During the presidential election period last year, more than 10,000 illegal election materials that abused generative artificial intelligence (AI) technology·deepfakes, including sophisticated face or voice synthesis, were produced and distributed. The government has finished developing a model that can detect AI deepfake fake-news election materials and decided to apply it starting with the 6·3 local elections.

According to the Ministry of the Interior and Safety on the 10th, illegal election materials such as AI deepfake (illegally synthesized) videos·images detected online during recent election periods numbered 388 during the 22nd general election in 2024, but during the last presidential election they surged to 10,510.

The government has decided to put cutting-edge AI detection technology into full-scale field use to respond firmly to deepfake video·audio manipulation crimes targeting elections.

The ‘AI deepfake detection model’ jointly developed by the Ministry of the Interior and Safety and the National Forensic Service increases detection accuracy by simultaneously carrying out ‘global analysis’, which examines the overall flow of a video, and ‘local analysis’, which precisely identifies traces of manipulation in specific areas such as faces. The ministry explained, “It is designed to respond to the latest generative AI-based content, and actual validation recorded a high detection accuracy of about 92%.”

The ministry plans to apply the detection model to the 9th nationwide simultaneous local elections to be held on June 3. An official from the ministry said, “It is becoming possible to generate content whose authenticity is difficult to distinguish with the human eye, and deepfake crimes abusing this are becoming more sophisticated or advanced,” adding, “In particular, as concerns grow over the spread of false·manipulated information during election season, we plan to use the model in the local elections to prevent distortion of voter decision-making.”

The Ministry of the Interior and Safety and the National Forensic Service plan to gradually expand the scope of use of the AI deepfake analysis model. They will work closely with relevant ministries and agencies, including the Ministry of Gender Equality and Family, the Korea Communications Commission, and the National Police Agency, to establish a comprehensive, whole-of-government response system to digital crime. The National Forensic Service plans to advance the forensic science system to stay one step ahead of evolving AI-based crimes.

Yoon Ho-jung, Minister of the Interior and Safety, said, “Fake news using deepfakes is seriously threatening electoral democracy,” adding, “We will steadily strengthen our capacity to respond to false·manipulated information during the election process to ensure the precious rights of the public are not infringed, and build a competent and trusted ‘AI democratic government’.”