Vietnam's AI Law is here. We're here too.
Vietnam has passed Southeast Asia's first standalone AI Law, No. 134/2025/QH15, in force since March 2026 and the detail is now arriving fast. Decree 142/2026/NĐ-CP set out how the law works in practice: risk classification, conformity assessment, transparency duties and a national AI portal. Then, in June 2026 the Prime Minister published Decision 33/2026/QĐ-TTg, the official List of High-Risk AI Systems, in force from 15 August 2026.
Decision 33 names 46 high-risk AI system types across six sectors: education, health, banking, transport, legal proceedings, and ethnic & religious affairs. If your system is listed, you must meet specific compliance obligations by set deadlines. 1 March 2027 for most sectors, and 1 September 2027 for health, education and finance, through one of two routes: self-assessment, or mandatory third-party certification.
The duties apply to organisations that use AI, not only those that build it, including foreign organisations active in Vietnam's markets. And more decrees are still to come.
What Law 134/2025/QH15 means
Binding
Extraterritorial
Extraterritorial
Passed in December 2025 and in force since March 2026. This is current law, not a proposal, and the first compliance deadlines arrive in 2027.
Extraterritorial
Extraterritorial
Extraterritorial
Applies across Vietnam's AI value chain. You can be in scope without a local office, and foreign providers must appoint a local representative.
Risk-based
Transparency
Transparency
Three tiers by impact: light-touch controls for low-risk tools, up to conformity assessment and human-oversight duties for high-risk systems. Know your tier first.
Transparency
Transparency
Transparency
AI-generated audio, image and video must be labelled and machine-readable. A visible duty that applies the moment you publish.
What Decree 142 introduced
Risk classification, defined
Transparency duties, live now
Transparency duties, live now
The decree sets out how AI systems are sorted into high, medium and low risk and crucially, what organisations must declare in each tier. Providers classify before a system goes live; deployers must re-check when they customise or repurpose one.
Knowing your tier is now a concrete step, not a concept.
Transparency duties, live now
Transparency duties, live now
Transparency duties, live now
AI-generated audio, image and video that could be mistaken for real people or events must be visibly labelled, and generated media machine-readable, with no grace period.
If you use open-source models, the marking duty can fall to you.
The national AI portal
Short clocks, real duties
Short clocks, real duties
A single-window portal now receives risk-classification notifications and issues each system an identification code. Medium and high-risk systems must be notified before use.
Filing is immediate; inspection comes later, so the quality of your dossier is what matters.
Short clocks, real duties
Short clocks, real duties
Short clocks, real duties
Serious incidents carry a 72-hour reporting duty; an upward change in risk must be notified within 15 working days.
With the Prime Minister's high-risk list published, listed systems get a transition period, but human-oversight and logging duties begin immediately.
The Decision 33 High-Risk list
Decision 33/2026/QĐ-TTg
Decision 33/2026/QĐ-TTg
Decision 33/2026/QĐ-TTg
In June 2026 the Prime Minister's office published Decision 33; the official List of High-Risk AI Systems, naming 46 system types across six sectors: education, health, banking, transport, legal proceedings, and ethnic & religious affairs.
What changed?
Decision 33/2026/QĐ-TTg
Decision 33/2026/QĐ-TTg
If your system is listed, you must complete specific obligations by set deadlines (1 March 2027 for most sectors; 1 September 2027 for health, education and finance) via one of two routes: self-assessment, or mandatory third-party certification.
What do you need to do?
Decision 33/2026/QĐ-TTg
What do you need to do?
Answer three questions immediately:
1. are you listed,
2. which route applies and
3. know what your conformity dossier needs for your system declaration.
Why act now?
The duties are concrete now
The law is moving fast, and the details keep arriving. First, Decree 142 then Decision 33 with further decrees still to come. The organisations that classify their AI against the list now will be ready as the rest lands, not scrambling after them.
Know: which of your AI systems are caught, at what tier and what you need to do to comply.
This is what CANDA delivers
Our dedicated assessment classifies each of your AI systems against the defined criteria, flags the duties that already apply and produces a filing-ready Article 12 dossier.
Fixed-fee, 3-5 week engagement, built to the standard regulators will inspect against.
Achieve : provable readiness, locally applicable and globally aligned conformity.
Contact Us
Book a discovery call or make an appointment with us in Thao Dien
+84 976 974 544
42 Ngô Quang Huy, An Khanh, Ho Chi Minh, Vietnam
Copyright © 2026 CANDA Consulting Limited - All Rights Reserved.