Officials piloting software to tackle deepfakes ahead of Scottish and Welsh elections | AI (artificial intelligence)
Electoral authorities are “rapidly” working on a pilot project with the Home Office to combat the use of deepfakes to target candidates in this year’s Scottish and Welsh elections.
Electoral Commission officials in Scotland said they and the Home Office expected software that can detect deepfake videos and images created by artificial intelligence to be operational before election campaigns begin in late March.
Sarah Mackie, the commission’s chief executive in Scotland, said if the software detected a fake video or image, officials would contact police about the candidate concerned and inform the public, but acknowledged this could not always be 100% certain.
He said they would then call on the social media platform to remove the content. However, as such action is currently voluntary, the commission is also seeking legally enforceable “takedown” powers that would require media platforms to remove fraudulent material.
He said the commission was calling on the UK government to introduce such powers.
“What we don’t have right now and what we want are called takedown powers; we approach social media companies and request that something be taken down,” Mackie said.
There are no known cases of deepfakes appearing during election campaigns in the UK, but their use has increased in elections abroad; this trend has accelerated dramatically with the proliferation of free AI rendering tools.
British elections and referendums have been repeatedly targeted, often by state-sponsored fake social media accounts From countries such as Russia, Iran and North Koreaoften designed to spread dissent or strengthen debate.
Speaking at a pre-election briefing for journalists in Edinburgh, Mackie said the commission was also working on a “safety and trust” project with the Scottish parliament and police to support women and black, Asian and ethnic minority candidates who have experienced harassment or harassment based on gender or race.
He said a 2022 study found that nearly half of female election candidates had been harassed, and many said they would never be able to resist again. Candidates from minority ethnic backgrounds also said the abuse left them too scared to stand up again and undermined diversity at Holyrood.
Mackie said the rise in artificial intelligence-driven and pornographic “strip” technology, particularly produced by Elon Musk’s Grok AI platform, would fall into this category and be reported to the police if used during the election.
While Musk’s X and Grok platforms have been criticized for failing to remove fake, pornographic and harmful content, senior politicians in Westminster have called for urgent action from the government and media regulator Ofcom.
Mackie said there was no clear legal role for the Electoral Commission or other bodies to regulate deepfakes during elections, but that both the commission and the Home Office should test what measures they could take.
If successful, the pilot project could be rolled out to all UK elections, he said.
He said: “We are not organizing the campaign but there is an empty area here where it looks like there are a lot of regulations surrounding ringside.
“So what we do is kind of, let’s see what we learn from this and then jump into the ring sharing it with other people.”
The Ministry of Internal Affairs was contacted for comment.
