The Dawn of a New Era: Artificial Intelligence Takes Charge in Landmark Election
In Argentina, Presidential Hopefuls Clash Using AI-Generated Imagery and Videos in a High-Stakes Political Battle
By: Rooz Aliabadi, Ph.D.
The streets of my favorite city in South America — Buenos Aires, are decorated with unique posters. An Argentine presidential candidate, Sergio Massa, is shown prominently in a shirt adorned with military-like medals, gesturing towards a blue sky. Encircling him, many elderly individuals in plain attire, bearing solemn and somewhat distorted expressions, gaze toward him with expectation. This distinctive Soviet-inspired style was intentionally crafted, following explicit directions given to the illustrator (Generative AI).
The Massa campaign used an AI program with specific instructions to create a striking political poster. The directive was to depict Massa in a manner reminiscent of Soviet political propaganda, as seen in Gustav Klutsis’s work, portraying him as a determined leader. The environment was to be imbued with symbols of unity and strength, conveying an aura of authority and determination. In response, Javier Milei, Massa’s opponent in the upcoming runoff election, has circulated AI-generated images portraying Massa as a Chinese communist figure and himself as an endearing animated lion, drawing over 30 million views. This election in Argentina is rapidly evolving into a battleground for AI usage in political campaigns, with candidates and their supporters actively manipulating and creating images and videos through this very powerful technology.
In recent months, Generative AI technology has dramatically altered the landscape of political campaigning, making candidates appear to say things they never did, featuring them in popular movies and memes, and even designing campaign posters. This has sparked intense discussions about the authenticity of videos. The significant impact of AI in Argentina’s elections and the resulting political discourse highlight the technology’s increasing ubiquity. As AI becomes more powerful and accessible, it’s poised to become pivotal in democratic elections worldwide. This scenario is reminiscent of the advent of social media, presenting both exciting opportunities and unforeseen challenges in the political theater.
The Massa campaign has developed an AI system capable of generating images and videos featuring key election figures — including candidates, their running mates, and political allies — in numerous scenarios. This technology has been employed to depict Mr. Massa, Argentina’s composed center-left economy minister, in a more dynamic light. Through videos, he’s been portrayed as a valiant soldier, a Ghostbuster, and even Indiana Jones. Additionally, the campaign has created posters reminiscent of Barack Obama’s iconic 2008 ‘Hope’ poster and a style similar to a cover of The New Yorker.
The Massa campaign has also utilized its AI system to cast his opponent, Mr. Milei — a far-right libertarian economist and TV personality known for his fiery outbursts — in a negative light. This has involved placing him in movies like Clockwork Orange and Fear and Loathing in Las Vegas to depict him as unstable. While much of this AI-generated content is overtly fictional, it is essential to remember that some pieces blur the line with disinformation. For instance, the Massa campaign created a ‘deepfake’ video where Mr. Milei is shown discussing the concept of a human organ market, aligning with his libertarian ideologies. In this video, the AI-generated version of Mr. Milei chillingly speaks about viewing children as assets, not in the traditional sense, but for the economic value of their organs. This video was shared on the Massa campaign’s Instagram account dedicated to AI content, named ‘AI for the Homeland.
The accompanying caption for the AI-generated post read, “We asked an AI to assist Javier in explaining the organ-selling business, and this was the result.” In a TV interview, Mr. Massa expressed his initial astonishment at the capabilities of AI, remarking, “I wasn’t mentally prepared for the world I’m entering. It’s a significant challenge. We’re riding a horse whose tricks we’re still learning.”
Upon viewing the deepfake of Mr. Milei discussing human organs created by his campaign, presented by a journalist, Massa seemed troubled and disapproved of such use. His spokesperson later emphasized that the post was meant as satire and was marked as AI-generated. The campaign stated that its AI usage aims to entertain and make political points, not to mislead. But this is very very very dangerous behavior.
I personally have been concerned about AI’s impact on elections, given its potential to mislead and confuse voters, further fueling rampant disinformation on social networks. These apprehensions were previously theoretical, as the technology needed to create such convincing fakes was complex, costly, and rudimentary. Now, we’ve noticed a massive wave in readily available and increasingly powerful democratized tools, altering the political landscape around the world drastically.
This year, a Toronto mayoral candidate deployed AI to create bleak images of homelessness, suggesting this would be the city’s future without his leadership. In the U.S., the Republican Party released an AI-generated video portraying a hypothetical Chinese invasion of Taiwan and other dystopian scenarios, claiming these would occur if President Biden is re-elected.
