evidoria

← Back to browse

Good practice Imported

South Korea's National Election Commission — AI Deepfake Detection and Enforcement Ahead of the 2024 Elections

South Korea · Gwacheon · See the South Korea profile

Evidence: Observational / pre–post Top 6% 87/100 · Ask Evidence Copilot about this practice

Ahead of April 2024 elections, South Korea banned AI-generated deepfakes in campaigning within 90 days of polling. The National Election Commission identified 129 violations, backed by a jointly built AI detection model, with penalties up to 7 years in prison.

129 items
AI-generated deepfakes identified violating election law (29 Jan-16 Feb 2024)
7 years
Maximum prison sentence for campaign deepfake violation
10-50 million won
Fine range for campaign deepfake violation
South Korea's National Election Commission — AI Deepfake Detection and Enforcement Ahead of the 2024 Elections

Details

Maturity
Established
Promoter
National Election Commission (NEC), with the Ministry of the Interior and Safety and the National Forensic Service
Period
2024–ongoing
Keywords
elections, public safety, AI detection, law enforcement

Context

Ahead of South Korea's April 2024 legislative elections, lawmakers amended the Public Official Election Act to criminalise producing or distributing AI-generated deepfake content for campaign purposes within 90 days of polling day, with penalties of up to seven years in prison or fines of 10-50 million won.

Objectives

The National Election Commission (NEC) and partner agencies aimed to detect and enforce against AI-generated deepfakes that could distort the election before voters went to the polls.

Activities

Enforcement was distributed across government: the Ministry of the Interior and Safety and the National Forensic Service jointly built an AI deepfake detection and analysis model for investigators; the National Police Agency targeted websites spreading flagged content; the Ministry of Justice used digital forensics to trace distribution networks; and the Ministry of Science and ICT funded further detection and blocking technology. Major technology platforms signed voluntary pledges to combat election-related deepfakes.

Results

The NEC identified 129 AI-generated deepfake videos and images violating election law between 29 January and 16 February 2024 alone. Independent analysis (East Asia Forum, The Diplomat) credits the coordinated legal-and-technical response with limiting deepfakes' practical impact on the election outcome.

Conclusions

Observers note the 90-day pre-election window leaves campaign-adjacent deepfake activity outside that window largely unaddressed, and that detection accuracy against rapidly improving generative models remains an open challenge.

Implementation

Indicative cost
High (€500k–€5M) — Multi-agency staffing and detection-model development funded by the Ministry of Science and ICT; not separately costed in sources.
Time to results
Medium (1–3 years) — Enforcement concentrated in the run-up to the April 2024 legislative election, ongoing as standing law thereafter.
Staffing & skills
National Election Commission (NEC) investigators, Ministry of the Interior and Safety, National Forensic Service, National Police Agency, Ministry of Justice, Ministry of Science and ICT

Conditions for success

  • Statutory amendment to the Public Official Election Act creating clear criminal penalties
  • Cross-agency coordination combining legal, forensic and technical detection functions
  • Voluntary pledges from major technology platforms to combat election deepfakes

Common failure modes

  • The 90-day pre-election enforcement window leaves campaign-adjacent deepfake activity outside that window largely unaddressed
  • Detection accuracy against rapidly improving generative models remains an open challenge

Where it fits

Governance type
national government, multi-agency
Scale
national election-cycle enforcement
Income level
high income

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful