AI Deepfakes Emerge as Concern Ahead of Welsh and Scottish Elections

Artificial intelligence-generated fake videos of Welsh politicians have begun circulating online, raising concerns about misinformation ahead of the upcoming devolved elections in Wales and Scotland on May 7.

The BBC has identified several AI-generated videos featuring Welsh political figures that have collectively garnered more than 200,000 views. These include fabricated footage of Prime Minister Keir Starmer and Wales’ First Minister Eluned Morgan kissing, and another showing Plaid Cymru leader Rhun ap Iorwerth shouting “I love Reform.”

The creators of these videos, posted on Facebook, claim they are “obvious” satire and indicated their fabricated nature in captions. They stated they trust their audience to use “common sense” when viewing the content.

Another video, shared by an anti-Reform UK page, depicts party leader Nigel Farage with a Welsh flag backdrop, shouting: “Britain is great; I don’t see the point of having four different sporting teams and four different parliaments.” The BBC found no record of Farage ever making these statements, and the video displayed visual distortions consistent with AI generation.

Professor Martin Innes, a disinformation expert, described these examples as “shallowfakes” rather than “deepfakes,” indicating they were created using less sophisticated software. However, he warned that the “barriers to entry” for creating convincing AI-manipulated content have significantly lowered over the past two years.

“Where previously doing this kind of image or video manipulation would have required quite a lot of computing power and technical knowledge, you don’t need that anymore,” Innes explained. This democratization of technology could create a “trickle-down effect” where manipulated videos increasingly impact local and devolved politics.

The growing threat has prompted action from the Electoral Commission, the independent body overseeing UK elections. The Commission is collaborating with the Home Office to develop software capable of identifying, tracking, and reporting deepfakes.

Electoral Commission Chief Executive Vijay Rangarajan said the initiative would “help voters identify misinformation during election campaigns and reduce activity which negatively impacts voters’ confidence in the conduct of campaigners.”

The upcoming May 7 elections will see voters in Wales and Scotland selecting representatives for their devolved parliaments, coinciding with local elections in parts of England. These votes represent the first major electoral test for political parties since the UK general election.

While concerns about AI-manipulated media have circulated for several years, their actual electoral impact has been limited thus far. The Alan Turing Institute, Britain’s national center for data science and AI, found “no evidence” that AI-enabled deepfakes or disinformation meaningfully influenced the outcome of the 2024 general election.

However, the rapid advancement of AI technology is changing that calculation. Detecting sophisticated fakes is becoming “harder and harder,” according to Innes, with his research team sometimes spending an entire day “poring over” the more convincing examples.

Experts remain skeptical about the Electoral Commission’s current approach. Innes noted that while the Commission’s plans “might help us understand things after the event,” they would not prevent deepfakes from potentially influencing the election as it happens.

As AI tools become more accessible and outputs more convincing, electoral authorities face a growing challenge in maintaining information integrity during campaigns. The proliferation of these technologies raises important questions about voter trust, media literacy, and the evolving nature of political communication in the digital age.