Deepfake Ad in Spain's Election: Has AI Crossed the Line in Campaigning?

A deepfake political ad released by Spain's main opposition party has put a question in front of Europe: where should AI stop in election campaigns? This time the fake content wasn't spread by an anonymous troll account. It was published by one of the country's largest parties on its own official channels. That difference turns the issue from a disinformation case into a question about whether this is the new normal in political communication.

What happened?

The conservative People's Party (PP) published a video ad on its social media accounts. The video features AI-generated likenesses of people linked to the Socialist Party, including Prime Minister Pedro Sánchez's wife. These figures appear to confess to things like paying for sex and taking bribes. With the country heading to a general election next month, the ad drew a fierce backlash.

Why are deepfakes so controversial in political campaigns?

Election deepfakes are nothing new. In 2023, just before Slovakia's election, a fake audio recording circulated that claimed to capture an opposition leader discussing vote-buying. In 2024, voters in New Hampshire received robocalls imitating Joe Biden's voice that urged them not to vote. In Argentina in 2023, candidates mocked one another with AI-generated images. In most of these cases, however, the creators were either anonymous or the content was openly presented as parody.

The Spanish case is different in two ways. First, the source is known and institutional: the main opposition party. Second, the content attributes crimes and moral failings to real people, some of whom aren't even active politicians. Caricature and satire have long been accepted as legitimate in politics. But a fake "confession" delivered with a realistic face and voice blurs the line between satire and accusation in the viewer's mind. Even if a disclaimer is added to the video later, the clipped copies shared on social media don't carry it.

The real risk is the precedent. Once a major party opens this door, its rivals will feel pressure to hit back with the same tools. In a polarized election, that means tit-for-tat escalation. There is also a separate cost when voters start viewing every video with suspicion. Through what experts call the "liar's dividend," genuine recordings can also be dismissed as "deepfakes."

What does the EU AI Act say about this?

Article 50 of the European Union's AI Act requires anyone publishing deepfake content to clearly disclose that it was artificially generated. But the article doesn't ban such content; it requires it to be labeled. That means a properly labeled deepfake ad may not technically break this rule. The law also provides a more flexible regime for artistic, satirical or fictional content. Parties may try to exploit this gap with a "satire" defense.

Here are the other pieces of the picture:

The Spanish government had previously drafted a bill imposing hefty fines for unlabeled AI content. But it remains unclear whether the bill can be enforced in time for this election. There's a timing problem: legal proceedings take months, while an ad does its work in a few days.

What changes for platforms and voters?

Much of the responsibility lies with the platforms. Meta and Google had decided to largely halt paid political ads in the region, citing the burden of complying with the EU's political advertising rules. That pushes parties toward organic posting, which is exactly the official-account channel the PP used. Organic content doesn't go through ad review. Labeling, meanwhile, often comes down to user self-disclosure or flawed automated detection tools.

Expectations for voters need to be realistic. Deepfake detection tools aren't reliable yet, and it's unreasonable to expect ordinary users to verify every video. That's why the burden needs to shift to creators and distributors. Using real people's likenesses without consent, especially when it involves accusations of crimes, shouldn't be legitimized by a label.

What's next?

There are three things to watch in the coming weeks. First, how Spain's electoral board and courts rule on the video. Second, whether platforms take the content down or simply label it and leave it up. Third, whether other parties respond with similar content. Together, these will set a de facto example for other elections across the EU.

Our take is clear: labeling is a reasonable solution where deepfakes are used as a technical tool. But it falls short for political ads that put fake confessions in the mouths of real people. Europe's rules regulate transparency, but they don't yet solve the problems of consent and reputation. The Spanish case shows that this gap is too wide to close within an election calendar.