"AI Cops," an artificial intelligence system introduced to detect illegal pharmaceuticals and narcotics distributed online, is significantly underperforming compared to human monitoring agents, contrary to initial expectations at launch. Experts say surveillance coverage must be expanded and technology upgraded to counter the rapidly spreading online drug trade.
According to data obtained by the Seoul Economic Daily through Rep. Ahn Sang-hoon's office from the Ministry of Food and Drug Safety, a total of 17,441 online drug-related posts were detected during the three months from November last year, when the ministry deployed its self-developed AI Cops system, through February this year. Of these, AI detected only 5,044 posts, accounting for just 28.9% of the total. In contrast, human monitoring agents directly identified 12,389 posts.
Online drug distribution is growing rapidly. The total number of online drug trafficking busts reached 51,493 last year, an eight-fold increase from 6,167 cases in 2021 over four years. Cumulative detections over the past five years have exceeded 130,000.
As online drug distribution increases each year, calls are growing for AI to be used to preemptively block transaction attempts. However, the current AI Cops system is seen as failing to adequately reflect actual distribution environments, as the monitoring system currently operates primarily on select search engines, social networking services, open marketplaces, blogs, and secondhand trading platforms.
Major overseas platforms such as X (formerly Twitter) and Telegram, identified as "hotbeds for online drug trafficking," are excluded from surveillance. Looking at online drug bust cases in January this year, one in four was identified on X. Of the 51,493 total detections last year, 7,430 were from X alone, showing a concentration of transactions on specific social media platforms. Posts with hashtags suggesting methamphetamine or ketamine can be easily found on X. Many of these posts lure buyers before directing them to one-on-one Telegram conversations to complete transactions.
Officials on the ground say improvements are needed as AI Cops is still in its early stages. The Food Ministry explained it is continuously adding and adjusting online media under surveillance by analyzing the types and characteristics of sales posts during monitoring. The ministry also plans to use AI Cops to detect illegal online advertisements and request site blocking, while pursuing functional improvements including data updates.
"Some social media platforms have strengthened security policies, making information collection difficult, so we are reviewing improvement measures. Closed messaging services are structurally difficult to monitor with AI, so we are blocking them through cooperation with platform operators," a ministry official said.
Experts and officials agree that both operational methods and cooperation frameworks must be improved to enhance the effectiveness of AI deployment. Analysis suggests that without coordination with overseas platforms, meaningful enforcement results will be difficult to achieve.
"Online drug distribution is characterized by speed and rapid spread, so enforcement at the current level has limitations," Rep. Ahn said. "We must upgrade AI-based surveillance systems, strengthen cooperation with platform operators, and also reorganize coordination systems among relevant agencies."
