South Korean police are building a next-generation detection system that analyzes video, audio and text together to counter crimes that exploit artificial intelligence, including deepfakes. Beyond identifying AI-generated fabrications, the force will also upgrade its digital forensics capabilities to trace the entire criminal process, from creation to distribution.
The National Office of Investigation at the Korean National Police Agency said on the 15th that it has established and begun implementing a basic plan to upgrade its system for detecting and analyzing illegal AI-generated content.
The measure aims to strengthen investigative agencies' detection and analysis capabilities as the criminal misuse of AI spreads beyond deepfake sex crimes into defamation, investment fraud and voice phishing. Police plan to pursue eight detailed tasks across four areas: technology, systems, personnel and organization.
At the core is the adoption of "multimodal" technology, moving away from existing detection methods centered on video and images to analyze audio and text as well. The approach determines whether content has been manipulated by comprehensively analyzing factors such as whether a person's mouth movements match their voice in a video, and whether the audio matches the subtitles.
Police plan to secure the relevant technology through a research and development project worth a total of 9.1 billion won ($6.5 million) from 2025 to 2027, and to build a system that all investigative agencies can use jointly in the future.
Since 2024, police have adopted and operated a deepfake detection system used overseas. Work is also under way to improve detection performance for the domestic environment by applying training data that reflects the characteristics of Korean nationals.
Forensic technology for tracing how AI-manipulated content is produced and distributed will also be strengthened. Police will develop investigative techniques to analyze program execution histories and command logs from major generative AI services and to identify the programs used in manipulation. They also plan to develop analytical tools that automatically extract digital evidence throughout the criminal process, from the creation of illegal content to its distribution. A research and development project worth a total of 10 billion won ($7.1 million) will be carried out from 2026 to 2029 for this purpose.
Police will deploy and operate "AI specialist analysts" dedicated to analyzing manipulated content at provincial and metropolitan police agencies nationwide, and will expand specialized training for field investigators and analysts. An expert group drawn from the private sector, government and academia will be formed to share the latest trends in AI technology, major detection cases and analyses of new programs.
"We will build advanced investigative capabilities that can not only quickly identify fakes created with AI, but also effectively trace how they were produced and distributed," said Kim Jong-cheol, acting commissioner general of the Korean National Police Agency.