This is an almost usual type of content for American Internet users. On Monday, August 12, on the social network X, a user published two unpublished photos of the Democratic candidate for the American vice presidency, Tim Walz, in response to a popularity poll. Surprise: standing in front of an audience, the governor of Minnesota is dressed only in his underwear, greeting the crowd with a big smile.
Perfectly smooth skin, inconsistent finger size, radically different weight… These two images were generated by generative artificial intelligence (AI), as noted by RFI. Although the deception seems obvious, this publication has more than two million views on the social network.
Just one example in recent weeks: a video of President Joe Biden swearing, a photo of former President Donald Trump being arrested, etc. AI-manipulated content is on the rise as the US presidential election in November approaches. This phenomenon is fueling a wave of disinformation and sowing doubt among American voters.
Content omnipresent in the campaign
While the example of Tim Walz’s fake photos may seem harmless, other AI tricks have had much more resonance. In another video Published on X in late July, outgoing President Joe Biden appears to insult his detractors, including by using anti-LGBT slurs, shortly after announcing his withdrawal from the presidential race.
Featuring the logo of American public television PBS, and as the country enters an unprecedented and uncertain period, the sequence quickly goes viral, recalls the AFP fact-checking service : Shared by a “conservative talk show host Dave Rubin”, the video was viewed more than thirteen million times on X. To the point of forcing PBS to issue an official deniala few hours later.
Fake images and videos are omnipresent, to the point of pushing some to cry wolf even when this is not the case. On Sunday, August 11, former US President Donald Trump accused on its Truth Social network Democratic candidate Kamala Harris of “cheating” in a campaign photo. According to him, the vice president’s team had added “a massive crowd of so-called supporters” at the foot of her plane as it arrived at the Detroit airport (Michigan). However, newspapers and news agencies quickly confirmed these cheers by publishing new images and videos of the event, as the daily newspaper analyses New York Times.
Web giants singled out
The cause, according to the American NGO Center for Combating Digital Hate (CCDH): an extremely simplified generation of fake content on artificial intelligence platforms. Thus, although the image generation platform Midjourney claims to have blocked any attempt at fake content related to American presidents, tests carried out by the CCDH have made it possible to create fake images of Joe Biden arrested or Donald Trump next to a double.
Social media is also being blamed. Last week, X boss Elon Musk faced a barrage of criticism after sharing another AI manipulated video (or “deepfake”): this one shows Kamala Harris calling Joe Biden senile and not knowing how to “run the country”, thanks to an artificially generated voice imitating her own. A publication distributed to her 192 million subscribers without specifying its initial parodic nature. Still online, this video illustrates in particular the decline in moderation standards on X since its acquisition by Elon Musk.
So, in a letter addressed at the beginning of April To social media CEOs, more than 200 groups or organizations have called for urgent efforts to strengthen the fight against AI-generated disinformation, including banning “deepfakes” in campaign clips and highlighting verified election content. While several tech giants have said they are working on systems to automatically label all AI-generated content, the American NGO Free Press says it has “seen little substance” in the platforms’ commitments this election year.
AI is already creating doubt about the election results
And this phenomenon of disinformation is likely to increase as the presidential election draws closer, according to some experts. While written disinformation remains “a concern,” explains Lucas Hansen, co-founder of the nonprofit organization CivAI, “the most likely use will be doctored images and videos, with the aim of generating anger and increasing political tension,” he believes.
In the case of Kamala Harris’ plane, it could even be a deliberate strategy on the part of Donald Trump. In a editorial published Mondayjournalist Philip Bump of Washington Post believes that “Trump and his allies are eager from the start to question the reliability of Harris’s support numbers – and, by extension, the reliability of the November results.”
A climate of tension and distrust that is already creating doubt among voters. According to a Axios and Morning Consult poll published in late 2023, more than half of Americans believe that fake news generated by AI will impact the outcome of the presidential election. And about a third of Americans say they will trust the final results less because of this technology. For Nora Benavidez, a director at the NGO Free Press, “this is a tipping point in our election.”