Skip to content
Submit incident
AII-1002-45679
Well documentedThree or more independent publishers report the core facts.

South Korean AI system for predictive policing and surveillance raises concerns

South Korean scientists developed an AI system called Dejaview that analyzes video footage to predict crimes in real time with reported accuracy over 82 percent, raising concerns about surveillance and potential bias in predictive policing. Seoul plans to deploy over 10,000 AI-equipped cameras by 2026, with a commercial version planned for late 2025.

20 September 2024 · Electronics and Telecommunications Research Institute · Privacy and surveillance
20 September 2024
AII-1002-456795 sources ↓
Quick facts
Category
Privacy and surveillance
Country
South Korea
Source
bTV Новините
Date
20 September 2024
Drafted by
Raidu Incident Detection

What happened

South Korean scientists at the Electronics and Telecommunications Research Institute (ETRI) developed an AI system called Dejaview that analyzes video footage to assess crime risks in real time.

The system learns from prior video footage, behavioral patterns and statistics, analyzes public behavior and compares it to criminal behavior. Dejaview can also differentiate between types of crimes.

The system can monitor individuals to analyze their behavior in depth to prevent recidivism. ETRI plans to release a commercial version by late 2025. Seoul plans to install over 10,000 high-resolution cameras with AI by 2026, in addition to over 160,000 existing security cameras.

Reported impact

According to experts, during testing the model predicted several crimes with accuracy over 82 percent.

Index analysis

The system's reliance on training data from prior video footage and behavioral statistics raises concerns about potential bias, as crime data may reflect historical enforcement patterns and socioeconomic disparities rather than actual crime occurrence.

Classification

Category
Privacy and surveillance
Industry
Government, security and defence
Country
South Korea
Harm status
Potential harm
Intent
Unintentional
Timing
Unknown
Harm types
Human or fundamental rights
Harmed
General public
Failure modes
Biased model
Consequences
None recorded
AI systems
Dejaview

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

bTV Новините
Създадоха изкуствен интелект, предсказващ престъпления
20 September 2024
dariknews.bg
Създаден е изкуствен интелект, предсказващ престъпления
20 September 2024
Novini.bg
Създадоха изкуствен интелект, предсказващ престъпления - Технологии - Новини Бг
20 September 2024
Telegraph.bg
КАТО ОТ ФИЛМ: Изкуствен интелект предсказва престъпления!
20 September 2024
Vesti.bg
Направиха изкуствен интелект, предсказващ престъпления
20 September 2024