On September 14, Agora Digitale Transformation published research identifying at least 138 women and nine men from national parliaments across 22 EU countries in a deepfake-abuse ecosystem. The analysis examined roughly 160 domains linked to deepfake abuse and nonconsensual sexual content, and Benjamin Shultz calculated that women MPs were 33 times more likely to be targeted within that ecosystem.
The finding is about more than fabricated images. Sexualized deepfakes can be used to intimidate, humiliate and damage the reputations of women in public life—raising the cost of political participation before a candidate ever reaches office.
A research snapshot of a wider abuse ecosystem
The study covered 5,872 national parliamentarians from all 27 EU member states. Members of the European Parliament were not included, so the figures concern national MPs rather than the EU institution’s elected members.
Germany, the Netherlands, Italy and France had the highest numbers of politicians appearing in the findings. The study did not produce a count of every accessible video or image. Instead, it mapped several ways that politicians could appear within the ecosystem.
At least half a dozen domains identified during the research had been removed before the investigation was published on September 14. The number of live depictions therefore changed as the sites disappeared, while the research snapshot remained tied to the period in which the domains were examined.
What the study counted
The research separated three broad categories:
- Hosted material: explicit content available on a site.
- Indexed remnants: pages still appearing in search results after material had been removed.
- Profiles and gateways: pages connecting a politician’s biography or images to tools that could potentially generate new material.
That distinction matters. A politician appearing in the study does not necessarily mean that a currently accessible explicit video existed for that person. The counts refer to people referenced or depicted across the ecosystem, not to a verified total of distinct live files.
Benjamin Shultz compiled the parliamentarian database, searched the sites and categorized the results. He also alerted affected MPs individually and provided guidance about possible removal routes.
The ecosystem is not just a collection of isolated pranks. AI adviser Henry Ajder described platforms designed to operate like businesses, with infrastructure that makes the abuse easy to reproduce at scale. A gateway page can be part of that process even when it does not host the final image or video itself.
Why the gender gap matters politically
The reported 33-to-1 calculation applies to women MPs compared with men within this particular research ecosystem; it is not a universal risk ratio for every politician. The gender imbalance nevertheless fits the mechanism described by affected lawmakers and experts: sexualized fabrication turns a person’s body and reputation into a tool for political intimidation.
Dutch MP Suzanne Kröger described the abuse as part of a broader pattern of online hatred and intimidation aimed at women and female politicians. Sarah Dobbe, another Dutch MP, said deepfake videos are used both to intimidate politicians and to belittle women.
That pressure can reach beyond the immediate victim. Benjamin Shultz said the abuse can discourage women from seeking office and push women already in public life out of it. Josephine Ballon of HateAid similarly described the dissemination of sexualized images as a way to degrade women and undermine their political standing.
The practical danger is also informational: fabricated sexual material can be mistaken for authentic content, especially when it appears alongside real photographs and biographical details. That makes the abuse useful not only as a personal attack, but as a way to contaminate the public record around a politician.
What changed after the research was gathered
The research was assembled in July 2026 and published on September 14. By publication, at least half a dozen identified domains had already disappeared. That development affected the sites’ live presence, not the number of politicians recorded in the study’s dated snapshot.
The disappearance of individual domains does not remove the wider problem described by the research. Profiles, indexed remnants and links to generation tools can connect the same person to an ecosystem even when one page is no longer available.
The policy response remains broader than one rule
The European Commission says transparency rules under the EU AI Act took effect in August 2026. The Digital Services Act also establishes transparency and accountability obligations for platforms. Those measures form part of the European response, while the supplied policy context does not make one EU-level rule a universal criminal ban or guarantee immediate removal in every member state.
In the United States, the TAKE IT DOWN Act became Public Law 119-12 on May 19, 2025. Its criminal prohibition took effect immediately, and covered platforms received removal obligations after notification, including a 48-hour period described in legislative materials.
For lawmakers such as Kröger and Dobbe, rapid removal and direct support for victims remain central alongside regulation. The study’s numbers put a sharper figure on why: women MPs were not simply present in the ecosystem; within the research population, they were calculated to face targeting at a dramatically higher rate.
