South Korea Orders Forums to AI-Scan Every Upload Starting July 1

A revised South Korean telecommunications law (전기통신사업법) takes effect July 1 requiring internet forums and online communities to run AI scans on every user-uploaded image and video. The government is not providing infrastructure, so operators must purchase datacenter-grade Nvidia GPUs at their own expense — a burden small forums and independent communities are unlikely to absorb.

The stated targets are non-consensual intimate imagery and CSAM, but the same AI filtering regime has been applied to messenger platforms like KakaoTalk since 2021, where it has flagged swimsuit photos, anime art, memes containing the word “sex,” and even linear algebra problems. South Korea has also been expanding the definition of CSAM to cover drawn and written fiction involving fictional minors, which broadens enforcement well beyond actual abuse material.

Critics argue the mandate either reflects technical illiteracy among lawmakers or a deliberate effort to shutter smaller, independent platforms that cannot afford compliance. With a recent controversy around memes of a deceased former president, observers also worry the infrastructure could quietly extend to political speech once the scanning pipeline is in place.