Vietnam: Ministry of Science and Technology opened consultation on draft Decision promulgating the List of High-Risk Artificial Intelligence Systems

On 9 February 2026, the Ministry of Science and Technology opened consultation on the draft Decision promulgating the List of High-Risk Artificial Intelligence Systems. The proposal establishes the principles and criteria for identifying high-risk artificial intelligence (AI) systems that must undergo conformity certification before being deployed. Concerning human rights, systems used for large-scale classification of individuals are considered high-risk if they affect at least 1'000 individuals/month or 10'000/year, or when 30% or more of the system’s results directly determine the decision. In employment, an AI system is high-risk if it screens, evaluates, or terminates contracts for 500+ workers annually. Systems performing remote biometric identification in public spaces must receive conformity certification before use. AI used in essential infrastructure is high-risk if a malfunction could affect 10'000 users or span 2+ provincial-level units. AI systems used for safety assurance in high-risk products require prior certification. Systems with the ability to monitor on a large scale, evaluate security risks, or affect management or control activities for 1'000 or more subjects are high-risk. Systems monitoring essential infrastructure is high-risk when they may affect 10'000 subjects or more. AI systems used in medicine require prior certification. AI systems used in education are high-risk when they determine education access for 500+ learners per year. Systems used in finance are high-risk where the outcome directly affects the access to financial services of 1'000+ customers per year, or when the decision is automated by 30% or more. Systems are classified as high-risk when their outputs are used as a formal basis for administrative or judicial decisions that determine legal or human rights, regardless of scale. Systems deployed in 2+ administrative units, affecting 50'000+ people or processes personal data are high-risk. Finally, AI systems integrated into work requiring certification also require certification. This requirement applies regardless of scale, with certification mandated prior to use.