Artificial intelligence and elections, the case of voting in Argentina

In a year the world will witness the election of the President of the United States of America. A vote that could be historic in several respects: in a race that – net of the former tycoon’s judicial problems – seems destined to pit Donald Trump and Joe Biden once again, whoever emerges as the winner will be the oldest president ever elected in Washington . And above all, the 2024 race could be the first to be marked by the use of artificial intelligence in political communication. To get an idea of ​​what could happen, it is enough to stay in America, moving south to Argentina, where the presidential elections will take place on Sunday. Sergio Massa and Javier Milei are the two contenders who will face each other in the vote to choose the new president of Argentina. And the Argentine elections, wrote the New York Times, “have quickly become a testing ground for AI in electoral campaigns” Massa and Milei are in fact “using artificial intelligence to create images and videos to promote themselves and attack the other”. And while “much of the content was clearly false,” some appears to fall more into the realm of disinformation.

Elections in Argentina and Artificial Intelligence

The electoral race, which has also attracted international attention for some of Milei’s impactful media releases, is also among the first in which the candidates have made extensive use of artificial intelligence. According to the NYT, Massa’s campaign has used AI to portray him in contexts that show him as a soldier at war, such as Indiana Jones, and even in posters reminiscent of Barack Obama in 2008.

At the same time, opponent Javier Milei seems to have used artificial intelligence to show himself as a lion in cartoon style. But AI has also been used to attack the opponent: one case, cited by the Brazilian version of CNN, is that of a video from Massa’s electoral campaign. The content seeks to relate the opponent to Margaret Thatcher, British Prime Minister at the time of the Falklands War, a defeat still burning in the public imagination of Argentina. And Milei seems to have done the same, publishing an image of her opponent’s portrait on

The use of AI, however, was not limited to portraying the contenders in a positive or negative way: it also led to the candidates saying sentences that they had never actually uttered. For example, in a video Milei was made to say sentences, which he never actually uttered, about the plan to make the sale of organs legal. A deepfake from which, the NYT reports, Sergio Massa distanced himself and which was subsequently removed. An example which, however, highlights the risks that lurk in the extensive use of artificial intelligence in electoral campaigns.

Artificial intelligence and disinformation

Researchers have long been concerned about the impact AI could have on elections, the New York Times points out. In particular, the content thus generated could confuse voters and fuel doubts about what is real or not, as well as broadening the impact of disinformation that already exists online. On the other hand, however, there are also those who think that the impact of AI in the field of misinformation risks being overestimated: this is the case of a recent publication – signed by Felix Simon of the Oxford Internet Institute, Sacha Altay of University of Zurich and Hugo Mercier of the Jean Nicod Institute – who reported that these fears could be at least partly exaggerated. What seems certain is that, in September, on the eve of the vote in Slovakia for the renewal of Parliament, deepfake videos with the voices of politicians were released on social media: in one of these, Bloomberg reports, the leader of a party discussing vote buying with a journalist. The video was analyzed by AFP fact-checkers, who with the help of several experts came to the conclusion that it was a deepfake, created by an artificial intelligence trained to reproduce the politician’s voice through the use of real audio files. A case, Bloomberg further underlines, which seems to show how these technologies can be exploited to deceive voters in the days preceding particularly close elections.

Meta and electoral advertising

The use of AI – and its possible impact – on elections is therefore a topic that is increasingly at the center of attention. Meta has announced that it will require election ads on Facebook and Instagram to disclose whether they use artificial intelligence. Meanwhile, in recent months the first cases of the use of artificial intelligence in the US electoral campaign have emerged: among others, in April, following Joe Biden’s announcement that he wanted to run for a second term, the Republican Party posted on YouTube an AI-generated video showing hypothetical catastrophic scenarios that could come true if the current president wins the 2024 elections.

AI4TRUST: fighting disinformation also thanks to AI

On the other hand, it is true that artificial intelligence itself represents a very useful tool for combating mis/disinformation. This increasingly crucial theme is also at the center of AI4TRUST’s work: the project, which involves 17 partners (including SkyTG24) from 11 countries, is financed by the Horizon Europe program of the European Union and aims to develop a platform against disinformation that combines the contribution of artificial intelligence with the checks of journalists and fact-checkers. The objective is precisely to combat online disinformation, exploiting the potential that artificial intelligence can make available to media professionals.