Governments ban the Grok chatbot due to nonconsensual bikini pics : NPR
UK communications regulator Ofcom has launched a formal investigation into Elon Musk’s social media platform X over its AI chatbot, Grok, following reports that Grok was allegedly used to generate non-consensual sexual deepfakes.
Leon Neal/Getty Images
hide caption
toggle caption
Leon Neal/Getty Images
Indonesia and Malaysia temporarily blocked X’s chatbot, Grok, over the weekend after it posted numerous fake images publicly sexualizing mostly women and, in some cases, children late last year.
Governments around the world are also launching investigations. The latest came on Monday as Britain’s media regulator Ofcom launched an investigation into the social media platform, which could result in a ban.
Grok had been generating sexually explicit images of people for some time. But the issue gained attention in late December, as people used the chatbot to edit a large volume of existing images by tagging the bot in comments and giving it prompts such as “put her in a bikini.” Although Grok did not respond to all requests, he agreed in many cases. In some cases, Bellingcat researcher and principal investigator Kolina Koltai noted, users can ask Grok to generate full-frontal nudes.
Countless women and, in some cases, children, as first reported by Reuters, have had their images sexualized online by Grok without their permission, including one of the mothers of X owner Elon Musk’s children.
It’s unusual for so many governments to take action against a social media company, but this case is different, said Riana Pfefferkorn, a policy researcher at Stanford University. “Sexually abusing children [material] is clearly illegal, almost everywhere on Earth. »
Last Friday, X restricted Grok’s AI image generation feature to make it only available to paid subscribers. Non-paying users can still put people in bikinis publicly with a few clicks, but they can only make a few such requests before being asked to sign up for a premium membership, which costs $8 per month.
NPR reviewed publicly available images of Grok generated earlier this month and found that she stopped taking images of scantily clad women several days into 2026. However, she sometimes still features men in bikinis.
xAI, the parent company of X, has been offering adult content with Grok since last year. In May, Koltai first noticed that the chatbot would generate sexually explicit images in response to queries about X such as “take off your clothes.” Last summer, Grok introduced “spicy mode” to its standalone app, which allowed users to put bikinis on AI-generated characters.
Ben Winters, director of AI and privacy at the consumer advocacy organization Federation of America, said Grok now not only allows editing of images of real people, but also provides an easy distribution platform through X. “It’s a further and significant escalation,” he said.
Governments are outraged by X’s decision to restrict access to the image generation feature to subscribers. UK Technology Secretary Liz Kendall told Sky News “it’s insulting to say you can still access this service if you pay for it.”
The Indonesian government found that Grok lacked effective safeguards to prevent users from creating nonconsensual pornographic content based on real Indonesian residents, the Associated Press first reported.
“The government considers non-consensual sexual deepfakes a serious violation of the human rights, dignity and safety of citizens in the digital space,” Indonesian Minister of Communication and Digital Affairs Meutya Hafid said in a statement.
The AP also reported that Grok would remain stranded in Malaysia until effective safeguards against abuse were put in place.
In response to questions from NPR, The statement echoes Musk’s own message earlier today.
Such an approach constitutes an “attempt to abdicate responsibility,” Winters said.
“It’s definitely not just the user asking for it,” he said. “It’s the fact that the image wouldn’t be created without…the tool that they created.”
Before the wave of explicit, non-consensual deepfakes generated by Grok reached X, other AI makers had added similar features to their chatbots. In November, Google released a new image generation model, Nano Banana Pro. In December, OpenAI updated its ChatGPT Images model. The two can also edit images to put people in bikinis, Wired reported last month. A thread on Reddit distributing such images has been deleted.
In another X article, Musk suggested that since other AI companies’ tools also have this undressing feature, pressure from governments is a form of censorship directed against his platform.
Koltai noted that the trend toward creating AI-generated, non-consensual intimate media has been underway for a year or two. “You’ll even see these ads on Instagram – it’s like uploading a photo of you and your crush and you can kiss,” she said. “So it’s kissing and cuddling, and then you see there’s this more extreme spectrum.”
“There’s obviously a huge problem within the tech industry, because we’ve seen it across multiple platforms, and [there is] not always good guidelines, limits or regulations,” Koltai said.
In the United States, criticism has been much more muted than in other countries. Sen. Ted Cruz, R-Tex., posted on X Wednesday that the images “should be removed and guardrails should be put in place.” He also said he was “encouraged that X has announced that it takes these violations seriously and is working to remove all illegal images and offensive users from its platform.”
Grok has sparked frequent controversy over the past year. Last summer, the chatbot called itself “MechaHitler” and spewed anti-Semitic conspiracy theories.
Apps with so-called nudification capabilities have existed for years, mostly in the shadows of the Internet. Winters said officials need to do more to police some features of X. “There are misrepresentations about the security of their products. There are violations of their terms of service,” he said.
“We have not seen any significant action from U.S. agencies, whether state or federal, that have the authority to enforce the law,” Winters said.
