‘AI election’ roundup: Deepfakes and AI memes are taking off

With less than two months to go until US voters head to the polls, hypotheticals about AI’s role in election-related misinformation may be starting to get real.

Donald Trump’s campaign and at least one prominent billionaire backer have taken a shine to AI-generated political images, and the Department of Justice (DOJ) is warning that the Russian government is tapping the tech to further its efforts to influence the election.

Meanwhile, initiatives to crack down on election-related deepfakes have stepped up as well. California passed a trio of new bills that would further restrict the creation and distribution of synthetic media in the state, and Google is also strengthening its election content policies.

As the first US presidential race since hyperrealistic AI imagery and text have become mainstream, some experts have dubbed the 2024 contest the “AI election” and warned that tools could allow disinformation purveyors to operate at a bigger scale. There has been some debate, however, about the extent to which those predictions have come to pass.

Here’s a quick recap of some recent events surrounding AI and the election:

• This month, Elon Musk posted an AI-generated image on X of Vice President Kamala Harris in a red Soviet-esque uniform emblazoned with a hammer and sickle. The post follows AI-generated media Trump has posted or shared on X and his Truth Social platform, including images of Harris as a communist dictator, himself astride a lion (really!), and purported fans of both Taylor Swift and Trump.
• The Justice Department said Russian actors are tapping AI-generated content and spreading “AI-generated false narratives on social media” in an attempt to influence the election, according to the DOJ. The department said it seized dozens of web domains that were part of this campaign.
• California passed three new laws that would add more rules around political deepfakes to those it already has on the books. One requires more disclosures of AI media in political ads, another mandates that online platforms label or block political deepfakes during certain periods around an election, and a third creates more restrictions for individuals making and distributing AI-generated deceptive content. The bills now await Governor Gavin Newsom’s signature.
• Google said it would extend blocks around certain election-related content to more of its AI products, including Search AI Overviews, YouTube AI-generated summaries for Live Chat, and image generation within Gemini. The rules previously applied to the Gemini apps and web experience.

Shomir Wilson, an associate professor at Pennsylvania State University whose focus is language AI and computational social science, said much of the threat AI poses to elections comes from the scale of misinformation it allows bad actors to produce. It can also be difficult to reliably detect AI’s fingerprints on text, he said, making it tough to measure the scope of the problem.

“Anyone can just write, but being able to write in large volumes, and to write using copy conventions to write well, basically—or some value of well—at a large scale, that’s pretty new, and that’s continuing to evolve,” Wilson told Tech Brew. “So, that enables folks who want to generate this information to generate large volumes of it at once. And sometimes quantity is useful for getting your message across.”

Despite the attention paid to AI’s role in this election, Wilson said “we’re not at the peak yet” in terms of how sophisticated this technology is likely to get in the future, especially once text-to-video generation evolves more.

“The big difficulty here is there is the cat-and-mouse game, and generation is always, I think, a little bit ahead of the detection,” Wilson said.