by Alessio Jacona*
The faces are relaxed and serene, in some cases smiling.
The clear glances turned towards the camera, as happens when you are among friends and take a photo to remember a beautiful moment.
In the center, there is former president Donald Trump, surrounded by young members of the African American community, but unfortunately it is yet another deep fake, a false image created with generative artificial intelligence that portrays a meeting that does not exist never been with the sole purpose of gaining consensus - and therefore votes - for the Republican candidate.
An attempt to deceive and manipulate the black electorate who, when Joe Biden won the presidential elections in 2020, was crucial to the success of the Democrats in the so-called "swing" states.
Disinformation “from below”
There are many images like these online: they were discovered by the BBC Panorama broadcast, which denounced the emergence of new misinformation in view of the US presidential elections in November.
And which, in doing so, has also highlighted a relatively new phenomenon, but one destined to grow exponentially: something we could call "bottom-up disinformation generated with AI", that is, by American citizens themselves.
FACT CHECK
Trump supporters are courting black voters using AI-generated images hoping to encourage African Americans to vote Republican.
FACT: THE PHOTOS ARE ALL FAKE
FACTS MATTER!
pic.twitter.com/OTZnAlvxYF
— PoliticsVideoChannel (@politvidchannel) March 4, 2024
Yes, because while the BBC journalists were unable to link the deepfakes in question in any way either to Trump's electoral campaign or to attempts at interference by foreign governments (as happened with the 2016 elections), they instead managed easily track down the creators of some of the deepfakes among ordinary US citizens.
Or rather, among conservative influencers with a large following on social media, who have long been conducting an incessant campaign against the Democrats and President Biden on their accounts;
and who now seem to increasingly appreciate the use of generative AI tools to fabricate deepfakes.
Alternative-facts powered by AI
In some ways, it is merely the natural evolution of what has been happening since the first half of the 1910s: that is, since the American Right transformed itself into the "alternative right", that is, into that part that recognizes itself in an ideology more conservative and reactionary, which rejects traditional politics and uses digital channels to spread false or at least controversial information, defined as "alternative facts".
A practice that over the years has been central to communication strategies and functional to Trump's success (to the point of costing him expulsion from Facebook and the former Twitter), and which today becomes more relevant than ever as tools for increasingly powerful generative artificial intelligence within everyone's reach, for which all it takes is a prompt to generate convincing texts, images and (soon also) photorealistic videos.
BBC Panorama intercepted two creators among those who spread false images of Trump with young African-Americans: one is the ultra-conservative Mark Kaye, host of a radio show in Florida and administrator of a Facebook page with over a million followers, who justified the choice to share a deepfake in question (now no longer available on his profile) simply by saying that he was not a photojournalist but a storyteller, that he had never said that the photo was real (but not even the opposite, ed.) and that if anyone thought it was, it's just their problem.
The other is a conservative US citizen - a certain Shaggy - who took an image portraying Trump among young people of color among the many produced with AI by the satirical account @Trump_History45, and reposted it, changing the caption.
The photo, accompanied by a text according to which the ex-Republican president would have stopped during a car trip to meet young African-Americans in the street, has totaled over one million and 300 thousand views on X.
It's only the beginning
In both cases, reading the comments it was possible to understand how many users mistook these images for real ones, and this despite the presence of several inaccuracies, including the usual wrong number of fingers.
One therefore wonders what will happen as these systems become (as they are becoming) increasingly precise and efficient in creating images, or when OpenaAI Sora's new generative AI, capable of creating incredible photorealistic videos from a simple prompt, is launched on the market.
Meanwhile, it appears increasingly clear that, even in the attempt to manipulate more or less large portions of the electorate, the strategy is to rely on a human-machine alliance, where false contents are generated by AI, but to spread (and therefore to make them credible) people of flesh and blood take care of them.
*Journalist, innovation expert and curator of the ANSA.it Artificial Intelligence Observatory
Reproduction reserved © Copyright ANSA