Ahead of the local elections on June 3, 2023, concerns are rising regarding the distribution of deepfake videos, which could mislead public opinion through social media. Such videos, created using artificial intelligence and deceptively realistic, resemble actual campaign footage, raising significant alarm among officials. In response, the South Korean government has decided to employ deepfake detection technologies at the polling stations to ensure the integrity of the electoral process.

On October 10, the Ministry of the Interior and Safety and the National Forensic Service held a demonstration of an 'AI Deepfake Detection and Analysis Model' at the Government Seoul Office. This event showcased the detection process, including a competitive project where private and academic experts developed detection models over a month last year, ultimately verifying five models for implementation in forensic systems. Despite plans revealing the development of a new detection model by January 2024, there were still limitations in identifying manipulated backgrounds or full-body images, highlighting the ongoing challenges faced by existing detection technologies in addressing sophisticated deepfakes.

The demonstration included analyzing five deepfake videos, including one of the fictional candidate Hong Gil-dong's campaign speech, using a total of six detection models (one existing and five newly developed). Each model evaluated the authenticity of the videos, with all four new models categorizing Hong's campaign video as 'fake', scoring between 73 to 100. This exercise underscores the urgent need for advanced AI detection solutions to combat misinformation and ensure fair electoral practices in the rapidly evolving digital landscape.