The Abundance Institute AI and Elections Update

Part I in our four-part series

Welcome to the first in our four-part series on AI and the 2024 elections. We will publish the remaining editions 90, 60, and 30 days out from the November 5, 2024, general election.

The U.S. votes in the 2024 presidential election in 180 days. This will be the first U.S. presidential election since the popularization of generative AI, which gives unprecedented access to tools capable of creating plausible, high-quality audio, visual, and textual content. 

Scholars and pundits claim that AI is going to transform our upcoming election, as one report puts it, in part  by “democratizing disinformation,” enabling “anyone [to] become a political content creator and seek to sway voters or the media.” In a widely viewed video, “The AI Dilemma,” Aza Raskin, co-founder of The Center for Humane Technology, claims that due to AI, “2024 will be the last human election.” CNBC host Brian Sullivan said, “There are terrifying possibilities with AI if used improperly to a vulnerable public, and the massive quick spread from social media.”

Policymakers are likewise worried. Senator Blumenthal (D-CT) opened a hearing on AI election deep fakes by stating, “AI is already being used to interfere with our elections, sowing lies about candidates and suppressing the vote.” In the same hearing, Senator Josh Hawley (R-MO) claimed, “[d]eepfakes are poisoning the public trust, especially around elections. Americans deserve to be able to go to the polls with confidence, not wonder whether every video clip or sound bite they see has been faked with AI. The Senate cannot wait to act.” 

The launch of ChatGPT in November 2022 sparked these and others’ public expressions of concern, which continue today.

What evidence of AI’s effects on elections have we seen thus far? 

Methodology

At the Abundance Institute, our AI (Generative) in Elections News Tracker (AI-GENT) has been tracking media coverage of AI and elections since January 21st, 2024. AI-GENT is built upon Mention, a subscription web application that crawls the web and searches for content that matches a user-defined criteria. Our tracker examines only articles that are published by web sites and newspapers based in the United States. The query criteria has evolved over the past four months as we iteratively narrow the query while experimenting to ensure we do not miss important news trends. Currently, a news article will get picked up by our tracker if the article contains at least one key word from each of the following bullet points: 

• Misinformation, false information, fake news, disinformation, deception, impersonation, deepfake, synthetic media, AI-manipulated media, fake, generative AI, AI-generated content, artificial intelligence, or robocall.

• 2024 election, presidential election, US election, election interference, voting integrity, election, democracy, Biden, or Trump. 

Using this methodology we have identified 7,573 articles related to AI-generated content being used in election-related information dissemination. The periodically updated dataset can be accessed here. 

Our tool also analyzes the sentiment of every article. Roughly 45% of the articles are neutral, 46% are negative, and 9% are positive. Preliminary sentiment correlation analysis shows that articles with a negative sentiment frequently use the word "deepfake” and articles with a positive sentiment are associated with use of the word "technology".

For this update, we used AI-GENT to identify publication trends. When the tracker indicated surging volumes of published content related to AI and elections, we examined media sources to identify what incident or incidents were driving the surge. Some surges were merely media discourse about the topic while others were about an actual incident. This required manual analysis to determine if the spike was a new incident, a past incident, general discussion of the issue or a related event such as the announcement of an initiative to combat AI-generated election material.

Identified Incidents

Our AI-GENT identified 7,573 articles. From these articles we have identified three prominent incidents of alleged election-related use of AI between January 21, 2024 and May 8, 2024. We also describe one high-profile incident that occurred before our tracker was operational. Below we describe the key facts of each.

The Biden Robocall in New Hampshire

First reported by NBC on January 22, 2024, the Biden Robocall story provided a concrete example of how recent advances in generative AI could propagate disinformation. In the lead up to New Hampshire’s Presidential Primary election, potential voters received a robocall that played an audio clip that sounded like Joe Biden encouraging New Hampshire residents to not vote in the state’s election. CNN estimated between 5,000 to 25,000 calls were made. Media coverage characterized the Biden robocall as an example of how generative AI is an ‘extraordinary threat’, although Reuters reported that the New Hampshire Secretary of State didn’t think that the robocall would have an impact on voter turnout. In a recent Senate Judiciary committee hearing, the New Hampshire Secretary of State said that New Hampshire had a record breaking turnout, making it difficult to assess the actual impact of the incident.

