Researchers are warning that time is running out for regulators to address the threat that artificial intelligence (AI) poses to the democratic process ahead of July's elections.
But the idea that deepfakes could influence election outcomes should not be taken with a pinch of salt because there is no clear evidence that election outcomes are influenced by AI, the researchers said.
In a new report published by the think tank Alan Turing Institute, researchers argue that there are early signs that AI-generated content is undermining democratic institutions.
In recent months, a number of documents purporting to depict politicians have been circulating.
Sadiq KhanThe Mayor of London spoke out about deepfake audio created to make it appear as if he was making inflammatory comments ahead of Armistice Day. It almost caused “serious injury”.
AI-Generated Audio US President Joe Biden Appeared Urging supporters not to vote in New Hampshire Primary elections earlier this year.
Sam Stockwell, a research fellow at the Alan Turing Institute and lead author of the report, said online harassment against celebrities who are the subject of deepfake attacks could lead some people to avoid participating in online forums.
“The challenge of distinguishing AI-generated content from real content creates a range of problems in the future… Dismissing deepfake content as suspect allows bad actors to exploit that uncertainty, leading to the existence of fake news and problems with fact-checking, all of which are harmful to fundamental principles of democracy,” he said.
The report called on media regulator Ofcom and the Electoral Commission to issue joint guidelines on the fair use of AI by political parties in election campaigns and to call for voluntary agreements.
It also recommended issuing guidance to media on reporting on fake AI-generated content and ensuring that voter information includes guidance on how to spot AI-generated content and where to seek advice.
Dr Christian Schroeder de Wit, a postdoctoral research fellow in AI and security research at the University of Oxford, said solutions to combat the impact of deepfakes include adding watermarks to AI-generated content, as well as provenance-based solutions that link images with metadata to indicate whether the original information has been tampered with. Only the owner can change the provenance record.
read more:
AI experts say world unprepared for biggest election year
AI is “an increasing threat to the next general election”
Artificial intelligence helps Meta combat election misinformation
Stockwell added that with just weeks to go until the general election, “there are no clear guidelines or expectations to prevent AI being used to create false or misleading election information.”
“That's why it's so important that regulators act quickly before it's too late,” he said.
Dr Alexander Babuta, director of the Centre for Emerging Technologies and Security at the Alan Turing Institute, said: “While we should not overemphasize the idea that elections are no longer secure, particularly given the lack of clear evidence globally that AI can change election outcomes, we must nevertheless use this moment to act and make our elections resilient to the threats they face.”
“Regulators can do more to help the public distinguish fact from fiction and to ensure that voters do not lose faith in the democratic process.”
“We are well placed to combat false information about the voting process and election administration,” an Election Commission spokesman said.
“During this election cycle, we will work to provide voters with factually accurate information and correct any misleading information about how elections should be run.”
The committee added that it would set up a voter support hub on its website with links to relevant media literacy resources, and urged activists to be accountable and transparent.
“We recognise the challenges posed by AI but a new legal framework will be needed to regulate the content of election campaign materials,” they said.
“We will monitor the impact on the general election and stand ready to engage in any discussions about the impact on voter trust and confidence going forward.”
An Ofcom spokesman said: “We have noted the report's recommendations and will consider them carefully.”
