Spain's main opposition party has put out a campaign video built with AI deepfakes, in which likenesses of people linked to Prime Minister Pedro Sánchez's Socialists appear to confess to wrongdoing.
The conservative People's party (PP) shared the video on October 21, five weeks before the November 29 general election. Deepfakes are media generated or altered by AI to show someone saying or doing something on screen. Here the technique is used to stage admissions of guilt by figures tied to the governing party.
The notable detail is the sender. Earlier election deepfakes, such as the fake audio that circulated before Slovakia's 2023 vote and the AI-generated robocall imitating President Joe Biden in New Hampshire in 2024, came from sources that were anonymous or hard to trace. Those cases turned on detection: who made it, and how fast could anyone prove it was fake. A party posting its own deepfake removes the attribution problem entirely. The PP put its name on the material.
That shifts the debate from forensics to norms. Platform policies and election rules have mostly been drafted with covert fakes in mind, where the harm is deception by an unknown actor. A major party with a press office and a campaign calendar is a different kind of sender, and the five-week window gives the video time to circulate before ballots are cast.
The open question is whether rules written for anonymous fakes apply cleanly when the sender is a party that is happy to be identified.