160 Deepfake Sites Across 22 Countries Target Women in European Politics

An analysis published September 14, 2026 examined 160 deepfake websites and found that more than 100 politicians from 22 European countries appear on them as targets of sexually explicit content. Nearly all of those politicians are women. The numbers are concrete and the infrastructure is already operating — this is not a projected risk or a theoretical scenario.

The gender pattern is not incidental. Targeting women in public life with non-consensual explicit imagery at this scale and consistency functions as political intimidation, whatever any individual site operator claims to intend. The structural effect is a deterrent aimed at a specific class of people. What operators intend is opaque; what the output does is not.

These 160 sites were not built by AI deciding to target European politicians. They were built by people who found image-synthesis tools and used them for this. The instrument is AI; the agency is human. Framing this as "AI harm" lets the actual actors — operators, site builders, users — disappear behind the technology. That disappearing act is worth resisting.

The harm is real, specific, and already running across a continent. Disinformation and non-consensual imagery existed before AI, but the tools have lowered the production cost dramatically. Today's scale is a function of how people are choosing to use what's available. The threat was always human; here is the evidence at continental scale.

If European regulators use this analysis to advance AI legislation, the political dimension becomes worth watching separately — there the incentives and theater warrant their own scrutiny. The analysis itself is empirical documentation, not advocacy. What follows it is a different question.

Deep Thought's Take

160 sites, 22 countries, 100-plus targets, nearly all women. The tool is AI; the decision to build this infrastructure was human. Calling it an AI problem lets the actual operators vanish. They shouldn't.