Swedish Deputy Prime Minister Ebba Busch addressing parliament naked. British actress Millie Bobby Brown, known from the Netflix series Stranger Thingswhose pants and shirt have been replaced by lingerie and stiletto heels. Samantha Smith, columnist for the conservative weekly The Spectator and the monthly magazine The Catholic Heraldwho stands at a demonstration in a tiny bikini with a British flag around her shoulders. American pop star Taylor Swift with white stuff on her face euphemistically called ‘doughnut glaze’. Democratic Congresswoman Alexandria Ocasio-Cortez in a swastika bikini.
Billionaire Elon Musk’s social media platform The images were created by Grok, X’s chatbot, which was instructed by users to edit existing photos without the consent of the people depicted. This has caused great outrage worldwide. Celebrities affected say they feel violated and humiliated by the exposed AI images.
Ashley St. Clair, a conservative influencer who has a son by Musk, is one of the victims. X-users had asked Grok to create sexually explicit images of her, including one based on a photo from when she was fourteen. X removed some of them at her request, but not others because they would not violate the guidelines. “As the mother of Musk’s child, I do not speak here from a privileged position,” she declared in return for The Washington Post. “I go through the same procedures as everyone else, and this is what I get.”
Political backlash
Government officials from the European Union, France, India and Malaysia have launched a criminal investigation and threatened legal action against xAI if Musk’s company that developed Grok does not take steps to prevent the generation of deepfakes and child pornography and remove the images. The British media watchdog Ofcom said in a message on
A spokesperson for the European Commission said during a press conference on Monday that the commission is investigating the matter “very seriously”. “This is not spicy. This is illegal. This is horrible. This is disgusting. That is how we see it. And this does not belong in Europe.” According to the Digital Services Act (DSA), which has been adopted by all 27 EU member states, social platforms must limit illegal content and negative consequences of sex-related violence for minors and the physical and mental well-being of individuals.
Screenshot of a post from Samantha Smith about her AI image.
Musk initially responded rather laconically to the fuss. He posted several jokes about the issue on X, such as a Grok-generated image of herself in a bikini and an image of a toaster in a bikini commenting: “I couldn’t stop laughing at this.” As criticism mounted, Musk proudly posted about Grok’s rising popularity in the app stores. According to analytics firm Apptopia, the number of daily downloads of the app increased by 54 percent last week.
Last weekend, however, Musk changed his tune slightly. He wrote on X that “anyone who uses Grok or encourages the creation of illegal content will face the same consequences as if they upload illegal content themselves.” X’s security account stated in a message that illegal content will be removed and accounts posting such content will be permanently suspended. The company also pledged to work with local governments and police forces to track down violators.
Digital undressing
AI programs that digitally undress women, so-called nudifyers, have been around for years. But Grok lowers the threshold enormously – especially since xAI launched Grok Imagine last year, including ‘Spicy Mode’, a less censored setting of the image generator. Moreover, because Grok is integrated into X, users can quickly add a sexual twist to existing images and make them go viral. The recent flood of deepfakes seems to be simply the result of people imitating what they see others doing.
By allowing sexually explicit images, xAI differentiates itself from competitors such as OpenAI, Meta and Google, which have stricter rules for their AI chatbots. In September, the company released an update that seemed to further expand its guidelines for sexual content. While the update banned the creation of child pornography, it also stated that “there are no restrictions on fictional sexual content with violent or dark themes” and that “‘teen’ or ‘girl’ does not necessarily imply minor.”
The images that Grok Imagine generates are also becoming increasingly explicit because users are increasingly aware of how to circumvent the guidelines. On discussion platform Reddit, users share tips for doing this. “You can basically do anything now,” someone wrote on Reddit. “I can’t believe they allow this. Have fun!” In the attached video, Princess Elsa from the Disney film appears Frozen takes off her bra and a man begins to perform oral sex on her. On messaging service Telegram, users let Grok much more explicit images create.
So Grok’s loose sexual morals are intentional. Musk has often said that his chatbot should be as uncensored as possible. However, the scale of the problem is difficult to measure. Genevieve Oh, an independent deepfakes researcher, said to Bloomberg news agency that Grok generates approximately 6,700 nude or sexually suggestive images per hour. The five other most visited websites for sexual deepfakes generate about 79 AI images of naked women per hour, according to Oh.
Legal patchwork
Following the launch of Grok Imagine, the Midas Project, an independent watchdog for the AI sector, sent a fire letter to US prosecutors and the Federal Trade Commission warning that the ‘Spicy Mode’ would lead to “a flood of non-consensual deepfakes”. “Back in August, we warned that xAI’s app was essentially a nudifier that was about to be weaponized,” Tyler Johnston, director of the Midas Project, told Reuters. “That is essentially what happened.”
After the worldwide outcry about the sexually explicit images, calls for government intervention are growing
After the worldwide outcry about the sexually explicit images, calls for government intervention are growing. On Monday, the Consumer Federation of America (CFA), a group of hundreds of consumer-focused nonprofits, publicly called for action against xAI for “creating and distributing child pornography and other non-consensual intimate images using generative AI.” The CFA sent a letter, signed by a number of organizations, to the Federal Trade Commission and prosecutors.
But xAI doesn’t care much about that for the time being. In response to the outcry, the company is not promising to limit Grok’s technical capabilities, but merely warning users not to do anything illegal. But it is not really clear what exactly is prohibited under American law. For example, AI images of minors in bikinis do not necessarily appear illegal. Many states have legislation against sexually explicit deepfakes. But the US has no federal legislation like the UK’s Online Safety Act or the European Digital Services Act, and has much stronger protections for freedom of expression.
It is also difficult to determine what responsibility social media platforms have in enforcing legislation against sexually explicit AI images. Many Republican free speech advocates find any form of content moderation problematic. They believe that victims should file a civil case. But the patchwork of legislation against sexual deepfakes has barely been tested, said John Langford, a visiting professor of law at Yale Law School, to the tech website The Verge. “This is all quite new – we are only now starting to develop case law.”
Holocaust denial
European governments have more legal options to hold xAI accountable. For example, the investigation opened in France into Grok’s sexually explicit deepfakes comes on top of an ongoing investigation into anti-Semitic statements and Holocaust denial by the chatbot. In addition, the French Justice X suspects political interference in Europe. The platform is said to have adjusted its algorithms to emphasize themes favorable to the radical right and to give radical right candidates more visibility.
The European Commission already imposed X last month fines of a total of 120 million euros due to violations of the Digital Services Act. The committee ruled that the blue check mark next to user names is misleading. It is not a sign of authenticity, but is for sale to every paying user. In addition, the committee found that X gives researchers too little access to study how harmful messages spread on the platform. It was the first time that the commission issued fines for violations of the DSA.
But Musk does not accept this and increasingly seeks confrontation with Europe. “The EU has imposed this insane fine not only on X, but also on me personally, which is even more insane!” he wrote on Musk received support from US government members. For example, Secretary of State Marco Rubio wrote on
The journalistic principles of NRC