Grok shared more than 3 million sexualized images, 23,000 of minors, with the option to generate images in X

The image generator powered by artificial intelligence (AI) assistant Grok filled social network X with around three million sexualized images during the 11 days that was available, including 23,000 depicting children and nearly 1.8 million featuring women.

Grok’s imaging tool saw a spike in user usage around December 29, coinciding with a new option available in Use Grok to edit images published on the social network, with just one click.

On January 2, the company itself acknowledged identifying flaws in Grok’s safeguards, after the appearance of multiple publications on X in which the ‘chatbot’ agreed to the request of some users to undress women and children modifying their images to put them in bikinis or appear in sexual positions.

These AI-modified images could be generated at through the chatbot or in response to posts from other users without the consent of the victims, even though they are illegal practices which are prohibited in the own acceptable use policies of xAI, the company responsible for Grok.

Thus, after collecting multiple complaints By users and authorities in the European Union, the United Kingdom and Spain, among others, the company limited the Grok image generator on January 9, allowing it only for paying members. Finally, on January 14, the generation of sexualized images through Grok was blocked for all users.

During the approximately 11 days the tool was most active, social network three million sexualized imagesas reported by the Center for the Fight against Digital Hate (CCDH) and The New York Times, which have prepared estimates of the volume of sexualized images produced by Grok and published in X during this period, from December 29 to January 8.

Specifically, the analysis prepared by The Times shows that Grok published more than 4.4 million images and estimates that at least 41 percent of the publications, that is, the equivalent of 1.8 million contained sexualized images of women.

Thus, cases were included in which images of famous women such as Selena Gomez, Taylor Swift, Billie Eilish, Ariana Grande or the Swedish Vice Prime Minister Ebba Busch and the former US Vice President Kamala Harris.

For its part, the report shared by the CCDH, which was generated using a statistical model, details that the 65 percent of images generated by Grok -that is, just over three million- contained sexualized images of men, women or children. Thus, the AI ​​tool of the company led by Elon Musk produced a number of images over 11 days equivalent to 190 sexualized images per minute.

Among them, AI generated around 23,000 sexualized images of minors including content of girls in bikinis or child actors. Furthermore, the organization verified that on January 15, when these functions had already been blocked, the 29 percent of sexualized images of children They were still accessible. The data also reflects that Grok generated approximately 9,900 images sexualized images depicting children in a cartoon format.

As the executive director of the CCDH, Imran Ahmed, has stated in statements reported by the aforementioned media, these types of practices “contribute to large-scale abuse against women and girls.”

Likewise, he has highlighted the severity of its distribution on social networks such as. “Nudity tools have existed, but they have never had the distribution, ease of use, or integration into a large platform that Elon Musk achieved with Grok,” he said.

It must be taken into account that, to prepare the report, the CCDH used a random sample of 20,000 images from the larger total of 4.6 million produced by Grok’s imaging function during the aforementioned time. They also used an AI tool to identify the proportion of these images that depicted sexualized content of people.