Korea AI Basic Act prior notice: tell users the product uses AI

MSIT's guidelines on AI transparency set two duties for AI business operators: give users prior notice that a product or service uses high-impact or generative AI, and mark AI-generated outputs. The notice can be given through terms, contracts or visible notifications (MSIT guidelines, read 2026-10-02). Most coverage focuses on labels; the notice is the half that lives in your sign-up flow.

The two duties side by side

The guidelines separate telling people a service uses generative AI from labeling what it produces. Different methods apply depending on whether output stays inside the service or leaves it.

What prior notice looks like in an API product

If you build an app on top of Sume's public API and offer it to people in Korea, you may be the business that provides the generative AI service to them. The practical question is where the notice sits: sign-up terms, a visible line on the generation screen, or both.

The page lists terms and contracts as acceptable carriers, so an accurate clause in your terms of service is not enough if the product screen shows an output that looks like a recorded human. Many teams do both: a sentence in the terms and a short visible notice next to the generate button.

A short plan

• Write one plain sentence saying the product uses generative AI to make video, images or audio.
• Place it in the terms and in the interface where people start a generation.
• Mark exported files with a visible or machine-readable label, using the stricter method for anything that leaves your service.
• Record who in your company owns the notice and when it last changed.

What Sume does and does not decide

Sume's docs do not state whether a given customer is a provider or a user under this law, and Sume does not insert your notice for you. The guidelines are a summary on a government page, and enforcement timing is covered separately; read the Korean text with counsel for anything binding.