Hanoi (VNA) – While many countries are still drafting policy blueprints or issuing non-binding guidance, Vietnam has moved ahead with a standalone Law on Artificial Intelligence (AI), effective from March 1, placing it among a select group of nations to adopt dedicated AI legislation at the parliamentary level.
Risk-based, human-centred governance
Minister of Science and Technology Nguyen Manh Hung emphasised that the law is grounded in a clear principle: AI serves as a support tool, and ultimate responsibility in critical social matters must remain with humans. This human-in-command approach shaped the legislative process.
The law introduces a three-tier risk classification system for AI applications – high, medium and low. Providers must assess and categorise their systems before deployment, in line with forthcoming technical guidance.
High-risk systems are subject to stricter safeguards. Developers must incorporate mechanisms enabling human oversight and intervention, while maintaining comprehensive technical documentation and operational logs for compliance reviews and post-market inspections. Users are required to adhere strictly to operational protocols and safety measures, and are barred from unauthorised modifications. Oversight and inspection mechanisms will be proportionate to the risk level.
To support enforcement, the Ministry of Science and Technology (MoST) is preparing a Prime Ministerial decision to issue an official list of high-risk AI systems.
Five criteria will guide the identification of such systems: impact on human rights and legitimate interests; implications for safety, security and public welfare; sector and usage context; degree of automation and extent of human control; and scale of application and potential impact. The draft list groups high-risk systems into four broad categories, covering those affecting human rights, public safety and security, essential public-interest sectors, and systems with wide-reaching or difficult-to-remedy consequences.
Transparency and safeguards for AI-generated content
The legislation also tightens transparency requirements. AI systems that interact directly with users must be clearly identifiable as such, except where otherwise stipulated by law.
Providers are required to embed machine-readable markers in AI-generated audio, images and videos in accordance with Government regulations. Content that replicates a real person’s voice or appearance, or simulates real-world events, must carry clear and visible labels distinguishing it from authentic material.
A dedicated chapter focuses on infrastructure development and safeguarding national AI sovereignty. Organisations and individuals engaged in research and mastery of core AI technologies will be eligible for preferential policies and targeted support.
AI enterprises are entitled to the highest level of incentives under current regulations, including improved access to infrastructure, datasets and regulatory sandbox environments to facilitate research, production and commercialisation. Small and medium-sized enterprises will receive assistance in funding, training and market access. Domestically developed AI products and services will be prioritised in public procurement under existing bidding laws.
The law further mandates the development of AI talent through integrated education pathways. Foundational AI knowledge, computational thinking, digital skills and technology ethics will be embedded in compulsory curricula, while vocational and higher education institutions are encouraged to expand AI and data science programmes and deepen partnerships with businesses, research bodies and international organisations.
Decision No. 367/QD-TTg, dated March 3, 2026 approves the implementation plan for the law, setting out detailed tasks, timelines and responsibilities to ensure timely, coordinated and effective enforcement, while raising awareness and strengthening accountability across ministries and local authorities.
Seven key task groups are outlined, including communications and training on the law and its guiding documents from 2026 onward; a review of implementing provisions under the Law on Digital Technology Industry; and a comprehensive audit of related legal instruments, with findings to be consolidated by the MoST and reported to the Prime Minister.
In the 2026–2027 period, MoST will establish a national AI development fund and build AI innovation clusters that combine concentrated physical hubs with interconnected digital networks.
Taken together, the new AI Law and its implementation roadmap not only complete Vietnam’s legal architecture for AI, but also lay a structured foundation for risk management, innovation and ecosystem development, aiming to ensure coherent and effective enforcement in the years ahead./.
VNA
