Alexios Mantzarlis

Director of the Security, Trust, and Safety Initiative (SETS)

Alexios Mantzarlis is director of the Security, Trust, and Safety Initiative at Cornell Tech. He found in January that Crush AI had run thousands of ads on Facebook and Instagram over the course of several months.

Mantzarlis says:

“Crush AI is not the only undresser advertising on Meta, but it has been one of the most persistent. They use dozens of different domains, hundreds of different accounts and thousands of different assets to circumvent moderation. Over the past year, I have flagged – and Meta has deleted – more than 10,000 ads for Crush AI. The number that got through is probably a multiple of that.

“Meta is not the only mainstream platform where these noxious tools of AI abuse find a home. Nudifiers have apps on Apple and Google stores, verified accounts on X, channels with 100,000 followers on Telegram.

“Millions of women and girls are being targeted with deepfake nudes – data indicates 6% of American teens have suffered from this type of AI abuse – and mainstream platforms need to do more to eliminate this threat.

“Meta's move is welcome, if late. The custom classifier to detect nudifier ads should help given the similarity of the lure used by many of these nudifiers, but even just yesterday I found more than 100 ads running on Facebook and Instagram promoting Crush AI and similar tools.

“Researchers, media and activists need to keep pushing platforms to do more.”