Insecure database reveals 1 million pornographic AI images and deepfakes

A massive collection of over one million AI-generated images and videos, predominantly adult content, was found unprotected online, according to recent research released on Friday. Some of the files appeared to depict minors, while others were based on photographs of real individuals.

Cybersecurity expert Jeremiah Fowler discovered the unsecured database in October. The server, accessible to multiple AI image-generation platforms including MagicEdit and DreamPal, contained files showing "faces of adults and young people digitally merged onto AI-created bodies," Fowler reported on the ExpressVPN blog. The dataset also included unedited images of real people, likely uploaded as references for the AI tools. It remains unclear whether any of these individuals consented to the use of their likeness.

Fowler emphasized the serious ethical and legal concerns surrounding AI tools that swap faces or bodies without consent. "AI-generated nudity or face-swapped images can be exploited for blackmail, harassment, or revenge," he said.

After notifying MagicEdit, the company secured the database, making it inaccessible to the public. MagicEdit stated it is "conducting a thorough investigation into the extent of the exposure." A DreamX spokesperson, representing both MagicEdit and DreamPal, highlighted the companys commitment to user safety, legal compliance, and transparency. The spokesperson also confirmed that the company does not tolerate child sexual abuse material (CSAM).

Following the exposure, both MagicEdit and DreamPals websites and mobile apps were taken offline. Wired reported that the MagicEdit app had previously been removed from Google Play earlier this year due to violations related to sexually explicit content. While MagicEdit stated it "does not encourage explicit sexual content," its website previously displayed images that transformed from dressed to bikini-clad AI-generated figures. DreamPal's site referenced adult-themed AI features, such as roleplay chat and AI companions, with some links mentioning sexual content. An FAQ on the site acknowledged the removal of NSFW AI chat filters that previously allowed adult-themed interactions.

In addition to notifying the companies, Fowler reported the exposed database to the National Center for Missing and Exploited Children due to the nature of some of the images.

Author: Logan Reeves