Seoul Launches AI System to Detect and Disrupt Online Grooming of Minors
SEOUL – The Seoul Metropolitan Government will implement a new artificial intelligence (AI) system designed to proactively identify and intervene in online grooming attempts targeting minors, officials announced today. The system, which goes beyond simple keyword detection, utilizes a “multimodal, lightweight language model” – a small large Language Model - to analyze slang, abbreviations, and conversational context, according to a city official.
The AI will function by recognizing “risk signals” within online conversations and instantly alerting victim support agencies like the DasiHamkke Center. These centers can then directly engage in the conversation,offering guidance from professional counselors and assisting with investigations. Accounts exhibiting repeated grooming behavior will be reported to authorities.
This initiative builds on Seoul’s existing efforts to combat online sexual exploitation of children. In 2023,the city became the first in South Korea to deploy AI for the swift removal of digital sex crime videos. Further advancements in 2024 included the growth of AI-based facial age-estimation technology to identify and delete child and youth sexual exploitation material.
“Digital sex crimes continue to evolve, and sexual exploitation through online grooming has been on the rise in recent years,” stated Ma Chae-sook, Deputy Mayor for the Women and family Office at the Seoul Metropolitan Government. “Because many sex crimes against minors stem from grooming, the city will strengthen efforts to prevent such crimes in advance.”
A public forum to raise awareness about online grooming and discuss response measures will be held Tuesday at City Hall’s Seosomun Building 1. Presentations will be given by Kim Joon-cheol, a senior researcher at the Seoul Institute, on the AI-based detection technology, and Kim Bo-hwa, a senior researcher at the Seoul Foundation of Women and Family, on the current state of online grooming in seoul and related policy directions.
