Inside AI-powered fight against digital sex crimes

The language of sexual predators grooming potential victims may look harmless. But spotting their patterns is necessary for protecting underage internet users.

“A teenager posts a short video and receives comments such as ‘You’re pretty’ or ‘How old are you?’ These may seem like harmless compliments or signs of interest. But when they are used to build trust and steer the child toward private contact, they can mark the beginning of online grooming,” an official from the National Center for Digital Sexual Crime Response said during a demonstration for reporters Tuesday.

Detecting predators’ pattern is the next challenge for the center, where artificial intelligence (AI) has already improved its capacity to detect suspected child sexual exploitation material and streamlined the process of requesting its removal.

“Because the individual words are not inherently suspicious, keyword searches alone cannot detect the pattern,” the official said. “That is why we’re now developing technology that looks beyond isolated words to examine how an exchange unfolds and identify attempts to draw young people into sexual exploitation.”

The center’s current AI system collects posts around the clock using about 220 keywords drawn from frontline monitoring. It analyzes text and images together, then presents its assessment and supporting reasons to staff for review.

Once a human staffer approves a report, the system prepares the information required by the relevant platform or authority, streamlining the submission process. It now takes about a minute or less to process each takedown request.

In recent years, the center has increasingly automated its monitoring and reporting procedures, allowing staff to focus on tasks that require human judgment.

According to the center, AI adoption has increased the average number of cases processed per worker each day from 20.8 in 2025 to 202.1, a roughly 9.7-fold rise. Between April and September this year, it processed 96,210 cases — more than twice last year’s total of 45,438.

The technology also helps protect staff from repeated exposure to disturbing images, which can take a toll on their mental health. Images are now blurred by default, allowing workers to reveal them only when necessary to make a reporting decision.

“From detecting whether material has been circulated to submitting a takedown request, the system handles the process mostly automatically, without requiring human intervention at every step,” said Kim Mi-soon, the center’s director.

“AI has not only accelerated the process but is also detecting deepfakes that are difficult to identify with the naked eye.”

The center plans to expand the system’s coverage from 14 platforms to 139 by 2027. Alongside contextual detection of potential grooming, it is also developing an AI chatbot to automatically send vulnerable young people information about support services.

Kim hopes these advances will help curb digital sex crimes, which disproportionately affect teenagers and people in their 20s among those seeking the center’s support. According to its data, the center supported 10,637 victims in 2025, up from 4,973 in 2020. Teenagers and people in their 20s accounted for 77.6 percent of those supported last year.

But faster detection and reporting do not guarantee removal. The center said 28.5 percent of its takedown requests went unfulfilled in 2025. Among websites whose server locations it could identify, at least 95.5 percent were hosted overseas, complicating enforcement.

Officials said technical improvements must be accompanied by international cooperation, stronger platform accountability and efforts to disrupt the revenue sustaining illegal sites.

Park Kwang-soo, head of the Women’s Human Rights Institute of Korea, which oversees the center, said the technological gains reflected the expertise and dedication of staff members determined to eradicate digital sex crimes.

“Whatever AI or technology we use, victim support must begin and end with people,” Park said. “And people are needed at countless points in between.”