Skip to content
AI.info

The Pulse

Study Finds 147 EU Politicians Linked to Deepfake Porn

An Agora Digitale Transformation audit found that 147 of 5,872 EU national parliamentarians appeared in or were associated with deepfake pornography. Women lawmakers were 33 times more likely than men to be exposed, the study found.

Study Finds 147 EU Politicians Linked to Deepfake Porn

AI.info Team ·

147 lawmakers appeared in the audit

An audit of 5,872 sitting members of national parliaments across the European Union found that 147 politicians appeared in or were associated with deepfake pornography. The group included 138 women and nine men, according to a policy brief published on September 14, 2026, by Agora Digitale Transformation.

The study found that women parliamentarians were 33 times more likely than their male colleagues to be exposed after the researchers adjusted for age, migration background, institutional prominence, political orientation and national legal context. Exposure affected roughly one in 14 women in the sample, compared with one in 500 men.

Researchers caution that the result does not mean every person counted had an active fabricated nude image circulating online. The audit included live depictions, pages linking a named politician to AI image-generation tools, indexed search results and records that remained visible after a domain had been removed.

Public photographs supplied the raw material

Benjamin Shultz, the study's author and a Data Fellow at Agora Digitale Transformation, examined the names of lawmakers from all 27 EU member states between July 16 and July 20. The team queried 160 domains associated with deepfake pornography, so-called celebrity leaks and other non-consensual or sexually invasive imagery, using eight Google Custom Search engines.

Thirty-six domains produced relevant results. The researchers say they did not contact producers or distributors, download the material or store depictions. Instead, they coded search results into categories based on whether a depiction was active, whether a page supplied biographical information and source photographs, or whether a search record survived after the original material disappeared.

Political office itself supplied much of the material needed for abuse. An official portrait, a photograph from a public event or a campaign image can give an image-generation system enough visual information to fabricate a sexualized depiction without any participation or consent from the person shown.

“Gateways” connected names to nudifier tools

The report identifies a class of pages it calls “gateways.” These pages assembled a politician's name, biography, photographs and links to AI tools designed to create nude or sexualized images. The audit found 61 gateway instances with source material present and 36 without source material, for 97 opportunities to assemble a deepfake locally.

Three politicians had gateway pages that displayed active deepfake non-consensual intimate imagery. Other pages provided enough identifying information and links to generation tools for a user to obtain or create such material independently.

That distinction matters because laws aimed at removing a finished image may not reach pages that provide the ingredients and instructions for making one. A site can avoid displaying an active depiction while still directing users toward the tools, photographs and identifying information needed to produce it.

Domain seizures did not erase the search trail

The audit also examined what happened after authorities seized the deepfake-hosting domain Cfake. More than a month after the seizure, the researchers retrieved 47 records from Google’s Custom Search API that still pointed to material associated with the domain.

The team could not reproduce those results through ordinary manual Google searches and says the finding requires independent replication. It nevertheless treats search indexing as a separate layer of exposure: removing a host does not necessarily remove the name, image or accusation from search systems.

For people targeted by fabricated sexual imagery, that persistence can extend the harm after the original site has gone offline. The report argues that regulators and platforms need to address both the host that stores the material and the index that continues to surface traces of it.

Legislation alone did not predict lower exposure

Countries with laws specifically addressing deepfakes did not show lower exposure in the audit. Agora does not interpret that result as evidence that legal protection has no value. The researchers say statutory language cannot show whether authorities investigate complaints, remove material quickly or devote resources to enforcement.

The brief points to the difference between having an offence on the books and applying it in practice. Some countries without a dedicated deepfake statute can prosecute the material under privacy or image-abuse laws, while countries with specific offences may show limited enforcement activity.

That finding also limits what can be inferred from a single cross-country comparison. The researchers say future audits should track prosecutions, takedown response times and enforcement resources rather than recording only whether a country has passed a deepfake-specific law.

Agora calls for intervention before images exist

The policy brief recommends shifting attention upstream from post-publication removal. Its proposals include consent-based safeguards in image-generation models and interfaces, stronger controls on hosts and search indexes, and separate enforcement against pages that connect public figures to deepfake tools.

Agora also argues that European policymakers should measure exposure repeatedly. The project plans to re-audit the same population after platform shutdowns, new laws and enforcement actions, allowing researchers to distinguish a genuine reduction in harm from displacement to another site or jurisdiction.

The first audit presents a narrower claim than the headline number might suggest: 147 EU lawmakers were found in a set of search and hosting records tied to deepfake pornography, not necessarily in 147 confirmed active videos or images. Its clearest result is the gender gap—138 women compared with nine men—and the finding that public office can provide enough material for abuse without any further action by the target.

Source

Agora Digitale Transformation

Explore

More articles