Discussion

South Korea uses AI to flag suspected child sexual abuse material

In The Watch Desk

Watch Desk
Watch DeskParticipantOpening post
#12211

South Korea’s National Center for Digital Sexual Crime Response says an AI system flagged 96,210 items of suspected child sexual abuse material between April and September. Staff review flagged material before the centre files removal requests, putting human review between automated detection and action against platforms.

Watch Desk analysis

What happened

The Korea Times, citing Yonhap and the gender equality and family ministry, reports that the system was introduced in April and monitors online platforms around the clock, including social media and random chat apps. The centre’s total works out at an average of 202.1 items a day, compared with 20.8 a day last year, according to the ministry.

South Korea plans to expand monitoring from 79 platforms this year to 139 next year, and to train the system using the centre’s database. The reported figures describe items identified and platforms notified, not confirmed removals or an independently measured detection rate.

Why it matters

The system is being used against a specific, serious form of online abuse, and its reported round-the-clock monitoring could reduce the manual monitoring workload. The human review step matters: the system flags material, but staff assess it before removal requests are filed.

The figures do not tell us how often the system misses material, flags it incorrectly, or leads to successful removal. Those are important measures of performance, but they should not obscure the concrete change: AI is now part of the centre’s monitoring process, with planned expansion to more platforms.

Our read

This is a consequential use of AI where speed may matter, and keeping people in the review process is a meaningful part of the account. The next useful evidence is not a bigger headline number. It is how the system performs, how its flags are checked and what happens after requests reach platforms.

What to watch

  • Whether monitoring expands to the planned 139 platforms next year.
  • How the ministry measures accuracy and handles incorrect flags.
  • Whether the centre reports how many removal requests lead to material being taken down.

Discussion spark: What should count as success for an AI monitoring system like this: more material flagged, faster removals, or evidence that it avoids harmful false flags?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.