Ahead of the recent general election, politicians, industry groups and the media were sure this would be the ‘AI election’ in which advances in artificial intelligence would threaten the democratic process with disinformation.
Among the biggest concerns was the potential for deepfakes – AI-generated images, video or audio designed to look real – to sway voters during the general election.
In the lead-up to polling day, one of many major global leadership races set for 2024, fears that AI-powered deepfake trickery would affect the results were running high, according to numerous surveys and reports.
One of those, with a sample size of more than 2,000 British adults from digital ID startup Onfido, found that almost half (42%) of respondents believed that deepfakes would influence the election outcome.
That number went even higher when looking at the 18-34 demographic, 54% of whom believed AI-created lookalike images and videos would manipulate the results.
A separate study, published by the Alan Turing Institute the week of the election, found that just under 90% of respondents to its survey were concerned about deepfakes affecting election results.
These concerns went beyond the public. Reports from the likes of the think tank the Centre for Policy Studies (CPS) and Scientists for Labour (SfL) made clear that deepfakes had the potential to impact this and future elections.
But were these fears founded? The impact of modern AI technology on every facet of society is under heavy scrutiny, however, so far, the UK may have been spared an election influenced by digital doppelgangers.
News and social media verification group Full Fact told UKTN that, for this election at least, concerns may have outweighed the reality.
“For all the discussion about this being the ‘AI election’ we saw very few incidents of faked or manipulated content,” a Full Fact spokesperson said.
The organisation noted just two examples during the general election campaign.
Faked audio clips of Prime Minister Keir Starmer and Health Secretary Wes Streeting were circulated on social media. However, Full Fact told UKTN that both examples were “debunked quite quickly”.
“Our sense is that neither were particularly widespread or credible as they were fairly amateurish and unconvincing.”
Full Fact chief executive Chris Morris told UKTN that the efforts of fact-checkers were required more for “unpacking the many misleading claims by politicians about tax and the economy”.
Morris said: “While we must remain vigilant and careful about faked content that can spread rapidly online, we can’t let this distract us from scrutinising what politicians actually say, and holding them to account.”
Reform AI candidate debunked
Rumours did appear that Nigel Farage’s Reform UK had fielded an AI-generated candidate in Clapham and Brixton Hill, however, the theory has been debunked.
CPS’s head of technology and innovation, Matthew Feeney, told UKTN there was a “noticeable lack of viral deepfakes” during the campaign.
Feeney suggested the overwhelming odds of a Labour victory ahead of votes being cast may have discouraged attempts to sway opinion through AI deepfakes.
Feeney, however, warned there is “still the potential for the large-scale proliferation of politically motivated deepfakes”.
Sanjush Dalmia, part of SfL’s deepfake report, admitted that the “volume of prominent election-related” deepfakes was “lower than expected”.
Dalmia similarly said “AI deepfakes continue to pose a risk to the democratic process”, however, particularly as AI tools “become more advanced”, making deepfakes “harder to identify”.
“As with many national security issues, the goal is not to make a likely event unlikely, but rather to make an unlikely event almost impossible,” Dalmia added.
The post Did AI deepfakes actually impact the general election? appeared first on UKTN.
previous post