“@Grok bikini now.” Three words are enough. A US entrepreneur writes it under a photo of Swedish Deputy Prime Minister Ebba Busch. She is wearing a blue dress. If you scroll down, you’ll see a new picture: Bush in a bikini. Generated by Grok, the AI chatbot from X, formerly Twitter. Another user adds: “Bigger breasts.” “String bikini.” “Let her pick up a pencil.”
For weeks, comments like: “Hey @Grok, put her in a bikini” have been piling up under posts from politicians, journalists and ordinary users. This is made possible by an image editing function that Grok introduced at the end of December. Elon Musk’s AI company xAI, which also owns X, has drastically lowered the threshold for deepfakes.
AI can be used to manipulate images and videos of real people – often pornographic, sometimes harmless. A study from 2019 showed that 95 percent of deepfakes circulating online are sexualized. Most of the time they show women without their consent. “It is therefore not at all surprising that an image generation tool is used for precisely this purpose.
HateAid is the first point of contact nationwide for those affected by digital violence, which also includes the creation of non-consensual deepfakes. The organization is in regular contact with those affected. “For them, such image manipulation means a massive invasion of their privacy and sexual self-determination as well as a serious loss of control,” says Ballon. In addition to psychological stress, there is also the risk of long-term reputational damage.
The creation of deepfakes is not yet explicitly punishable in Germany; the law usually only comes into effect when they are distributed. The right to one’s own image, data protection and personal rights apply. In the case of derogatory or sexualized deepfakes, insults, slander or slander are also possible criminal offenses, and copyright violations are also punished.
Those affected should first request the platform to delete the content. There are also three ways to take action: request a cease and desist under civil law, file a criminal complaint or contact the data protection authority – ideally in parallel. How quickly content disappears depends largely on the platform; If they don’t respond, legal action can be taken against them.
At EU level, the directive to combat violence against women and domestic violence also requires better protection against digital violence. This also includes the unauthorized falsification of images.
Restrained response to real damage
The EU Commission doubts whether X complies with European digital laws and asked the company to secure internal documents for an investigation by the end of the year. Allegations of AI-generated pornographic content are also being examined.
The damage caused by these deepfakes is “very real,” said EU Commission President Ursula von der Leyen Spiegel. “We will not outsource child protection and consent to Silicon Valley. If they don’t act, we will.” Last week, the EU Commission announced that it would examine complaints about the distribution of child pornography content on X.
Images of children are also generated
Even in the crowd, Grok takes the deepfake problem to an unprecedented level. According to a deepfake researcher Thousands of images every hour. One Investigation The European non-profit organization AI Forensics shows: More than 50 percent of the images generated by Grok show people in bikinis or underwear. In 97.5 percent of cases, the requests came not from the people depicted themselves, but from third parties – most likely without their consent. In the majority of cases, it was male-read accounts that had images created by women. Not only celebrities are affected, but especially private users, including minors.
AI Forensics documented 30 cases in which Grok generated explicit images of minors, including girls who appeared younger than five years old. The Internet Watch Foundation (IWF) also found criminally relevant images of children between the ages of 11 and 13 that were created with Grok and distributed on the dark web. These images are now being used to generate “even more extreme” videos with other AI tools.
Justice Department wants tougher penalties
Federal Justice Minister Stefanie Hubig announced that she would take more decisive action against the creation and distribution of sexualized images created with the help of artificial intelligence. The Ministry of Justice is working on a digital violence protection law that should punish deepfakes with appropriate penalties. At the same time, it must be easier for those affected to defend themselves against violations of their personal rights.
It is “terrible” when AI tools like Grok generate masses of lewd images of women without their consent, Hubig told the Neuen Osnabrücker Zeitung. Such practices systematically violated personal rights and reduced women to objects. The creation of sexualized AI images of children and young people is particularly serious. “We cannot accept that,” emphasized Hubig. Although there are already legal remedies against manipulated images, they are not enough. The new law should remedy this.
In Great Britain, the media regulator Ofcom has already launched an investigation into the incidents. In extreme cases, X could be blocked in the UK. Elon Musk reacted to the threat of legal consequences as usual: The British government wanted to restrict freedom of expression, he wrote on X. However, anyone who uses Grok for illegal content must expect consequences.
Under increasing pressure, X restricted the functionality for non-paying users. However, many requests and images remain publicly available. That’s not enough for the British authorities. X is turning a problematic function into a “premium service,” said a spokesman for British Prime Minister Keir Starmer. This is “not a solution”, but shows that X can act quickly – if it wants to.
Legal requirements are not consistently enforced
Josephine Ballon, HateAid
Starmer himself criticized the content on X as “disgusting” and called for it to be removed immediately. Otherwise you will intervene yourself. According to him, even the AI-friendly politician JD Vance expressed clear criticism during a phone call.
France’s prosecutors also expanded proceedings against X – because of the creation and distribution of sexualized content. Indonesia was the first country to completely block Grok, followed by Malaysia. In the USA, senators are calling for the X and Grok apps to be removed from the app stores.
“Platforms and developers like X and xAI have a clear responsibility to check and minimize the risks of new functions before they are introduced,” criticizes Ballon. X apparently failed to do this and was clearly endangering the rights of the users. Ballon sees this as a strategic step: the function increases user interaction, brings more clicks and therefore higher profits.
Voluntary commitment is not enough and legal requirements are not consistently enforced. “The risks must be addressed through regulation,” demands Ballon. With the Digital Services Act, the EU has an effective instrument that must now be applied consistently.
This is not the first time that AI has provided sexualized image manipulation and chat functions while neglecting the protection of women and children. Back in the summer, xAI made it possible to create nude deepfakes of well-known personalities in seconds. Meta also came under pressure because the chatbot Meta AI allowed romantic interactions with minors on Facebook, WhatsApp and Instagram. OpenAI announced in October 2025 that it wanted to introduce a so-called erotic mode for ChatGPT.
HateAid views these developments with great concern. “They have created a new potential danger for women and people who read as female, who now always have to fear that naked pictures of them will be spread as soon as they speak out online,” said Ballon. She emphasizes that the technology is now no longer just freely available, but is directly integrated into everyday applications such as ChatGPT and social networks. This would further reduce the hurdles and gender-specific digital violence would become increasingly normalized.
Self-help and emergency help
The Swedish Deputy Prime Minister has now himself commented on the pictures on X. With self-irony she films herself in snowy Sweden, where she is currently wearing thick winter clothing rather than a bikini. But the tone changes quickly. She makes it clear that she never consented to the creation or distribution of these images. “As a woman, I decide when, where and for whom I show myself in a bikini,” says Busch.
Then she talks about her daughter. She is 9 years old, at some point she will be 14, later 16. She doesn’t want her child to experience something similar one day. “What if that was your daughter?” she asks X users. Busch encourages people to think twice before sharing content. “The world needs more good men and women – and fewer assholes.” Busch’s photos have now been blocked.
xAi’s press department responded to inquiries from international media with the comment “legacy media lies”.