AI Images of Trump with Black Voters Spread Before Election Campaign
The 2024 US Presidential Election is the first major American election since the explosion of artificial intelligence tools – their use in campaigning, advertising and narrative-forming is already evident. With few case studies currently available, there remains significant dispute between professionals, experts, and academics as to the implications of this technology for democratic participation. One point of agreement, however, is that voters must be careful of the increased risk of disinformation percolated using AI software. This was recently proven by the online success of a collection of AI-generated images purporting to show former US president and leading Republican nominee Donald Trump associating with African American people, in a bid to court black voters to vote for Trump in the next election.
These images – created and distributed by Trump supporters online – display the former president in various positive interactions with minority voters, such as one fake image showing him smiling with a group of black women at a party, or another showing him sitting with some young men, purporting that he “stopped his motorcade” to pose with them. Published on X, formerly Twitter, many of these posts attained millions of views, with users split on whether they were real or fake.
Nonetheless, there remained some telltale signs that these posts were, in fact, faked using generative software. Experts pointed out a series of issues with each photo, such as AI’s persistent problems with generating hands which leave some people with too many fingers, and some with too few. Furthermore, one Floridian conservative radio host, Mr Mark Kaye, admitted to BBC journalists that he created and disseminated one such viral pro-Trump image using AI tools.
The posts, published from a variety of sources over the prior weeks, appear to pre-empt a planned advertising push by the Trump campaign targeting black voters in Georgia, Michigan and Pennsylvania, key swing states that may prove vital in electing the next president. Trump has enjoyed increasing levels of support from the African American electorate, with the proportion of this demographic who would consider voting Republican increasing from 12% to 16% between 2020 and 2024, and it is clear that Republican strategists see this group as potentially important in securing Trump a second term as president. [1] In addition to concerns over the economy and the rising cost of living, many black voters hold the view that Trump is supportive of their community, a belief reinforced by these AI fakes.
However, this is not the first time that the Trump campaign and its supporters have been accused of utilising online misinformation to draw floating voters to their side. In the 2016 and 2020 elections, the Trump campaign faced a raft of misinformation accusations, ranging from false narratives that the 2020 presidential election was “stolen” to alleged attempts by foreign powers, such as Russia, to influence voter behaviour through the use of networks of bots and troll accounts sowing division and spreading pro-Trump talking points. If this issue persists, the use of AI to manipulate and strategically control narratives surrounding various issues during the election may become a defining threat to the US electoral process.
If a substantial proportion of voters are exposed to false AI-generated information online, made by provocateurs and partisans to inflame tensions and spread anonymously online, public knowledge of current affairs may be significantly compromised and electoral outcomes manipulated by a loud minority of malicious actors (ranging from homespun extremists to hostile foreign powers). It is increasingly clear that, in a regulatory climate labelled the “AI Wild West”, the recent growth of these fake pro-Trump posts may simply be the tip of a larger spear of misinformation, manipulation and confusion generated by the purposeful use of AI tools to influence democratic elections. [2]
Ultimately, what is needed is a significantly more rigorous system of watermarking, labelling, and identification on all AI-generated posts, emblazoning on each image whether it was genuinely taken or produced using AI tools. Whilst this is certainly an interim solution until a wider and more expansive regulatory framework for generative AI products is formed, it may deal with the primary concern associated with AI images: the average consumers’ struggle to distinguish between what is real and what is generated.
Every major social media company, such as X and Meta, have instituted their own policies to attempt to combat AI misinformation and deal with fake generated content during election seasons, but nonetheless the risk remains high that a wide selection of voters will be exposed to, and potentially influenced by, AI-generated political propaganda in subsequent elections. This is clearly unacceptable, and only time will tell if democratic systems, governments and regulators can handle the challenge presented by savvy propagandists wielding AI-generated misinformation in the future.
References
Poll: 20-point deficit on handling economy highlights Biden’s struggles against Trump (nbcnews.com)
The Wild, Wild West of Artificial Intelligence Is Coming | InvestorPlace
By James Barber
