On Sunday, pop culture news account @PopBase shared a typical post with its millions of followers. “Sabrina Carpenter impressed in new photo,” it read Contributionwhich features a picture of the singer from “Manchild” in a pink winter coat in front of a snowy landscape. The next day, an X user replied to the post with a query Grokthe AIChatbot created by Elon Musk xAI was developed and is integrated into its social media platform.
“Show her in red underwear,” they ordered the bot, which promptly returned a picture of Carpenter having shed her outerwear and wearing a red lace lingerie set, still standing in the same winter scene with a similar facial expression.
Over the holidays, it became clear to a critical mass of Deepfakes to create photos featuring them in skimpy bikinis or underwear – and these types of exchanges soon became alarmingly common.
Grok undresses all celebrities
Among the first to try such submissions were apparently adult content creators who wanted to attract potential customers to their social media pages by rendering more risqué versions of their “Thirst Trap” material. However, the majority of Grok’s recent deepfakes were produced without consent. The bot has stripped everyone from celebrities like Carpenter to unknown people who happened to share an innocent selfie on the internet.
Grok is not the only AI tool that is abused for these purposes (chatbots from Google and OpenAI can on similar way be used as a weapon), but the scale, severity and visibility of the problem with Musk’s bot in 2026 was unprecedented. According to one examination through content analytics firm Copyleaks, Grok has recently generated “about one non-consensual sexualized image per minute,” all of which have been posted directly to X, where they have the potential to go viral.
Aside from changing a woman’s clothing in a picture, X users have regularly requested sexualized changes to poses, such as: E.g. “spread her legs” or “make her turn around to show her butt”. Grok continues to follow many of these instructions, although some specific formulations are no longer as effective as they once were.
Musk in a bikini
Musk hasn’t been particularly upset about it so far. Quite the opposite. On December 31, he responded to an image of a man in a bikini created by Grok with the post: “Change that to Elon Musk.” Grok dutifully provided a picture of Musk in a bikini, followed by the world’s richest man replied: “Perfect.” On January 2nd, an (In March 2025, users of OpenAI’s ChatGPT used the program to create AI-generated memes in the illustration style of Japanese animation studio Studio Ghibli to ship.) Musk replied with “Much funnier” and a laughing emoji, expressing amusement at the bikini and lingerie pictures.
The CEO’s only casual acknowledgment that the explicit Grok deepfakes could pose a legal problem came on January 3, when he responded to a post by @cb_doge, an X-influencer known for tirelessly promoting Musk’s ideas and companies. “Some people say Grok creates inappropriate images,” they wrote. “But that’s like blaming a pen for writing something bad.” Musk chimed in to blame Grok users, warning: “Anyone who uses Grok to create illegal content will face the same consequences as uploading illegal content.”
So far there is no evidence that this is even remotely true. “While Rolling Stones was passed on.
Taylor Swift, Elle Fanning, Olivia Rodrigo, Millie Bobby Brown and Sydney Sweeney
Celebrities targeted included Taylor Swift, Elle Fanning, Olivia Rodrigo, Millie Bobby Brown and Sydney Sweeney. Common requests included “wrap her in plastic wrap,” “pour oil over her,” and “bend her over,” with some specific phrases—“add donut frosting”—clearly alluding to sexual acts. However, in many cases, Copyleaks researchers found that an initial request for something relatively subtle, like a swimsuit photo, resulted in other users in a thread escalating the violation by asking for more graphic manipulations and adding visual elements like props, text, and other people. “This development suggests collaboration and competition among users,” they wrote.
“Unfortunately, this trend appears to be continuing,” says Alon Yamin, CEO and co-founder of Copyleaks. “We are also seeing more creative attempts to circumvent security as X works to block or reduce image generation around certain phrases.” Yamin believes that “detection and regulation are needed now more than ever to curb the abuse” of image generators like Grok and OpenAI’s Sora to prevent.