Additionally, Governor Ron DeSantis of Florida’s campaign circulated a video with AI-generated imagery showing Donald J. Trump embracing Dr. Anthony S. Fauci, a figure criticized by many on the American right for his leadership during the pandemic.
In Argentina, the AI-generated materials shared by political campaigns have been either explicitly marked as such or are so evidently artificial that they’re unlikely to mislead even the most gullible voters, at least for now. Rather than deceiving, this technology has enhanced the capacity to quickly produce viral content, which traditionally would have required extensive time and effort from graphic design teams.
Meta, the parent company of Facebook and Instagram, recently announced that political AI advertisements must disclose this fact. However, other non-paid posts on these platforms, even if politically related and AI-generated, are not mandated to have such disclosures. The U.S. Federal Election Commission is also deliberating on regulations for AI use in political ads.
The Massa campaign employed AI to modernize the image of Peronism, a 78-year-old political movement, aiming to resonate with younger voters by blending Mr. Massa’s image with elements of popular and meme culture. To achieve this, campaign engineers and artists used open-source software Stable Diffusion to train their AI system with photos of Argentine political figures, enabling the creation of synthetic images of these individuals. This system allows the rapid production of images or videos featuring prominent Argentine politicians in various scenarios.
Throughout the campaign, Mr. Massa’s communications team has guided artists in using the campaign’s AI to convey specific messages or emotions, such as national unity, family values, and fear. The artists brainstorm ideas to feature Mr. Massa, Mr. Milei, and other political figures in content that references movies, memes, artistic styles, or historical events.
For Halloween, the Massa campaign instructed its AI to produce whimsical images depicting Mr. Milei and his associates as zombies. Additionally, the campaign leveraged AI to craft an intense movie trailer, showcasing Buenos Aires in flames, with Mr. Milei portrayed as a diabolical villain in a straitjacket and Mr. Massa as the nation's savior.
These AI-generated images have also made their way into physical spaces. The Soviet-themed posters are just one of many designs that Mr. Massa’s campaign and his supporters have printed and displayed throughout public areas in Argentina.
Some of these images were directly produced by the campaign’s AI, while others were created by supporters using AI technology. Among the most notable is an image of Mr. Massa on horseback, styled after José de San Martín, a revered figure in Argentine independence history.
Octavio Tome, a community organizer involved in crafting this image, commented on the approach, stating, “Massa appeared too rigid. We aimed to present him as a leader with a strong Argentine character.”
The increasing use of AI in Argentina’s elections has led to growing skepticism among voters about the authenticity of what they see. This was evident when a video surfaced showing Mr. Massa looking fatigued after a campaign event, leading to accusations of drug use by his detractors. In retaliation, his supporters claimed the video was a deepfake. However, the campaign later confirmed that the video was indeed genuine.
Mr. Massa pointed out that AI obscures past errors or scandals. He noted in an interview, “It’s now quite simple to blame artificial intelligence when something you said is revealed unintentionally.”
Pivotal Year Ahead for Elections Around the World
The upcoming year is poised to be pivotal for elections globally, with major nations including the U.S., India, the U.K., Mexico, Indonesia, and Taiwan, all gearing up for the polls. While there is widespread concern about the potential influence of deepfakes and misinformation on these elections, experts note that, to date, the actual impact of these factors on electoral outcomes remains minimal. They acknowledge the steps technology companies took to safeguard election integrity but emphasize that more profound reforms in political systems are necessary to effectively address the issue of misinformation.
Many suggest that the impact of misinformation is often overestimated because people assume it’s easy to alter deeply held beliefs, like voting preferences, and overrate the capabilities of technologies like AI in spreading misinformation.
Although AI-generated misinformation currently has a low probability of significantly swaying public opinion in elections, its future implications could be more troubling. I foresee potential challenges with highly personalized AI-driven targeting and convincing AI agents, which might be used in large-scale persuasion campaigns. Furthermore, the mere presence of deepfakes in 2024 might undermine trust in essential information sources.
These concerns are not purely hypothetical. In 2019, rumors that a video of Gabon’s President Ali Bongo was a deepfake fueled an attempted coup. Similarly, the impact of a controversial video allegedly involving a Malaysian cabinet minister was diminished by claims that it was altered by AI.
This is an often-missed risk that seems more probable in this election cycle. It’s not so much that generative AI will create deepfakes deceiving people into believing a candidate has done something terrible. My genuine concern is that the mere existence of these technologies could be leveraged to devalue evidence and erode trust in crucial information sources.
This article was written by Rooz Aliabadi, Ph.D. (rooz@readyai.org). Rooz is the CEO (Chief Troublemaker) at ReadyAI.org
To learn more about ReadyAI, visit www.readyai.org or email us at info@readyai.org.
