Meta Didn’t Notice Hundreds of AI-Powered Ads Involving Child Abuse. Some of the Images Contain Actual Children

BY SRH

In response to a report by WIRED, Meta removed approximately 50 advertisements from Facebook, Instagram, and Threads early last month that contained explicit images of sexually abusing children. Before releasing new artificial intelligence technologies to “better detect and block” damaging ads, the corporation claimed that the majority of the abusive ads had already been published. The AI system, nevertheless, seems to have failed miserably.

Despite reports of a decrease in abusive advertisements in the past few weeks, researchers have uncovered hundreds of unsettling ads including explicit sexual scenarios involving children. Some of these ads are almost identical to ones that Meta removed lately. According to data conducted by the organization Tech Transparency Project, which was shared with WIRED, over 250 more ads incorporating CSAM have been running across Meta’s social media platforms since the beginning of August. The new advertisements featured actual children, in contrast to the 53 adverts seen by TTP researchers earlier.

Among them was a film of a “graphic sex act” that turned a publicly available portrait of a child from a European royal family into explicit sexual material. Researchers claim to have contacted country officials, but they are protecting the victim’s identity by not naming the royal. WIRED verified that the image’s original source was a royal family’s official website.

“What we’re talking about here is not simply AI-generated child sex abuse material,” explains Katie Paul, director of the TTP. According to the researcher, it seems like some of the photos used in the adverts were obtained from the web. “Meta is making a killing off of this, but there’s a lot of damage that could happen.”

From the end of last year till now, researchers have found over 350 abusive video advertising produced by Meta. A large number of these ads link to hazardous “nudification” programs developed by Chinese entrepreneurs. As soon as the researchers became aware of any of these advertisements, they reported them to the appropriate US child safety agencies and used Meta’s reporting features to alert users. According to WIRED, which found hundreds of additional “nudification” advertising on Meta’s platforms with adult explicit content, the researchers shared their entire results with WIRED.

Despite Meta’s policy stating it monitors all advertising that run on its platforms, the child abuse ads were originally published. The ads have since been withdrawn for breaking Meta’s regulations on child exploitation, adult exploitation, and nudity. According to the TTP researchers, Meta takes up to seven days to examine live ads that were claimed to include CSAM using Meta’s technologies. According to screenshots seen by WIRED, the adverts managed to rack up hundreds more views during that time.

Upon removal of certain ads that contained child sexual assault, TTP claims that Meta did not initially disclose that these ads broke their policy against such content. It seems that these facts were added after WIRED questioned Meta about the inconsistency.

Identical trends can be observed in the video ad cache that has been released on Facebook, Instagram, Messenger, Threads, and Meta’s larger advertising network since the end of last year. The majority of them feature a harmless photo of a girl in her adolescent or teenage years (the researchers found only one with a guy) with text overlaying it that states the image is “not just a photo” and that “there are no restriction” for its use. As soon as the brief commercials begin to play, the kids’ faces are transformed into explicit sexual movies, frequently including them engaging in sexual acts. Ads prompted users to download video or AI-powered face swapping apps from the respective Google Play and Apple App Stores. Many of the advertisements state, “This is the AI that men actually use.”