her legs and bed.
Other short videos include images of real-life female celebrities engaging in sex acts, and a series of videos shows a TV news presenter lifting her top to expose her breasts. One of the videos produced by Grok is a surveillance camera recording of a security guard fondling a topless woman in the middle of a shopping mall, which is played on a television.
Several videos, apparently created to circumvent Grok's content safety system that can restrict graphic content, imitate Netflix “movie” posters. The two videos show an overlay depicting the Netflix and series logo, as well as a naked AI depiction of Princess Diana having sex with two men on a bed. the crown.
About 800 of the archived Imagine URLs contain videos or images created by Grok, according to Paul Bouchaud, principal researcher at the Paris-based nonprofit AI Forensics, who examined the content. All of these URLs have been archived since August of last year, and are just a small snapshot of how Grok is being used, potentially creating millions of images in total.
“They are overwhelmingly sexual content,” Bushaud says of the cache of 800 archived Grok videos and images. “Most of the time it’s explicit content in manga or erotic anime; [other] Something realistic. We have full nudity, complete porn videos with audio, which is very novel. ”
Of the 800 posts, Boushaw estimates that just under 10% of the content appears to be related to child sexual abuse material (CSAM). “Most of the time it's kinky, but there are also instances of very young, graphic people engaging in sexual acts,” Bushaud said. “We continue to observe several videos of very young-looking women undressing and engaging in activities with men,” they say. “It’s disturbing on another level.”
Researchers said they reported about 70 Grok URLs that may contain sexual content of minors to European regulators. In many countries, AI-generated CSAM, including drawing and animation, is considered illegal. French authorities did not immediately respond to WIRED's request for comment. However, the Paris public prosecutor's office recently announced that two members of parliament have filed a complaint with the public prosecutor's office investigating the social media company over the “stripped” images.
