South Korea mandates AI image scanning by 2026. Platforms must cover all server costs, potentially burdening smaller operators amid global regulatory shifts.
From mid-2026, every picture uploaded to a forum or social network in South Korea will be scanned by artificial intelligence for illegal content — and the companies running those platforms must foot the server bill themselves. Experts warn the financial strain could be ruinous for smaller communities.
The mandate, confirmed by the Korea Communications Commission on June 4 during a technical briefing for affected firms, takes effect on July 1, 2026. After a six-month transition period, roughly 80 platform operators — among them Google, Meta, and local giants Naver and Kakao — must have the system fully integrated.
Each uploaded image will be checked in real time against a government-maintained reference database designed to flag illegal recordings. The law’s origins trace back to the so-called N-th Room scandal of 2020, a mass sexual exploitation case that shocked the nation. In October 2025, South Korea’s Constitutional Court rejected complaints challenging the expanded surveillance duties.
Operators carry the full cost of the computing power needed for the AI scans. Smaller platforms, which often lack the revenue streams of global tech companies, are expected to struggle most. No state subsidy or cost-sharing mechanism has been announced.
On the same day as the commission’s briefing, the opposition Democratic Party introduced a separate bill targeting harmful speech. It would impose fines of up to three percent of a platform’s annual revenue, or even shutdown orders, for failing to curb the spread of hateful misinformation.
The South Korean push is part of a wider global tightening. In Europe, a labeling requirement for AI-generated content — including deepfakes — takes effect on August 2, 2026. Responsibility for missing labels falls on publishers and distributing companies. Market data cited in the article indicates that between 40 and 75 percent of people can no longer distinguish AI-generated images from real photographs.
Other jurisdictions are moving in parallel. In June, the U.S. government issued an executive order establishing a voluntary 30-day pre-review for advanced AI models by the National Security Agency. Austria’s Justice Minister Anna Sporrer announced new liability rules on June 5: site operators are only responsible for defamatory comments after receiving a take-down notice. A recent Supreme Court ruling also clarified that merely “liking” an offensive comment does not automatically constitute defamation, often reflecting only unspecified antipathy instead.
In Germany, coalition parties in Saxony — the CDU, SPD, and BSW — agreed on June 5 to revise the state police law, permitting AI-assisted video surveillance and facial recognition searches on the internet.
A report published on June 5 underscores the urgency of robust filtering: 92 percent of tested image generators can produce fake identity documents. Cybercrime losses in the United States alone reached roughly €20 billion in 2025. Industry representatives are pressing for stronger identity-verification measures and tighter monitoring of digital content.
Leave a Reply