AINEWS
An audit of about 160 deepfake abuse domains found women lawmakers in 22 of 27 EU national parliaments on them. Targeting tracks seniority, the researchers say.
An audit of roughly 160 domains hosting sexually explicit deepfakes found women members of national parliaments across the European Union are about 33 times more likely than their male counterparts to appear on those sites: 138 women lawmakers from 22 of the bloc's 27 countries, against 9 men, according to a research-grade dataset published Wednesday by the German digital democracy think tank Agora Digitale Transformation and written up by WIRED.
Deepfakes are AI-generated videos or images that place a real person's likeness into fabricated scenes, in this case explicit scenes, without consent. Researcher Benjamin Shultz, a fellow at Agora and lead researcher at the American Sunlight Project, compiled the names of more than 5,800 national lawmakers across all 27 EU member states in July and ran each one through custom search engines tuned to surface deepfake domains. The dataset covers national parliaments only; members of the European Parliament are not in scope. The pattern, he found, scales with seniority: more senior politicians, including committee chairs, party leaders, and ministers, are more likely to appear, regardless of media profile. Germany, the Netherlands, Italy, and France account for the largest shares.
"This is really deeply disturbing," Dutch MP Suzanne Kröger of Progressive Netherlands, one of the women in the dataset, told WIRED. She framed the findings as "part of a much bigger trend of online hate and intimidation against women and female politicians," and noted that politicians at least have a platform to speak out, where most victims do not. Kröger's reaction is the only named-victim voice in the published coverage; the other 137 women MPs are not named in the article, in part to keep the audit from functioning as a target list.
The abuse takes two main forms. Some victims' likenesses are used directly in explicit deepfake videos. Others are catalogued in database-like entries that list names, headshots, and biographical details, including age, party, and parliamentary role, and sometimes link to deepfake creation tools that can produce new imagery on demand. The same sites also include celebrities, journalists, and other public figures, but the political slice is large enough to be structurally distinct. WIRED is withholding the domain names to avoid amplifying the abuse; the full methodology and country breakdown are published in the Agora and American Sunlight Project report.
Shultz individually notified every impacted MP with removal guidance, and Agora published the audit openly so that national cyber-nonconsensual-intimate-imagery (NCII) units across the bloc can reuse the same search method in their own cases. The publication lands as several member states are still writing takedown playbooks for synthetic NCII material, and a research-grade methodology is the kind of evidence those playbooks need to justify a coordinated response.
The central finding, the 33x disparity, rests on a single researcher's dataset. A second independent audit that fails to reproduce the same gender and seniority skew would weaken the claim; that replication is the most consequential next step before the disparity is treated as settled. The infrastructure layer, by contrast, is already documented. A 2025 Bellingcat investigation traced one of the most-trafficked deepfake sites to a single operator, evidence that the abuse economy is concentrated in a small number of hands rather than scattered across thousands of amateurs. The next test is whether the women-and-seniority pattern holds when a second research team runs the same audit on a different set of national chambers.