Police to Post AI Analysts Nationwide to Catch Deepfakes
Is that video real or fake? Korean police now plan to settle it with AI. On the 16th the National Investigation Headquarters unveiled a basic plan for precise detection of AI-manipulated content, with AI expert analysts in every provincial police agency.
Every AI-assisted crime is a target
Deepfake sexual crime, defamation, investment fraud and voice phishing are all covered. The goal is to spot fakes quickly and analyse the whole chain from creation to distribution.
The plan has eight detailed tasks spanning technology, systems, personnel and organisation.
Checking lips against sound
Police have used a deepfake detection system since 2024 and are training it on Korean-specific data. They are also researching "multimodal" detection that checks whether lip movements match speech and whether audio matches subtitles.
Once the 9.1 billion won, 2025–2027 authenticity-checking system is built, every investigative agency will share it.
Tracing back to how it was made
Police are developing techniques that analyse program execution records, command logs and altered files to identify the program that made a fake. Manuals for major generative AI services are being distributed.
A 10 billion won, 2026–2029 project will build forensics for locally run AI models, meaning evidence extraction even when the fake came from an AI running on someone's own computer.
The numbers so far
Police said 419 suspects were arrested for deepfake crimes in January–July this year, 17 of them detained. The detection software has been used in 1,636 investigations since it was developed (M Economy News, 13 August).
The new plan adds AI expert analysts and a public-private-academic expert group, and sharply expands training for field investigators and analysts.
What the police said
Acting Police Commissioner Kim Jong-chul said, "As criminal methods using AI change fast, the police response must change too."
The plan also covers standard procedures and an expert group to share the latest cases and analysis results.
Sources & context
2026-10-16T22:30:20.940Z
Korea Policy Briefing · National Police Agency AI-manipulation detection plan (2026-10-16) ↗M Economy News · deepfake three-layer response report (2026-08-13) ↗Photo credit
Gohsuke Takama (Wikimedia Commons) · CC BY 2.0The Korean National Police Agency building in Seodaemun-gu, Seoul (Nov 2006). An archive photo, not a scene from this announcement. Gohsuke Takama / CC BY 2.0.