AI is coming to an election near you

A new player enters the political stage

By Amal Anzari

In 2024, some of the most significant geopolitical players of our time, such as the US, India, Indonesia, the European Union, Russia, and the UK, are set to witness pivotal elections that will play a critical role in determining the world’s trajectory. However, amidst the anticipation, a new player has emerged on the political stage — AI-generated content.

Since OpenAI unveiled ChatGPT in November 2022, several ethical concerns have cropped up. The capacity for generative AI tools to craft text that convincingly mirrors human expression has led to apprehensions regarding the integrity of our electoral systems. OpenAI itself has repeatedly highlighted the considerable threats posed by this technology, including its ability to “produce content that is nonsensical or untruthful,” as well as to “generate realistic and targeted content, including news articles, tweets, dialogue, and emails.”

Donald Trump recently shared a video of a Twitter spaces conversation between Ron DeSantis, Elon Musk, George Soros, Klaus Schwab, Dick Cheney, an FBI agent, Adolf Hitler, and the devil. It was fake, of course. But while this kind of Generative AI is more commonly used to make memes and joke videos, it’s also started creeping into real-world elections and politics. History teaches that impassioned speeches, persuasive books, informative leaflets, and social media affect our political outcomes by influencing voters. Generative AI can blow them all out of the water. 

A study published by the Proceedings of the National Academy of Sciences of the United States of America reveals that AI is “capable of creating faces that are indistinguishable—and more trustworthy—than real faces.” South Korea’s 2022 presidential election gives us an early example. During Yoon Suk Yeol’s campaign launch, the candidate introduced “AI Yoon,” a digital avatar of himself. South Korea allows real candidates’ AI “candidates” to campaign on their behalf — so long as it’s clear that it’s not the real person. While it is difficult to say whether AI Yoon helped Yoon Suk-yeol win, AI Yoon reached voters in a way the human candidate could not — by interacting with them on social media and online forums. It was also good press: AI Yoon was featured in media reports (even when the real Yoon was not).

In Canada in 2023, Toronto mayoral candidate Anthony Furey ran on what you could call an ‘AI platform.’ Although his campaign does not take a position on AI, his campaign appears to have used it to generate his platform. Not only does his campaign use many (quite obviously) AI-generated images, but a quick scan of the About page of his website reveals a 99.9% probability that these words were AI-generated too.

Turkey’s elections may be the most impacted by AI manipulation. According to École Polytechnique Fédérale de Lausanne, almost half of local Twitter trending topics in Turkey are fake, “a scale of manipulation previously unheard of.” Presidential candidate Muharrem İnce pulled out of the race, partly due to fake sex pictures of him being circulated. At the same time, incumbent President Erdoğan shared fake videos linking his main rival to a Kurdish militant organization. It was, effectively, a generative AI propaganda war.

Generative AI makes it easy for candidates to create campaigns without expensive strategists and videographers. A simple prompt can produce a hard-hitting appeal, social post, or slogan. Messaging on-demand will run into counter-messaging on demand, and elections then become bots versus bots — all battling for influence over the citizen audience. 

Back in the US, it’s still mostly memes, for now. Ron DeSantis’ campaign shared AI-generated images of former President Donald Trump hugging Anthony Fauci, his avowed enemy. In response to President Biden’s re-election announcement, the Republican National Committee posted a video depicting a bleak future, asking, “What if the weakest president we ever had was reelected?” The video then goes on to show us realistic-looking, albeit fake, images of our future dystopia were that to happen.

But it will not just be memes forever. Just as AI has a place in the South Korean government, Turkish propaganda wars, and in Canada’s mayoral race with the utilization of AI-generated images and potentially automated language, the American electorate should expect extremely convincing lies in the near future. Currently, no federal law in the US requires individuals or organizations to disclose that they used generative AI to manufacture videos or develop specific campaign appeals. Better policy could help, as the Federal Election Commission recently voted to regulate deepfake political ads. But there is no stopping what’s coming — a much more persuasive, insidious, and entertaining democracy.