- The guidelines provide detailed implementation measures for the transparency obligations under the AI Basic Act, which takes effect on January 22.

- The guidelines also clarify labeling requirements for AI-generated content, incorporating comments from industry stakeholders.

On January 22, the Ministry of Science and ICT (MSIT, Deputy Prime Minister and Minister: Bae Kyung-hoon) unveiled the Guidelines on Ensuring AI Transparency, which set out detailed implementation methods for Article 31 “Obligation to Ensure AI Transparency” of the Basic Act on the Development of Artificial Intelligence and the Establishment of a Trustworthy Foundation (“AI Basic Act”).

The AI Basic Act, enacted on January 21, 2025 to systematically promote the AI industry and lay a foundation for the safe use of AI technologies, enters into full force on January 22. However, as a grace period of at least one year will be granted, fact-finding investigations and imposition of penalties under the transparency provision will also be deferred during this period.

So far, the AI industry has raised concerns that the Act and its Enforcement Decree alone do not provide sufficient clarity on how the transparency obligation should be applied in practice. In response, MSIT released a draft of the guidelines in September 2025, gathered feedback from industry participants, and developed implementation standards based on the types of AI products and services currently available on the market.

These guidelines seek to balance the legislative goal of securing trust in AI with the need to ease burdens on businesses. In particular, for AI-manipulated content (deepfakes), which has generated significant social concern, the guidelines require clear labeling so that users can easily recognize such content as AI-generated. This measure aims to address public anxiety over the increasing misuse of AI technologies.

On the other hand, the guidelines allow more flexible labeling requirements by distinguishing between AI-generated content that remains within a service environment* and content that is exported or distributed outside the service environment**. These standards are intended to ensure that companies do not adopt excessive measures that could undermine the convenience and usability of their services.

* Refers to cases where AI-generated outputs are displayed only within the service environment, such as on screens, on user interfaces (UIs) or inside applications.

** Refers to cases where AI-generated outputs are exported outside the service environment through downloading and other means.

The guidelines clearly define “AI business operators” as those who directly provide AI products and services to users and are therefore subject to the transparency obligation. International companies that provide such products and services to users in Korea are also subject to the obligation.

However, the obligation does not apply to users who simply use AI technologies or services as tools for work or creative activities. This is because they are not providers of AI products or services, but rather users who make use of such products, services, and AI-generated content.

For example, a film production company that produces and distributes movies using AI video-generation tools is regarded as a user applying AI in its operations, rather than as a provider of AI products or services. Accordingly, it is not subject to the transparency obligation.

The obligation to ensure transparency is divided into two parts: (i) a prior notice requirement to inform users that products or services are operated based on high-impact or generative AI (Article 31(1)), and (ii) a labeling requirement indicating that AI-generated content has been created using generative AI (Article 31(2) and (3)).

The prior notice requirement (Article 31(1)) is intended to ensure that users are informed in advance that high-impact or generative AI technologies are being used in the products or services they intend to use, enabling them to exercise appropriate caution.

AI business operators must specify in their terms and conditions, contracts, or software and application interfaces that their products or services are based on high-impact or generative AI. In the case of offline services, operators may use appropriate notification methods depending on the nature of the service, for instance, by posting a notice in a location easily visible to users.

With respect to the labeling requirement (Article 31(2) and (3)), the guidelines establish differentiated standards by distinguishing between AI-generated content that remains within a service environment and content that is exported or distributed outside the service, taking into account technological and practical considerations.

When AI-generated content is provided solely within a service environment, operators may apply flexible labeling methods through UIs or visual indicators (e.g., logos). For example, in the case of conversational AI services such as chatbots, operators may provide prior notice before use or display visual indicators on the screen. In games or virtual convergence environments (metaverse), the guidelines permit notifications at the login stage or indications that characters are generated by AI.

By contrast, when AI-generated outputs are exported outside the service environment, more explicit labeling is required. When AI-generated texts, images, or videos are downloaded or shared, they must be labeled (i) using methods recognizable to humans such as visible or audible identifiers (e.g., watermarks), or (ii) by embedding machine-readable methods (e.g., metadata) into the content after providing a text or voice notification.

In particular, for AI-generated content that is difficult to distinguish from reality — such as AI-manipulated videos (deepfakes) — the guidelines require labeling in a manner that is clearly recognizable to users, in order to prevent confusion.

MSIT stated that “the application of watermarks to AI-generated content is a minimum safeguard to prevent the misuse of AI technologies, such as AI-manipulated content (deepfakes), and reflects a global trend already adopted by leading international companies.” The ministry further added “Given concerns over insufficient preparedness in the early stage of implementation, we will grant a sufficient grace period of at least one year and closely communicate with industry stakeholders to ensure the smooth implementation of the system.”

MSIT also plans to continuously refine and enhance the guidelines by collecting feedback and addressing challenges faced by businesses, while reflecting newly emerging services and technological characteristics.

The Guidelines on Ensuring AI Transparency are available on the websites of the Ministry of Science and ICT (https://www.msit.go.kr), and the Telecommunications Technology Association (https://www.tta.or.kr).

For further information, please contact the Public Relations Division (Phone: +82-44-202-4034, E-mail: msitmedia@korea.kr) of the Ministry of Science and ICT.

Please refer to the attached PDF.