Elon Musk’s Grok chatbot is testing Europe’s ability to clamp down on deepfakes and digital nudity of online images, even after regulators won a rare victory by forcing Musk’s xAI to curb the creation of sexualized images.
xAI said late Wednesday that it had restricted image editing for Grok AI users after the chatbot produced thousands of sexualized images of women and minors that alarmed global regulators.
The backtracking by Musk, who initially laughed off the trend, highlights the difficulty of policing AI tools that make it cheap and easy to create explicit content. It is the latest confrontation between Europe and Musk after disputes over electoral interference, content moderation and freedom of expression.
Many regulators continue to struggle to craft laws and rules governing AI, with questions about what constitutes nudity, how to define consent, and who is responsible: the user or the platform.
“This is a gray area when it comes to creating nude images,” Ängla Pändel, a data protection and privacy lawyer at Stockholm-based Mannheimer Swartling, told Reuters.
British regulator Ofcom, one of the most vocal on the issue, welcomed Musk’s move but said its investigation into xAI over the Grok images would continue.
“Our formal investigation remains ongoing,” a spokesperson said. “We are working tirelessly to move this matter forward and get answers about what went wrong and what is being done to fix it.”
Continue reading: Musk’s xAI restricts editing of Grok images after concerns raised by California and Europe
Earlier this month, Grok created hyper-realistic images of X-shaped women manipulated to appear to be in skimpy bikinis, in degrading poses, or even covered in bruises. Some minors were digitally stripped down to their swimsuits.
As of Wednesday, Reuters found that the chatbot was still privately producing sexualized images on demand. That appeared to have been curbed at least in certain geographies on Thursday.
Musk’s xAI said it was blocking users from generating images of scantily clad people in “jurisdictions where it is illegal.” It did not identify those jurisdictions.
In Malaysia and Indonesia, the government has imposed temporary bans on Grok, while EU and UK regulators called the images illegal. The United Kingdom, France and Italy launched investigations but faced calls for stricter measures.
Christian Democrat MEP Nina Carberry, who called the latest measure a “positive step”, said that “stronger enforcement of the Digital Services Act is needed to put an end to apps and platforms that sexualise or strip women and children.”
A European Commission spokesperson said that if Grok’s changes were not effective, the Commission would continue to use all EU DSA enforcement tools against the platform.
Legal gray area
According to Alexander Brown, a British data protection lawyer at Simmons & Simmons, the UK Online Safety Act makes it a “priority offence” to share intimate images without consent, including AI-generated deepfakes.
“This means that X must take proactive and proportionate measures to prevent such content from appearing on its platform and to remove it quickly when detected,” it said.
The British regulator can fine a company up to 10% of its revenue in the most serious cases of non-compliance or ask a court to force internet service providers to block the site.
For individuals, taking the platforms to court is “a really difficult and burdensome process,” said Anders Bergsten, a lawyer at Mannheimer Swartling, citing the emotional toll on victims.
Deepfakes have been around for years, long before the advent of AI applications, although they were largely limited to the darkest corners of the web. X’s publishing ability gives Grok unprecedented reach.
“The ability to publish without friction allows deepfakes to spread on a large scale,” says American lawyer Carrie Goldberg, who works with victims of cyberbullying.
British and Swedish laws make the non-consensual sharing of nude images illegal. Britain is expanding the law to include the creation of such images.
According to the DSA, suspending a service is considered a last resort. According to experts, the EU AI Law also does not cover images of naked adults, but only transparency obligations for deepfakes.
British Prime Minister Keir Starmer welcomed X’s move on Thursday, but warned: “Freedom of expression is not the freedom to violate consent. Images of young women are not public property, and their safety is not up for debate.”
“If we need to further strengthen existing laws, we are prepared to do so.”
With information from Reuters
Follow the technology information in our specialized section
