Uni of Surrey researchers warn AI deepfakes could influence elections
The report warns it could be 'easier than ever' for misinformation to spread to voters
Researchers at the University of Surrey say there's "three months to save democracy" - as they warn of the impact artificial intelligence (AI) and deepfakes could have on elections.
A new AI election report has been published ahead of polls in the UK, US and to the European Parliament.
It warns it could be "easier than ever" for misinformation to spread to voters.
The report is calling for clearer labelling of AI-generated content, and for more research into how it can influence voters.
It's also calling for laws making social media companies responsible for content on their platforms.
Speaking to Greatest Hits Radio, Dr Andrew Rogoyski, Director of Innovation and Partnerships at the Institute for People-Centred AI, said the possible influence of deepfakes was "very worrying."
"Something like two-thirds of the population consume news online, with nearly half getting their news from social media," he said.
"I don't we're taking it our dependence on digital media seriously enough to realise that actually it makes us vulnerable to some of the misuses, the disinformation, that's starting to appear."
Dr Rogoyski was particularly concerned these might not appear "until the last minute" of an election campaign.
"You've got no chance to respond to it and change people's minds, but these bits of news may have a profound impact on the way that people vote."
He's urging people to learn how to spot deepfakes and AI-generated content, and check any stories with multiple sources.
"If a story looks interesting, if it's provocative, it may be perfectly true, but it also may be something that somebody's cooked up to cause problems.
"So go to different news outlets, and check the story, and see whether reliable, reputable people are echoing that story."