AI-generated fake flood videos cause public confusion in South Korea
AI-generated flood videos circulating on YouTube and other platforms in South Korea confused viewers by depicting realistic but false scenes of flooding, including Gyeongbokgung Palace and subway stations. Broadcasters mistook some AI videos for real footage, and deepfake-related police reports have increased significantly since 2021.
What happened
Since Google released Veo3 in May 2025, AI-generated videos have proliferated on platforms like YouTube, with over 40 million created in the first two months. Videos depicting flooding scenarios in Seoul, including a flooded Gyeongbokgung Palace with a seal swimming through it, circulated online and fooled viewers who thought they were real.
Some South Korean broadcasters mistook AI-generated video of sparrows attacking love bugs for real footage and reported it as a genuine pest-control event. A video of a female environmental activist purportedly cursing at love bugs was widely shared but later confirmed to be AI-generated.
Reported impact
Online users expressed alarm at the realistic appearance of the fake videos, with comments such as 'Soon it will be hard to tell real from fake.' A police official stated that 'Recently, crimes are being committed with sophisticated manipulated videos that are incomparable to the past.'
Deepfake-related police reports increased from 156 in 2021 to 964 in 2024. A professor warned that 'uncontrolled proliferation of AI videos could foster social distrust and conflict, leading to community destruction.'
Index analysis
MBC's use of AI video in its 'Surprise' program to recreate historical scenes prompted YouTube comments expressing concern about job losses for actors, makeup artists, and production staff.
Governance controls
South Korea's AI Basic Law will mandate watermarking of AI-generated content starting January 22, 2026, though experts question the effectiveness of watermarks that can be easily removed.
A professor at Korea University suggested that AI model creators providing technology to identify AI-generated content would be a more effective approach without imposing social costs. Some experts argue that excessive regulation could hinder domestic AI technology development.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.