Further investigation revealed that the Biden robocall was perpetrated by Steve Kramer,  a Democrat political consultant. Kramer hired a street magician to create the fake audio of president Joe Biden. The street magician used Eleven Labs to generate the audio, which Kramer then distributed using robocalling services. Kramer claims his goal was to raise awareness that generative AI could be used to create disinformation, and hoped to provide Congress ammunition to pass AI regulation. 

Steve Kramer and his other co-conspirators are being sued by New Hampshire residents. The League of Women Voters, a nonpartisan organization, filed the lawsuit demanding injunctions and tens of thousands of dollars in damages.  

In this prominent instance, generative AI was used to create the fake audio but there were no reported negative effects.

Trump Supporters Spread AI-Generated Images of Donald Trump

The BBC first reported on March 4, 2024 that Trump supporters were distributing AI-generated images of Donald Trump posing with black voters. These images could not be tied to the official Trump campaign. Some of the distributing social media accounts have more than a million followers. A conservative radio and TV host in Florida self-reported that he used Midjourney to create the image of Trump with his arms around black women. At this time, there is no evidence of the effect of these photos on voters’ perception of Donald Trump, and understanding the effect is complicated by the unique electoral environment: according to a November 2023 poll by the New York Times and Siena College that predates the fake images, black voters were supporting Trump at levels that are previously unseen for a Republican Presidential candidate. 

Taylor Swift Holding “Trump Won” Flag

On January 31, 2024, a photo was posted on X (and subsequently removed) by Benny Johnson that showed Taylor Swift at the Super Bowl holding a flag that said, “Trump Won, Democrats Cheated.” In early February, a video was posted on X that showed Swift holding the same flag on the red carpet at the Grammys. NBC reporting on the Grammy Awards image noted that, “[o]ne post containing the video on X (formerly Twitter) has over 10.3 million views, according to X’s metrics.” X removed the posts of the video as violating X’s terms of service regarding manipulated media. 

Both instances of faked content received widespread mainstream media coverage and X community notes debunking their veracity. In both instances, an image of the “Trump Won” flag was inserted into a real image of Taylor Swift at the Super Bowl or a real video of Taylor Swift at the Grammys. Generative AI was not necessary to produce these fake images. Regular photo editing software was likely used in the case of the still image. It is possible that some kind of generative AI was used to create the Grammy Awards video. There is no evidence to date that these images changed any opinions about candidate Trump.

DeSantis Campaign Using AI-Generated Images in Video

Although this instance predates our tracker, it was a prominent enough example we considered it worthy of inclusion. On June 5, 2023, DeSantis War Room, a communications arm of the former Ron DeSantis presidential campaign, posted a video on X that showed real clips and audio of President Trump explaining why he didn’t fire Anthony Fauci. Interspersed in the video (between 00:24 and 00:30) was an image collage that included both real and apparently AI-generated images of President Trump hugging and showing affection to Fauci. The video did not note it included AI-generated or manipulated images. According to NPR reporting, the fact-checking organization AFP detected the fake images two days after they were posted. The post received a community note on X. The story was widely covered by major news organizations. The three fake collage images have distinct characteristics of being AI generated using a tool like DALL-E or Midjourney. However, it is not known which system was used. On January 21, 2024 DeSantis dropped out of the presidential campaign. 

Conclusion

180 days from the national presidential election, generative AI has not yet transformed the U.S. electoral process. None of the four incidents created any identifiable material impact. In three of these incidents it appears generative AI systems were used to produce the material. In the other incident, media coverage implied the use of new AI systems; however the content could easily be produced without AI.

No one can precisely predict how people will use generative AI technologies to create election-related content. It is also difficult to determine how such content would impact the 2024 election. Early media coverage of an event can leave a lasting impression even if later investigations undermine that impression. The Abundance Institute will continue to track usage of generative AI in the context of the 2024 election, running down all of the reported facts around any particular incident. This factual grounding can be useful to avoid sensationalist perspectives and to understand the legitimate impact that generative AI is having on the U.S. electoral process. 

Invitation to Collaborate

If there are any stories we missed or you would like to help our efforts, please fill out this Google Form. If there are search terms we should consider including please let us know. As the November election approaches in the U.S. we would especially appreciate your help in sending us examples of AI use in election material and examples of media stories that cover AI in elections. Our updates and tracking are primarily focused on U.S. elections although international stories are helpful for context. We welcome your help!