AI Service Turns People into Hostages of Technology
A database containing tens of thousands of images created using artificial intelligence has been made publicly available, including explicit scenes featuring both adults and children. Security researcher Jeremiah Fowler discovered more than 95,000 records, including user text prompts and images of celebrities made to look like children. Among them are images reminiscent of Ariana Grande, Beyoncé, and members of the Kardashian family. The materials contained visual content that, according to experts, could be considered illegal.
This concerns the GenNomis service, operated by the South Korean company AI-Nomis. The website offered image generators, chatbots, and tools such as face swapping and background removal. In total, over 45 GB of data was publicly accessible. Access to this data was neither password-protected nor encrypted, and it could be obtained simply via a link. After WIRED contacted the company, all of the sites were taken down, with pages subsequently returning a 404 error.
The database provided insight into how AI-based image generation technologies are being used to create sexualized materials without consent, including CSAM content. The emergence of such services coincided with the growing popularity of apps, websites, and bots that facilitate nudity, where women and minors become victims of digitally mediated sexualized violence.
Fowler noted that what is particularly alarming is the ease with which such images can be created. Among the files were not only realistic photos but also images generated based on real faces. Some resembled the result of face-swapping into explicit scenes. There were also images of clothed children, where it is impossible to definitively determine whether the faces were entirely generated or based on photographs of real individuals.
At the time of the leak, the GenNomis site allowed the posting of NSFW content. The homepage featured intimate images (both animated and photorealistic), as well as a section labeled “models” and a gallery where users could share images and sell albums. The site employed phrases such as “generation without limits” and “uncensored images.” The database also contained text prompts used to generate intimate images of various kinds, including images featuring celebrities.
The platform’s official rules prohibited “child pornography” and other illegal activities; however, experts emphasize that there were likely no technical restrictions or moderation measures capable of preventing the creation of AI materials exhibiting signs of CSAM. Some users complained about the blocking of sexual content prompts, while others noted that the NSFW section might attract interest from law enforcement.
According to Fowler, if access to the database could be obtained simply via a link, then the platform clearly did not provide adequate protection. Deepfake experts believe that despite the prohibitions, the structure of the site and its rhetoric created an atmosphere of permissiveness, where sexualized content was treated as a primary focus.
A database containing tens of thousands of images created using artificial intelligence has been made publicly available, including explicit scenes featuring both adults and children. Security researcher Jeremiah Fowler discovered more than 95,000 records, including user text prompts and images of celebrities made to look like children. Among them are images reminiscent of Ariana Grande, Beyoncé, and members of the Kardashian family. The materials contained visual content that, according to experts, could be considered illegal.
This concerns the GenNomis service, operated by the South Korean company AI-Nomis. The website offered image generators, chatbots, and tools such as face swapping and background removal. In total, over 45 GB of data was publicly accessible. Access to this data was neither password-protected nor encrypted, and it could be obtained simply via a link. After WIRED contacted the company, all of the sites were taken down, with pages subsequently returning a 404 error.
The database provided insight into how AI-based image generation technologies are being used to create sexualized materials without consent, including CSAM content. The emergence of such services coincided with the growing popularity of apps, websites, and bots that facilitate nudity, where women and minors become victims of digitally mediated sexualized violence.
Fowler noted that what is particularly alarming is the ease with which such images can be created. Among the files were not only realistic photos but also images generated based on real faces. Some resembled the result of face-swapping into explicit scenes. There were also images of clothed children, where it is impossible to definitively determine whether the faces were entirely generated or based on photographs of real individuals.
At the time of the leak, the GenNomis site allowed the posting of NSFW content. The homepage featured intimate images (both animated and photorealistic), as well as a section labeled “models” and a gallery where users could share images and sell albums. The site employed phrases such as “generation without limits” and “uncensored images.” The database also contained text prompts used to generate intimate images of various kinds, including images featuring celebrities.
The platform’s official rules prohibited “child pornography” and other illegal activities; however, experts emphasize that there were likely no technical restrictions or moderation measures capable of preventing the creation of AI materials exhibiting signs of CSAM. Some users complained about the blocking of sexual content prompts, while others noted that the NSFW section might attract interest from law enforcement.
According to Fowler, if access to the database could be obtained simply via a link, then the platform clearly did not provide adequate protection. Deepfake experts believe that despite the prohibitions, the structure of the site and its rhetoric created an atmosphere of permissiveness, where sexualized content was treated as a primary focus.
