Nearly 150 European politicians have been targeted by websites facilitating sexually explicit AI deepfakes, according to new research showing how low-cost generative tools are turning ordinary public photographs into raw material for nonconsensual sexual imagery.
WIRED reported that researcher Benjamin Shultz, working with Germany’s Agora Digitale Transformation, identified 138 women politicians and nine men across 22 EU countries who were listed or depicted by sexually explicit deepfake services.
The findings show a striking gender imbalance—and illustrate how generative AI can transform an abuse problem once limited by editing skill and production effort into something that can be automated and commercialized.
Public photographs have become source material
Shultz’s project examines 5,872 members of national parliaments across all 27 EU member states, comparing exposure to deepfake abuse against factors including gender, age, political prominence and national regulation.
Agora Digitale Transformation says producing manipulated sexual images once required technical ability and significant effort. Today, “nudify” applications and easily accessible models can create convincing material with a few clicks. Some services combine biographical information, photographs and generation tools into the same interface.
WIRED reported that researchers examined 160 domains, finding women lawmakers from Germany, France, the Netherlands and Italy among those most frequently targeted. More prominent politicians were also more likely to appear on the sites.
Dutch politicians Suzanne Kröger and Sarah Dobbe were among those discussing the damage such material can cause, including psychological harm and the possibility that sexualized abuse may deter women from participating in public life.
The problem is becoming a software ecosystem
What makes the research particularly important for technology companies is that the abuse is not confined to isolated fake images.
The underlying ecosystem includes generation models, “nudify” interfaces, hosting providers, search indexes and websites that organize targets. Agora’s research notes that traces can remain discoverable through search infrastructure even after individual content or entire domains have disappeared.
Commercial distribution also matters. WIRED previously found that Meta had run advertisements for an AI tool promising to create sexualized images resembling female politicians, despite platform rules restricting sexual material.
That means the technical problem extends beyond the model generating an image. Distribution systems, advertising platforms, app stores and hosting infrastructure can all become part of the pipeline.
Takedowns appear capable of changing the numbers
The research is designed to measure the same population repeatedly, allowing researchers to examine whether platform shutdowns, new laws or enforcement actions actually reduce exposure rather than merely moving the material elsewhere.
WIRED reported that earlier interventions in the U.S. and U.K. were followed by reductions in victim depictions on some services, suggesting that enforcement can have measurable effects.
For AI builders, this is a Human Impact story with a product-design lesson.
Generative technology has lowered the cost of producing synthetic media. But when the cost of abuse approaches zero, safety cannot depend entirely on victims discovering and reporting every generated image afterward.
The harder engineering challenge is increasingly upstream: deciding what generation tools permit, how platforms identify nonconsensual imagery, and whether the software infrastructure distributing those outputs is designed to make abuse scalable—or difficult.
