"War Against Fake Information"... Global Elections Shaken by AI Deepfakes

Generative AI Content Emerges in Elections Including Turkey
Next Year's US Presidential Election Expected to Face 'Great Chaos'... "A Watershed Moment for Campaigning"

The generative artificial intelligence (AI) craze that began with ChatGPT has triggered alarms worldwide in the election arena. In elections, where a single word, expression, or action can easily sway voter sentiment, the emergence of AI-generated fake information, namely deepfakes, has repeatedly had a profound impact on election dynamics, causing deep concerns among election authorities in various countries.

Especially in the United States, with the 2024 presidential election just over a year away, despite calls for regulation, AI is already being used in election campaigns, igniting an "AI election battle."

[Image source=Reuters Yonhap News]

◆ AI Fake Images Appear in T?rkiye and Canada Elections

In the Turkish presidential election, regarded as the most notable election in the first half of this year, AI emerged as one of the last-minute variables. This election saw Recep Tayyip Erdo?an, the long-term incumbent president of T?rkiye for 20 years and dubbed the "modern-day Sultan," locked in a close race with Kemal Kılı?daro?lu, the opposition Republican People's Party candidate, with even the possibility of Erdo?an's defeat being raised.

According to The Guardian and others, AI appeared two weeks before the runoff vote held in May. At that time, a video shown by President Erdo?an to his supporters became problematic. The video depicted the Turkish separatist group PKK singing a song supporting Erdo?an's opponent, Kılı?daro?lu. Erdo?an, who had emphasized a nationalist stance, provoked his supporters by saying, "Look at this. This is very important." The video was an AI-generated fake, and Kılı?daro?lu accused Russia of being behind it.

President Recep Tayyip Erdo?an of T?rkiye giving a speech in front of supporters last May [Image source=Yonhap News]

The result was a victory for President Erdo?an. The margin between him and Kılı?daro?lu was less than 5 percentage points. While various factors contributed to Erdo?an's presidential win, one was his use of AI-generated fake images to strengthen his nationalist line amid declining support and high inflation and economic difficulties.

In the Toronto mayoral by-election held on the 26th of last month, AI-generated fake images also drew attention. Toronto, the fourth largest city in North America, is one of the fastest and most active cities globally in AI research. More than 100 candidates ran in this mayoral election. Capturing voters' attention itself was a mission. Among them, Anthony Furey, a conservative-leaning former news columnist, gained attention for AI-generated election materials.

According to The New York Times and others, the 40-page material included several AI-generated images that contained many errors impossible in reality, causing a stir. In one image, a woman crossing her arms has three arms instead of two, with the other hand resting on her chin instead of crossing arms. In another photo, a long-haired woman wears a necklace, but the pendant is attached to the front hair rather than a gold chain. Additionally, badges on a cardigan worn by a woman and text on tape at a construction site were illegible.

Appearance of Anthony Fury, candidate in the Toronto, Canada mayoral by-election (Photo by Fury candidate SNS)

The New York Times noted, "Another candidate made fun of these images during a debate on the 16th of last month," but also assessed that "in a situation with over 100 candidates, such features acted as a driving force to make a name for oneself." Although Furey was not considered a major candidate early in the campaign and did not participate in TV debates, he ultimately ranked fourth and was recognized for building public awareness.

◆ "2024 Presidential Election Will Be an AI Election"…Republicans and Democrats in Dilemma

AI has already entered the global election battlefield. Now, the world's attention is turning to the U.S. presidential election scheduled for November next year, the "largest election in the world." With both incumbent President Joe Biden and former President Donald Trump having declared their candidacies, a fierce campaign is expected. Experts worry that AI could mimic candidates' voices to confess crimes they did not commit, produce racially discriminatory remarks, or create videos of speeches or interviews that never happened, resulting in various deepfakes.

Tom Newhouse, Vice President of Digital Marketing at Convergence Media, told the U.S. political media Axios, "If 2008 and 2012 were Facebook elections, this will be an AI election," adding, "But this time, it will be far more destructive." He predicted that a "October surprise" flood of AI-generated content could shake the election landscape just one month before the November election.

AI is already being used in U.S. political campaigns.

The Republican National Committee (RNC) released a 32-second video in April on the day President Biden officially announced his bid for re-election next year. The video included AI-generated images depicting a dystopian world expected if Biden's second term begins, which caused controversy.

Leading US Republican presidential candidate Ron DeSantis, Governor of Florida (left), and former President Donald Trump [Image source=AP Yonhap News]

AI fake images were also used within the Republican Party.

Ron DeSantis, Florida Governor and a leading Republican presidential contender competing against former President Trump, posted a composite photo on his Twitter account last month showing Trump hugging and kissing Dr. Anthony Fauci, former director of the National Institute of Allergy and Infectious Diseases, in the White House briefing room. This fake image targeted conservative voters dissatisfied with Fauci's pandemic measures.

DeSantis's team did not immediately clarify that the video containing this image was false but acknowledged it was a composite photo days later after media pointed out it was fake.

In response to these developments, over 70 Democratic election strategists held a virtual meeting on the 14th of last month. According to U.S. political media Politico, they discussed ways to counter AI-generated fake content in 2024. They expressed concerns about generative AI's ability to produce fake information at tremendous speed and scale and suggested educating voters on how to identify such content to make informed judgments, Politico reported.

The Democratic Party has been exploring AI use since early this year, including drafting fundraising emails with AI assistance.

Given the rapid spread and adoption of AI in U.S. politics, there are calls to establish guardrails and take measures accordingly. Senate Majority Leader Chuck Schumer unveiled the "SAFE Innovation Framework," an initiative for AI regulation, last month. He warned, "As early as next year's election, AI could completely undermine trust," adding, "We will soon live in a world where we see photos of Democratic or Republican candidates that are manipulated but fully trusted during political campaigns."

◆ "2024 U.S. Election Will Be a Mess"… "People Can't Distinguish Fake from Real"

While discussions on AI regulation are ongoing in the U.S., Europe, and Asia, experts generally believe that AI-related guardrails for elections will be difficult to implement ahead of next year's U.S. election.

Eric Schmidt, former Google chairman, appeared on a broadcast on the 26th of last month and said, "The 2024 election will be a mess because social media will fail to protect us from AI-generated misinformation," adding, "We are working to solve this problem but have not yet succeeded. In fact, trust and safety groups are shrinking."

He pointed out broad concerns about AI's long-term societal impact and diagnosed the short-term risk as "misinformation." He emphasized, "(Social media) should allow freedom of the press for humans, not computers. What social media must do is indicate who the user is for all content and hold them accountable if laws are violated," adding, "While it cannot resolve factual discrepancies, it at least establishes that the claimant is human."

Daryl M. West, senior fellow at the Brookings Institution, recently released a report on AI's impact on next year's election, stating, "Through easy-to-use and inexpensive templates, claims and counterclaims will continue, and election campaigns will face a situation similar to a 'lawless Wild West,'" predicting, "The ability to distinguish fake from real will be limited, and it will be uncertain how these claims will affect the election."

Paul Barrett, deputy director of the Center for Business and Human Rights at New York University's Stern School of Business, told Politico, "The public's belief that ordinary people won't create deepfakes has collapsed," diagnosing, "More people seem resigned to 'not knowing what is true.'" He also predicted that next year's U.S. election will be a major turning point in campaign methods.

This content was produced with the assistance of AI translation services.