What responsibilities bind AI system operators in the August 2025 Draft Guidelines?

The Draft Concept of Artificial Intelligence (AI) Ethics Guidelines issued by the Ministry of Communications. Furthermore, Digital (public consultation draft, August 2025) defines an AI system operator as any party that implements. Furthermore, Operates an AI system within a specific user context. The draft does not provide an isolated list of operator duties. instead, it frames obligations collectively for "Sector Actors." Among the fully legible protective safeguard clauses, only one names operators directly. The text is not yet enacted law and remains subject to revision.

Who qualifies as an AI system operator

The draft defines an operator as any party that implements. Furthermore, Operates an Artificial Intelligence system within a specific user context, such as in public services, industrial sectors, or specific institutions (Definition point 8). This role is distinguished from developers who train models (point 6) and system providers who integrate models into applications (point 7). Together with data providers, these roles constitute Sector Actors (point 3).

A clinic or hospital utilizing a vendor-supplied AI documentation or screening tool within its healthcare operations represents a specific institution acting as an AI system operator.

General obligations and operational safeguards

Chapter VIII point 2 establishes four core Sector Actor duties: developing and operating systems in accordance with ethical principles. managing risks across all lifecycle phases. mitigating risks through prevention, early detection, remediation, periodic audits, and continuous monitoring. and guaranteeing system transparency and accountability.

Table 1 (Chapter VII.2) details key operational safeguards:

• explaining system workflows, underlying data, and operational boundaries to end users;
• designating and documenting specific personnel accountable for system adoption and operational decisions;
• maintaining internal and external grievance and incident reporting channels;
• establishing emergency procedures to halt system operations quickly and safely. and
• ensuring meaningful human oversight without automated final decisions in matters significantly affecting fundamental rights, human life, or health.

Among the fully legible safeguard items, one clause specifically addresses operators alongside developers: both must maintain adequate, continuously updated technical and ethical competence.

High-risk markers and grievance resolution

The high-risk classification criteria identify negative impacts on security, safety, health. Alternatively, Human rights, characterized by biometric data usage, personal identifiable information (PII), and the direct impact resulting from public or community use of the system.

Sector Actors must complete self-assessments provided by sector oversight authorities (Chapter IX.1). When end users submit complaints, authorities mediate the report. The Sector Actor must trace root causes, apply system-wide fixes if the issue is widespread or individualized corrective measures if personal. In addition, Document and report the resolution to the supervisory authority (Chapter IX.2).

The draft proposes a two-year preparation timeframe following formal enactment (Chapter X). As an administrative consultation draft, it does not currently impose binding statutory obligations.

Legal basis

• Draft Concept of Artificial Intelligence Ethics Guidelines, Ministry of Communications. Furthermore, Digital, August 2025 (not yet in force): Definition points 2, 3, 6, 7, and 8. Chapter I. Chapter VII.2 Table 1. Chapter VIII point 2. Chapter IX.1 and IX.2. Chapters X and XI.