Taylor Swift has not endorsed Donald Trump, nor have Lady Gaga or Morgan Freeman, and Bruce Springsteen has not been photographed wearing a “Keep America Trump Free” T-shirt. Yet these false reports have been shared on social media and are polluting the election campaign.
These so-called messages from American actors, singers and athletes revealing their political opinions and views on Kamala Harris or Donald Trump, often generated by artificial intelligence (AI), are multiplying, according to researchers.
Social media — like an X owned by Elon Musk that is breaking down barriers against disinformation — is allowing these messages to spread, and their potential influence on voters is worrying, just weeks before the election.
Last month, Donald Trump shared manipulated images showing the ultra-popular singer Taylor Swift in the United States, giving her support to the Republican. Some of these images bear the characteristics of images modified by AI, according to specialist Hany Farid, combining real and fake images, which makes the montage “particularly sneaky”, according to the expert.
Taylor Swift, who has hundreds of millions of followers on social media, actually endorsed Democrat Kamala Harris last week.
After the announcement, Donald Trump said on his Truth Social network: “I hate Taylor Swift!”
“Confusion and chaos”
A database from the News Literacy Project, a US NGO that works on disinformation, has so far listed 70 false messages from personalities expressing their support or rejection of a candidate — and the list is growing by the day.
“In this era of polarization, fake celebrity endorsements can capture voters’ attention, influence their views, confirm their biases and sow confusion and chaos,” warns Peter Adams, one of the NGO’s leaders.
Included in his list is a manipulated image of Lady Gaga holding a “Trump 2024” sign.
Others claim that actor Morgan Freeman, a Republican critic, said a second Trump term would be “good for the country,” or show rock singer Bruce Springsteen wearing a T-shirt that reads “Keep America Trump Free.”
“The platforms have made this possible,” Adams said. “By rolling back moderation and hesitating to remove false information, they have become a significant means for trolls, opportunists and propagandists to reach mass audiences.”
“Facilitator”
Especially X (ex-Twitter), after having revised downwards its content moderation policy and reinstated accounts spreading disinformation.
Its owner, Elon Musk, who supported Donald Trump in the presidential election, is regularly accused of sharing electoral lies.
After spreading false information, US officials overseeing the election asked Musk to fix X’s chatbot called Grok, which allows users to generate not only text but also images.
Lucas Hansen, co-founder of CivAI, which fights against the dangers of AI, showed AFP how easily Grok can generate a fake message. For example, simply write the command “image of an outdoor rally of women wearing “Swifties for Trump” T-shirts” to create a fake photo of Taylor Swift fans supporting the former president.
“If you want a fairly mundane situation where the people in the image are either famous or fictional, Grok is a great facilitator” for creating visual manipulation, Hansen said. “I expect it to be a major source of fake celebrity endorsement images.”
And as the technology develops, it will be “increasingly difficult to identify fakes,” says Jess Terry, who works as an analyst at Blackbird.AI.
“There is a very real risk that older generations or other groups of people who are less familiar with the development of AI-based technology will believe what they see,” says Jess Terry.