The problem is that using an AI censorship tool requires purchasing a solution from a specific vendor. And the deadline is effectively less than a month. There’s nothing particularly unusual about this—South Korea especially has many IT zombie companies that sustain themselves through government contracts. In practice, there’s a local CMS structure in place, and Korean programmers, who are generally weak in English, have to rely on that local CMS, which makes them weak in programming as well. (This is why, despite being a country with a high proportion of highly educated people, South Korea has relatively few prominent programmers.)

South Korea was the first country in the world to implement an internet censorship law. There is a historical record of censorship, regardless of which administration—left or right—was in power.

That said, it’s a complicated issue because these censorship systems also tend to create state IT contracts and job opportunities.

To make things more concrete: most local bulletin board systems and forum platforms are heavily tied to a specific commercial CMS. This is not a coincidence — government-affiliated projects often mandate that CMS, and developers here, lacking both English proficiency and exposure to global open-source alternatives, end up locked into its ecosystem. As a result, even basic AI censorship features become dependent on that vendor’s proprietary modules. When a tight deadline (less than a month) forces a purchase, there’s no room to explore better, cheaper, or more transparent options. The structure itself perpetuates vendor lock-in, weak technical capacity, and a cycle of superficial compliance rather than genuine innovation.