Ministry of the Interior and Safety and National Forensic Service Secure 5 Additional AI Models
Misinformation Detection Rate Improves to 92% Compared to Previous Models
Collaboration with Relevant Agencies to Strengthen Response to Digital Crime

The government will begin actively deploying artificial intelligence (AI) detection technology in the field to counter deepfake manipulation crimes targeting elections.

Yoon Hojung, Minister of the Ministry of the Interior and Safety, is delivering a welcome speech on the afternoon of the 10th at the Government Seoul Building in Jongno-gu, Seoul, ahead of a demonstration of the AI deepfake detection analysis model.

Yoon Hojung, Minister of the Ministry of the Interior and Safety, is delivering a welcome speech on the afternoon of the 10th at the Government Seoul Building in Jongno-gu, Seoul, ahead of a demonstration of the AI deepfake detection analysis model.

View original image

The Ministry of the Interior and Safety and the National Forensic Service announced on March 10 that they will provide the five top-performing models, selected through last year’s "AI Detection Model Competition for Deepfake Countermeasures," to the National Election Commission.


On this day, Yoon Hojung, Minister of the Interior and Safety, described fake news using deepfakes as "a new type of information crime that threatens democracy," and stressed, "Protecting voters’ rights and their ability to make informed decisions from distorted information is a key responsibility of the government."


Recently, the advancement of generative AI technology has made it possible to create sophisticated deepfakes that manipulate individuals' faces or voices, making this type of crime a growing social issue.


According to the National Election Commission, the number of requests to delete deepfake videos during election periods surged from 388 cases during the 2024 general election to 10,510 during last year’s presidential election.


The deepfake detection model previously held by the National Forensic Service was limited in its ability to respond to advanced deepfakes, as its detection range was focused solely on facial areas.

AI Deepfake Detection Analysis Model. Provided by the Ministry of the Interior and Safety.

AI Deepfake Detection Analysis Model. Provided by the Ministry of the Interior and Safety.

View original image

By securing five additional top-performing models through last year’s private-sector competition, which saw participation from 1,077 people, the agency has now acquired global detection technology capable of analyzing entire videos—including backgrounds and bodies—beyond just localized facial analysis.


The Ministry of the Interior and Safety explained that these new models have a misinformation detection rate of 92%, a significant improvement over the previous model’s 76%.


After logging in to the National Forensic Service’s deepfake detection service and uploading a video, image, or audio file suspected of deepfake manipulation, the system determines its authenticity.


At the demonstration, after uploading a fake election campaign video and activating the video and audio detection models, the result "Fake 99.99%" was returned. Results are produced using a majority voting method across the different models.


Scores are given from 0 to 100, where a score closer to 0 indicates a higher probability of the content being authentic, and a score closer to 100 indicates a higher probability of it being a deepfake.


Jo Bonggi, Director of Investigation at the National Election Commission, stated, "If voters discover a deepfake video, they can report it to the Election Commission. We are also operating 400 monitoring agents."


He added, "The Election Commission immediately requests deletion from the relevant platform or party upon discovering deepfakes in order to minimize their influence on voters' decisions."


According to the Public Official Election Act, producing, distributing, or posting deepfake videos for election campaigning from 90 days before the election until election day is punishable by up to seven years in prison or a fine between 10 million and 50 million won.


The Ministry of the Interior and Safety and the National Forensic Service plan to gradually expand the scope of AI deepfake analysis model applications. They will strengthen digital crime response by cooperating with relevant agencies, such as the Ministry of Gender Equality and Family, the Korea Communications Commission, and the National Police Agency.



Minister Yoon stated, "When images or videos suspected of being deepfakes are detected, scientific analysis will be used to quickly verify whether they have been manipulated, and all relevant agencies will respond together in a timely manner."


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

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing