With 6 Weeks Left in Local Elections, Deepfakes Run Rampant [Kim Hyun-ji's with AI]

As the June 3rd local elections approach, South Korea finds itself in a deepening battle against the proliferation of sophisticated deepfake technology. Despite regulations enacted in December 2023, which prohibit AI-generated election videos 90 days before the election, the problem persists. The National Forensic Service (NFS) recently identified a video of a candidate, seemingly indistinguishable from the real person in appearance, gestures, and voice, as an AI-generated fake. This incident underscores the growing challenge of discerning truth from fabrication in political discourse.

“

Simply manipulating photos or videos is no longer the case; videos generated from realistic photos now have AI voices embedded, or synthesized voices are added over actual music in the form of campaign songs.

The sophistication of these deepfakes is escalating. Park Nam-in, an industrial researcher at the NFS's digital division, notes that creators are moving beyond simple photo or video manipulation. They are now embedding AI-generated voices into realistic videos and even creating campaign songs with synthesized voices. This blending of real and fake content makes it increasingly difficult for the average citizen to verify authenticity using only their senses. The methods of distribution further complicate matters; while original files shared via email might be easier to analyze, videos uploaded to platforms like YouTube and then re-shared through messengers can have their metadata corrupted or deleted, rendering deepfake detection a formidable task.

From a South Korean perspective, this technological arms race poses a significant threat to the integrity of our democratic processes. While law enforcement and government agencies are committed to a strict response, relying solely on post-regulation and punishment is proving insufficient against the rapid pace of technological evolution. As Park emphasizes, fostering the public's ability to discern false information is paramount. This requires a proactive approach, encouraging citizens to actively verify suspicious content rather than passively consuming it.

“

These mixed real and fake contents make it even more difficult for the general public to distinguish between truth and falsehood with just their eyes and ears, requiring special caution.

We at Dong-A Ilbo urge our readers to adopt vigilant verification habits. Simple yet effective methods, such as using Google Image Search to check for original sources of suspicious images or carefully observing subtle discrepancies in videos like unnatural blinking or slight audio-visual sync issues, can make a difference. By integrating these 'reverse search' practices into our daily information consumption, we can collectively combat the spread of misinformation and safeguard the electoral process. The challenge is immense, but through heightened awareness and shared responsibility, we can strive to maintain a more informed and trustworthy political environment.

“

Citizens must also develop the capacity to discern false information.