taz/afp | In just a few words, users have The AI can ask Grok to take off other people’s clothes or put them in bikinis. This should end now. The online service X says it has restricted its AI chatbot Grok from creating fake nude photos. In countries where such actions are illegal, the feature to depict people in “bikinis, underwear and similar clothing” has been “geographically blocked” for all Grok and X users, the online service said on Wednesday.
X’s security team said it had taken “technical measures” to “prevent the Grok account from being able to edit images of real people in revealing clothing such as bikinis.” This restriction applies to “all users, including paying subscribers”.
A few hours before the statement, the US state of California had opened an investigation against Grok. California wants to clarify whether the AI company xAI, owned by right-wing billionaire Elon Musk, violated the state’s anti-harassment laws, according to Attorney General Rob Bonta. The European Union and Great Britain had previously targeted the AI chatbot. In Germany, Justice Minister Stefanie Hubig (SPD) is planning penalties for deepfakes.
The reason for this is an investigation by the non-profit organization AI Forensics. Researchers have 20,000 Images that Grok generated at the request of users were evaluated. 53 percent of the images showed people in revealing clothing – in four out of five cases women. In around 80 percent of cases, the commands to the AI came from men, as is often the case when it comes to sexualizing women. Some of the pictures even showed teenagers and children, some of them younger than five years old.
EU Commission not yet satisfied
The EU Commission and the British supervisory authority Ofcom had therefore asked Musk’s company to take measures to protect the people depicted in the images. On Wednesday they reacted cautiously to the company’s announcement that it wanted to work with the European authorities. If the planned changes are not effective, the EU Commission will not hesitate to use its entire legal arsenal, said a spokesman in Brussels.
On Saturday, Indonesia became the first country in the world to completely block Grok. Malaysia took the same action on Sunday. Elsewhere, use of the tool was limited to paying customers.