A stark warning from the CEO of a prominent artificial intelligence company suggests the use of AI could “threaten democracy” in the United States unless it is controlled.

Simona Vasytė, head of Perfection42, told Newsweek that the 2024 presidential contest might be remembered as the first major “AI election,” with bad actors able to use the technology to create deepfake videos of major candidates who appear to say things they never actually said.

A surge in AI capabilities, particularly regarding image and video generation, over the past couple of years has sparked increased concern over their impact on the upcoming presidential election. The contest is widely expected to be bitter, with both main candidates suggesting their rival is a threat to American democracy following the 2020 contest, when Donald Trump refused to concede defeat to Joe Biden.

Speaking to Newsweek, Vasytė said there are multiple ways AI could be used for election interference and predicted foreign powers are likely to get involved.

“We’ve seen an AI-generated video of Ukrainian President Volodymyr Zelensky ‘asking’ Ukrainian soldiers to lay down their weapons,” she said. “It’s entirely possible to see similar generated videos of presidential candidates right before the election.

“Another tactic might be to encourage youth voters not to vote in the election with fake videos on TikTok…If the election is close, it might become a determining factor, and that is definitely concerning.”

Photo-illustration by Newsweek/Getty

In March 2023, unknown hackers uploaded an AI-produced video, or deepfake, onto a Ukrainian news website of Zelensky telling his troops to stop fighting. While swiftly debunked, it triggered renewed concern over how deepfakes could be used to influence politics.

Newsweek reached out to the Federal Election Commission for comment via email.

Vasytė said we should “expect” the Russian state to try and interfere with the election, potentially using AI, as “American support for Ukraine might determine the outcome of the war” which has been raging since President Vladimir Putin’s invasion of the country in February 2022.

In terms of impact, Vasytė said AI is most likely to play a decisive role if the presidential contest is tight.

“The closer the race, the bigger the impact of AI will be during the election period,” she said.

“It’s important to note that AI can be used not only for malicious but also for positive content generation, so in that case, it might be the first major ‘AI election’ where both parties will try to keep up with the changing game of campaigning.”

Ensuring AI doesn’t subvert democratic processes will require both legislation and individual vigilance, according to Vasytė.

Donald Trump and Joe Biden
From left, Republican presidential candidate Donald Trump in New York on March 25 and Democratic President Joe Biden at the White House on March 26. The 2024 presidential race could be remembered as the first…

Alex Wong/Andrea Renault/Star Max/GC Images/GETTY

“Laws on AI regulation will soon be a necessity,” she said. “But to catch malicious use of AI, we need to invest in AI recognition software. Think of it as an antivirus on your PC—catching anything malicious before it can harm you. As the threat of malicious AI use grows, so will the need of these programs who can check videos for their authenticity.

“Rule of thumb: If it’s hard to believe someone would say that, it’s probably not true. Always fact-check information with reliable sources. Don’t trust everything you see on social media. Regarding AI-generated content, look for inconsistencies in movement, especially when looking at the background. Also, look for distorted objects, anything that looks out of the ordinary.”

Perfection42 specializes in providing AI tools for artists and other content creators. According to its website, the company creates “solutions for automating repetitive and mundane tasks” so artists can focus on creativity.

Speaking to Newsweek in 2023, Henry Ajder, a deepfakes expert who presents a podcast on the subject for the BBC, predicted the widespread use of deepfakes in 2024 for political purposes.

“So, for me, 2024 is looking increasingly likely to be an election where deepfakes are deployed quite broadly,” Ajder said.

“The real key question is whether there is an incredibly high-quality deepfake, which is very hard to authenticate or falsify and is linked to a critical period in the electoral process, say the eve of an election or before one of the debates, and becomes part of the narrative from the opposition.”