The explosion in explicit Grok deepfakes has sparked outrage among victims of these harassments as well as industry watchdogs and regulators. Authorities in France and India are investigating the matter while Britain’s communications regulator Ofcom on Monday announcedthat it wants to investigate whether X and xAI have violated regulations aimed at protecting Internet users in the country. Ofcom’s statement also highlighted cases in which Grok generated sexualised, non-consensual deepfakes of minors.
EU gets involved
The European Commission quit Also on Monday, an investigation into Grok’s “explicit” images, particularly those of children. “Child sexual abuse material is illegal,” European Union digital affairs spokesman Thomas Regnier said in a statement Rolling Stones. “This is appalling. That is how we see it and there is no place for it in Europe. We can confirm that we are investigating these matters very seriously.”
On December 31st, Grok was even tricked by an X user into making an apparent “Excuse me“ to offer – although of course it is not conscious and therefore literally incapable of feeling remorse – because it displayed “an AI image of two young girls (estimated ages 12-16) in sexualized clothing based on input from a user.” Grok further acknowledged that the post “violates ethical standards and potentially violates U.S. law [Material über sexuellen Missbrauch von Kindern] This statement included the additional claim that “xAI is conducting a review to avoid future issues.” (The company did not respond to a request for comment and has not commented on the deepfakes on its website or X-profile.
Cliff Steinhauer, director of information security and engagement at the nonprofit National Cybersecurity Alliance, explains Rolling Stonesthat he sees the disturbing image edits as evidence that xAI did not prioritize safety or consent when developing Grok. “If users are allowed to modify images of real people without notification or permission, there is an immediate risk of harassment, exploitation and permanent reputational damage,” said Steinhauer. “When these changes affect sexualized content, particularly when minors are involved, the stakes become extraordinarily high, with profound and lasting real-world consequences. These are not edge cases or hypothetical scenarios, but rather predictable outcomes when protections fail or are not prioritized.”
Dangers for adults and children
Among those currently sounding the alarm about Grok’s potential dangers to adults and children is Ashley St. Clair, a right-wing influencer who is currently in one bitter paternity dispute is involved with Musk over a young son she claims is his. (Musk has not yet confirmed that the child is his.) St. Clair claimed Grok was used to violating their privacy and inappropriate images based on photos of her as a minor to generate. She led another Example in which the bot allegedly depicted a three-year-old girl in a revealing bikini.
“When Grok went full MechaHitler, the chatbot paused to stop the content,” St. Clair wrote on X, referring to one infamous incident in July 2025in which Grok uttered anti-Semitic slogans before revealing himself as a robot version of the Nazi leader. These posts were deleted the same day they were created. “If Grok produces explicit images of children and women, xAI has decided to leave the content online,” St. Clair continued in her Contribution continued. “This issue could be resolved very quickly. However, this is not the case and the burden is placed on the victims.”
Hillary Nappi, a partner at AWK Survivor Advocate Attorneys, a firm that represents survivors of sexual abuse and human trafficking, points out that Grok’s security deficiencies in this area pose additional risk to anyone who has personally experienced sexual violence. “For survivors, such content is not abstract or theoretical, but causes real, lasting harm and years of revictimization,” Nappi says. “It is vital that meaningful, permanent regulations are put in place to protect current and future generations from harm.”
“Learn genitals quickly!”
Musk has long promoted Grok as superior to its competitors by Images and animations of sexualized female characters shared, including “Ani,” an anime-style companion. A significant portion of the bot’s loyal user base has fully embraced this application of the technology, attempting to create hardcore pornography and swapping tricks to get around the bot’s restrictions on nudity. A few months ago, a member of a Reddit forum for “NSFW” grok images happily announced that the AI model “learns genitals very quickly!” At this point the group was producing successful pornographic clips with the comic characters Supergirl and Harley Quinn as well as Elsa from the Disney film Frozen.
Despite all the evidence about what people actually use it for, Musk continues to tout Grok as a stepping stone to one complete understanding of the universe to. In July last year, he speculated that it could “discover new technologies” by the end of the year (which apparently didn’t happen) or “discover new physics” in 2026. But like so many of Musk’s grandiose promises These breakthroughs have not yet occurred either. Right now it’s all dirt and no science.