A corner of an exhibition showcasing AI technology products in the central city of Đà Nẵng.— VNA/VNS Photo Quốc Dũng

HÀ NỘI — The Law on Artificial Intelligence, which took effect on March 1, establishes a unified legal framework for the development and deployment of AI in Việt Nam.

It imposes stringent requirements on implementation, oversight and the accountability of all parties involved in translating science and technology into practical application.

The new legislation devotes a dedicated section to building the AI ecosystem and mobilising resources, including provisions for the establishment of a National Artificial Intelligence Development Fund, a State financial fund operating outside the State budget on a not-for-profit basis.

Under the law, the fund will be set up by the Government to mobilise and allocate resources for the research, development, application and governance of AI in service of socio-economic development, national defence and security.

The fund will operate under a special financial mechanism that allows for risk-taking in scientific, technological and innovation activities.

Capital will be disbursed in line with project timelines and implementation requirements, and streamlined procedures may be applied to strategic or urgent tasks.

Management and use of the fund must ensure proper purpose, transparency and compliance with relevant legal provisions.

At a time when AI demands substantial resources in data, computing infrastructure and high-quality human capital, the introduction of a flexible financial mechanism is regarded as a key step in supporting priority AI research, development and deployment projects.

Outlining the Government’s vision for AI development, Minister of Science and Technology Nguyễn Mạnh Hùng said: “AI must become a national intellectual infrastructure, serving the people, promoting sustainable development and enhancing national competitiveness.”

Hùng added that AI is no longer merely an applied technology, but is evolving into a form of national infrastructure, akin to electricity, telecommunications or the Internet.

“We will pursue AI-isation in the same way we once pursued electrification, but we will move faster, as fast as possible,” he said.

He also said the drafting of the AI Law was guided by several core principles: risk-based regulation, transparency and accountability, a human-centred approach, encouragement of domestic AI development and technological self-reliance, positioning AI as a driver of rapid and sustainable growth and safeguarding digital sovereignty, with data, infrastructure and AI technology identified as three strategic pillars.

“AI is a powerful tool, but humans remain the ultimate decision-makers,” he said.

Human-centred approach

A central tenet of the law is its human-centred orientation. AI is defined as a tool to serve people, not to replace human authority or responsibility.

All AI systems must ensure the maintenance of human oversight and the ability to intervene in decisions and actions generated by such systems.

The law requires the protection of human rights, privacy, national interests, public interests and national security, as well as system safety, data security and information confidentiality.

It mandates inspectability and oversight throughout the development and operational lifecycle of AI systems.

Principles of fairness, transparency, non-bias, non-discrimination and non-harm are enshrined, alongside adherence to ethical standards, Vietnamese cultural values and accountability for decisions produced by AI systems.

One notable provision concerns the labelling of AI-generated content.

AI systems that interact directly with individuals must be designed and operated in a way that enables users to recognise they are engaging with an AI system, unless otherwise provided by law.

Audio, images and videos generated or modified by AI to simulate a real person’s appearance or voice, or to recreate real events, must carry clear and recognisable labels distinguishing them from authentic content.

For cinematic works, artistic productions or other creative products, labelling must be applied in an appropriate manner that does not hinder performance or audience experience.

In terms of risk governance, the law classifies AI systems into three categories: low, medium and high risk.

High-risk systems are those capable of causing significant harm to life, health or the lawful rights and interests of organisations and individuals, as well as to national interest, public interest or national security.

Medium-risk systems are those that may mislead, influence or manipulate users who are unaware that they are interacting with an AI system or AI-generated content.

All other systems are deemed low risk.

Classification is based on the extent of impact on human rights, safety and security; the field of application, particularly essential sectors or those directly affecting the public interest; the user base; and the scale of potential impact.

This approach is intended to ensure strict control over high-risk systems while preserving space for innovation in low-risk applications.

Under the law, prohibited acts are clearly identified, including the misuse of AI for deception or manipulation, exploitation of vulnerable groups and the creation or dissemination of fabricated content causing serious harm to national security, social order or public safety.

The unlawful collection, processing or use of data, obstruction of human oversight and control mechanisms and concealment of information required to be disclosed in AI operations are likewise prohibited.

In terms of legal liability, organisations and individuals in breach of the law may face administrative penalties, criminal prosecution and civil compensation, depending on the nature and severity of the violation.

The law is believed to place Việt Nam among a small number of jurisdictions to enact dedicated AI legislation.

Codifying such a foundational technology amid an accelerating digital transformation is seen as a significant step in strengthening the institutional framework for the digital economy.

Only three entities worldwide have adopted standalone AI laws so far: the European Union, South Korea and Japan.

Most other countries remain at the level of posting strategies, guidelines or dispersed regulatory provisions, while continuing to research and draft comprehensive legislation. — VNS