[AI Infused in the General Election] ① Election Year, Deepfake Targeting Swing Voters

Ban on Deepfake Election Campaigns from the 29th
Widespread Side Effects if Misused in Public Elections
Up to 7 Years Imprisonment or Fines for Violations
US Issues AI Regulation Executive Order
EU Discusses Measures to Prevent Technology Misuse

The deepfake war?manipulation technology that is fake but looks real?is shaking up the election scene. The moment the boundary between what is real and fake becomes blurred, fake content transforms into reality, clouding voters' judgment and leading to irreversible choices. Deepfakes are so sophisticated that even experts find it difficult to determine whether content has been manipulated, and if used in public elections, the side effects could escalate exponentially.

In South Korea, a representative case was during the 2022 presidential election when the People Power Party's candidate Yoon Seok-yeol promoted a public-friendly image through 'AI Yoon Seok-yeol.' Although no major cases raising awareness have emerged domestically since then, overseas, deepfakes have recently become one of the biggest variables in important elections. In the previous U.S. presidential election, deepfake videos showing then-candidate Joe Biden confusing states during campaign rallies or dozing off during interviews were circulated mainly among supporters of then-candidate Donald Trump, causing controversy. Recently, in Slovakia, just days before the election, an AI-generated audio recording fabricated to make it seem as if a liberal candidate discussed beer price hikes and election rigging plans spread as if it were true, sparking public uproar. On the 26th of last month (local time), the Associated Press predicted, "This year's U.S. presidential election will be the first election held since the emergence of sophisticated AI tools that can create fakes in just a few clicks and seconds."

Public Official Election Act Regulating Deepfakes Passes National Assembly

Deepfake is a neologism combining 'Deep' from deep learning and 'Fake,' meaning false. It began to be used in 2017 when a Reddit community member with the username 'Deepfakes' posted manipulated videos. Jo Won-yong, a professor at the Central Election Commission's Election Training Institute, warned, "In a paper published last year in the Korean Public Law Association's 'Public Law Research,' deepfake technology is advancing day by day," adding, "If abused in public elections, it could become a grave crime shaking the foundation of a democratic republic."

Due to concerns about the side effects of deepfakes, legislative responses have begun both domestically and internationally. In South Korea, ahead of the general election on April 10, minimal legal measures to prevent the misuse of deepfakes have been established. On the 20th of last month, an amendment to the 'Public Official Election Act' regulating the indiscriminate use of deepfake videos during election campaigns passed the plenary session of the National Assembly. Nam In-soon, chairperson of the National Assembly's Special Committee on Political Reform (Democratic Party), stated, "The act of spreading false information using deepfake technology that synthesizes images and videos to look real can distort voters' judgment. This bill aims to regulate the indiscriminate use of AI technology during election periods."

Accordingly, from 90 days before the election until election day, it is prohibited to produce, edit, distribute, screen, or post deepfake videos for election campaigning. The law will take effect from the 29th of this month. An official from the Election Commission explained, "In principle, it should apply from the 11th, but a supplementary provision in the law sets the enforcement date as the 29th." Violations may result in imprisonment of up to seven years or fines ranging from 10 million to 50 million won. Experts regard this as the first legislative step against the misuse of deepfakes in public elections, but academia views it as just the 'beginning of the discussion.' As early as 2017, Oxford University demonstrated that a 'fake' video could be produced with just one photo and one voice file of the target individual. Now, anyone can become a target of deepfake technology with only a single photo, and if abused in public elections?from presidential to general and local elections?the impact could spread exponentially, highlighting the need for active political interest.

The Wall Street Journal mentioned that deepfakes could have an even greater impact on moderates or swing voters?those undecided about whom to vote for. Since swing voters pay attention to deepfakes and may change their preferences if the content seems plausible, Jo said, "In situations where a politician's preference is unclear, the appearance of a persuasive deepfake can shake voters' perceptions."

It is especially difficult to verify the truth when actual candidates claim deepfakes to cover up their wrongdoings or gaffes. With AI technology advancing, even experts take time to determine authenticity, and if voters cast ballots after watching deepfake videos just before elections, it could lead to irreversible choices, which is a major concern. In situations where public sentiment is hard to gauge, the sudden proliferation of deepfakes before elections could become a variable that disrupts the election landscape.

U.S. and EU Take Action... Need for Response to Advanced Technology

Governments worldwide have also taken action. In October last year, U.S. President Biden signed an 'Executive Order on Safe and Trustworthy AI.' The order includes regulations addressing AI threats across various fields, from personal information to U.S. national security. President Biden said, "I've seen my own deepfake," adding, "I even asked myself, 'When on earth did I say that?'"

Choi Jin-woong, a legislative researcher at the National Assembly Research Service, said, "In the U.S., some state laws define the scope and standards of illegal deepfake election campaigns," adding, "In the European Union (EU), discussions are ongoing on general legal regulations to prevent the misuse of deepfake technology rather than specific regulations on deepfake election campaigns. More detailed discussions on the legitimate use of deepfake election campaigns are needed going forward." Awareness of deepfake videos remains low domestically. However, as deepfake technology advances, it could negatively impact not only elections but also the development of democracy in our society, underscoring the need for serious reflection and countermeasures.