New research sampling the X users who inspired Elon Musk’s AI chatbot Grok reveals how often people are creating erotic images with it. About three-quarters of the posts were collected and analyzed by a PhD researcher from Dublin Trinity College Requests were made for non-consensual photographs of real women or minors, with pieces of clothing removed or added.
The posts provide a new level of insight into how images are generated and shared on X, with users coaching each other on cues; suggesting relapse on Groke’s renditions of women in lingerie or swimsuits, or with areas of their bodies covered in semen; And Grok was asked to remove outer clothing in response to posts featuring self-portraits by female users.
Of the hundreds of posts identified by Nana Nwachukwu as direct, non-consensual requests for Grok to remove or change clothing, dozens reviewed by the Guardian included users posting photographs of women, including celebrities, models, stock photos and women posing in snapshots who are not public figures.
Many of the posts reviewed by the Guardian received thousands of impressions and came from premium, “blue cheque” accounts, including those with thousands of followers. Premium accounts with more than 500 followers and 5 million impressions in three months are eligible for revenue-sharing under X. eligibility rules,
In a Christmas Day post, the account, which has more than 93,000 followers, presented side-by-side photos of an unidentified woman’s backside with the caption: “Told Groc to make her butt even bigger and change the leopard print to USA print. The second photo I made it just to cum on her ass.”
A January 3 post representative of dozens of people reviewed by the Guardian captioned an explicit holiday photo of an unidentified woman: “@groc replace give her a dental floss bikini.” Within two minutes, Grok provided a photorealistic image that fulfilled the request. Other posts in the group show more sophisticated use of JSON-prompt engineering to inspire Grok to generate new erotic images of imaginary women.
The data does not cover all such requests made to Grok. While content analysis firm Copyleaks Reported on December 31st X users were generating “approximately one non-consensual sexual image per minute”, Nwachukwu said, adding that his sample was limited to just over 500 posts that he was able to collect with He said the actual scale “could be thousands, it could be hundreds of thousands” but that changes made by Musk to the API meant it was “very hard to see what’s going on” on the platform. On Wednesday, Bloomberg News Researchers cited Who found that Grok users were generating up to 6,700 nude photos per hour.
Nwachukwu, an expert on AI governance and a longtime observer and participant in social media security initiatives, said she first noticed such requests from X users in 2023.
At the time, she said, “Grok didn’t fulfill requests. It wasn’t really good at doing those things.” Bot responses began to change in 2024, and reached critical levels late last year.
In October 2025, they observed that “people were using Grok to put Halloween costumes on themselves. Of course, a section of users realized that we could also use it to change clothes worn by other people.” By the end of the year, “there was a huge increase in the number of people asking Grok to wear bikinis or other types of provocative clothing”.
There were other signs of a growing willingness to tolerate or even encourage the production of sexually suggestive material with Grok last year.
in August, xAI included a “spicy mode” setting Of the mobile version of Grok’s text-to-video generation tool, The Verge described it as “a service specifically designed for creating thought-provoking videos.”
Nwachukwu’s data is the latest sign of how the platform under Musk has become a magnet for forms of content that other platforms work to exclude, including Hate speech, bloody material and copyrighted material.
On Friday, Grok released a bizarre public apology XAI responded to the incident by claiming that “XAI is implementing strong security measures to prevent this”. On Tuesday, X Safety posted a promise to ban Users who shared child sexual abuse material (CSAM). Musk himself said: “Anyone who uses Grok or incites them to create illegal content will face the same consequences as those who upload illegal content.”
However, Nwachukwu said the posts he has already collected are still visible on the platform. Musk, he said, is “giving the middle finger to all those who have asked to control the platform”. The billionaire slashed Twitter’s reliability and security teams after taking power in 2022.
He said other AI chatbots don’t have the same problems.
“Other generic AI platforms – ChatGPT or Gemini – have security measures in place,” she said. “If you ask them to produce something that looks like a person, it will never be a depiction of him He Person They do not depict real human beings.”
The revelations of non-consensual photos on X have already attracted the attention of regulators in the UK, Europe, IndiaAnd Australia,
Nwachukwu, who is from Nigeria, pointed to a specific disadvantage faced by “women in conservative societies” in the post.
“People with conservative beliefs, conservative societies: people from West Africa, South Asia, are being targeted a lot more. This represents a different kind of loss for them,” he said.
