Vietnam's AI Law Explained: A Beginner's Guide to Southeast Asia's First AI Regulation

On 1 March 2026, Vietnam became the first country in Southeast Asia to enforce a comprehensive, standalone artificial intelligence law. Law No. 134/2025/QH15 — passed by the National Assembly on 10 December 2025 — spans eight chapters and 35 articles, establishing a risk-based framework that covers every stage of the AI lifecycle, from research and development through deployment and end-user interaction.

For businesses, developers, and policymakers watching Asia's fast-moving AI landscape, this law is more than a regulatory curiosity. It sets a precedent that could shape how the rest of ASEAN approaches AI governance^✦ over the next decade. Whether you are running AI-powered^✦ tools for a fintech startup in Ho Chi Minh City, or simply curious about how governments are catching up with the technology, this guide breaks down the essentials.

What the Law Actually Says

Risk-Based Classification

At the heart of the legislation sits a three-tier risk classification system — a structure that will feel familiar to anyone who has followed the EU AI Act. Article 9 sets out the criteria for sorting AI systems into high-risk, medium-risk, and low-risk categories.

The risk-based approach allows Vietnam to target its heaviest compliance requirements at the systems most likely to cause harm, while keeping the regulatory burden light for everyday AI tools," said Dr Tran Dai Hai, a technology law researcher at the Vietnam National University.

High-risk systems are those with the potential to cause significant harm to life, health, national security, or the lawful rights and interests of individuals. Think AI used in criminal sentencing, autonomous vehicles, or critical infrastructure monitoring. Providers of high-risk AI must complete a pre-market conformity assessment and register on the national one-stop AI portal managed by the Ministry of Science and Technology (MoST).

Medium-risk systems cover scenarios where users may be confused or misled by undisclosed AI interactions or AI-generated content. Chatbots that do not clearly identify themselves as non-human fall into this bracket, as do deepfake generation tools. Providers must self-classify and notify MoST before deployment.

Low-risk systems

Innovation Incentives

This is not purely a restrictive framework. The law establishes a National AI Development Fund and introduces top-tier tax incentives for qualifying AI projects. A voucher scheme targets startups, helping them access computing resources, datasets, and sandbox^✦ testing environments. Vietnam's government has made it clear that the law is meant to attract investment, not repel it.

How Vietnam Compares to the Region

Vietnam is not operating in isolation. Across Southeast Asia, governments are grappling with the same question: how do you regulate AI without stifling growth? Singapore has opted for a voluntary, sectoral approach rather than a single binding law, relying on frameworks like the Model AI Governance Framework and the newer Agentic^✦ AI guidelines from IMDA. Thailand is drafting its own AI governance bill, expected later in 2026. The Philippines has introduced an AI bill that remains under committee review.

What sets Vietnam apart is the decision to go binding and comprehensive, rather than voluntary and piecemeal. By adopting a risk-based approach that echoes the EU AI Act, Vietnam positions itself as a serious regulatory player — one that global companies will need to factor into their compliance strategies alongside Brussels.

The ASEAN Guide on AI Ethics and Governance, released in 2024 with a supplementary guide for generative AI^✦ in 2025, provides a regional baseline. But it remains voluntary. Vietnam's law effectively operationalises many of those principles into enforceable obligations — a move that may push neighbouring countries to follow suit.

What This Means for Businesses and Developers

If you are building or deploying AI tools that touch the Vietnamese market, here is what you need to do now:

First, audit your AI systems against the three-tier classification. Determine whether any of your products fall into the high-risk or medium-risk category. If you are running agentic AI workflows, pay particular attention to systems that make autonomous decisions affecting individuals.

Second, check your disclosure practices. Medium-risk classification triggers transparency requirements. If your chatbot or AI assistant does not clearly identify itself as AI-powered, you will need to fix that before the grace period expires.

Third, plan for local presence requirements if you are a foreign provider of high-risk AI. The law requires either a commercial entity or an authorised representative in Vietnam — a significant operational consideration for companies currently serving the market remotely.

Finally, explore the incentives. The National AI Development Fund and voucher scheme represent genuine opportunities for startups and SMEs. Vietnam is signalling that it wants to grow its domestic AI ecosystem^✦, and early movers may benefit from practical AI skills training and favourable regulatory treatment.