Deepfake abuse sites index 138 women MPs beside links to nudifier tools

Shultz emailed every MP whose name he found and told her how the material might come down ^[11]. Some of what he found is a profile page. His profile and gateway category covers pages holding a woman's name, her official photograph and sometimes her social accounts or personal details ^[6][7], sitting next to a link to a nudifier app or to the site's own tool claiming to digitally undress people ^[8].

"It has all the images, all the information, and a direct link to the tool, and in three clicks you can generate," Shultz said ^[9]. "That's just as bad as actually hosting the content" ^[10].

The removal workflow starts with a hosted file, and these pages are assembled out of material a parliament publishes on purpose. Removal only goes so far. Shultz's indexed category counted results where the content had gone and the MP's name still came back ^[6].

He compiled more than 5,800 national MPs in July and ran custom searches against about 160 domains ^[2][3]. The 147 names that came back are roughly one in 40 of that list ^[18], and they run about 15 women for every man ^[19]. Shultz puts the likelihood of a woman MP being targeted at 33 times higher ^[5], a figure WIRED reports without showing the working behind it ^[21]. German, Dutch, Italian and French politicians appeared most often, and the more senior the politician, the more likely she was targeted ^[12]. The work was published by Agora Digitale Transformation, a German digital democracy think tank where Shultz is a fellow, and it excludes members of the European Parliament ^[13][14].

"Deepfake porn is part of a much bigger trend of online hate and intimidation against women and female politicians," said Suzanne Kroger, a Dutch member of parliament with the Progressive Netherlands group who was affected ^[15]. "We are politicians, we have a platform and speak out," Kroger said. "But meanwhile, there are millions of women victims of these criminal acts" ^[16]. Sarah Dobbe, a Dutch Socialist Party MP who was also affected, said it is "very important that if a woman or girl is victimized in this way, she can get direct support if needed, and that the content is removed quickly" ^[17].

For a team with an image tool in the product, sort the cases on two axes: whether your surface holds the output or helps make it, and whether the subject is named. Held and unnamed is what the abuse queue was staffed for. Held and named adds a legal escalation. Made and unnamed is a model policy argument. Made and named is the case in this research, and most terms of service cover it with a prohibited-use clause while the input stays open.

If you ship face editing, refuse a named third party's photograph at the input. That has a price. Separating a consenting subject from a target needs either an identity step you probably do not want to build or a review queue slow enough to annoy people using the feature honestly. Teams that decline to pay it find out which public figures their users are naming when a researcher's email arrives. That is how these 147 MPs found out ^[11][18].