Skip to content
Menu ▾
Get the Receipt, $349/mo Search

Use Regulayer

For attorneys The Regulayer Receipt HumanMark The free Sealer See the product Pricing How to buy

License Regulayer

Enterprise What is licensed Agentic AI Life sciences Insurance Deployment Begin with the problem

Proof

Live demos Verify free Open verifier

Learn

The Consequence Library Learning Center The Drift Desk Detectors compared Authorship marks compared

Company

About Investors Why now FAQ Support Legal
Patent pending

The law library · Vietnam · National law

Vietnam: Law on Artificial Intelligence, No. 134/2025/QH15

Luật Trí tuệ nhân tạo · passed by the National Assembly 10 December 2025 · in force 1 March 2026

Page verified 13 August 2026 against the Government of Vietnam’s official legal document portal (vanban.chinhphu.vn) and the drafting ministry’s own report of passage (mst.gov.vn). The signed official text is a scanned gazette document; this page states only what those government sources state, and says so.

Vietnam’s first AI law: a deliberately short framework of 35 articles that classifies AI systems into three risk tiers and regulates the outputs and use of AI rather than the building of models. High-risk systems need safety documentation prepared before deployment, checked by post-market audit rather than pre-approval. It has been in force since 1 March 2026.

Status

  • 10 Dec 2025Passed by the National Assembly, 429 of 434 delegates in favour, per the Ministry of Science and Technology’s report of the vote.
  • 1 Mar 2026In force, per the Government’s legal document portal record for Law No. 134/2025/QH15.
  • In forceThe law applies now. Signed by National Assembly Chairman Trần Thanh Mẫn.

What the law does

  • Risk tiersThree tiers: low, medium and high risk, with prohibited-conduct provisions alongside.
  • OutputsThe law regulates the outputs and usage behaviour of AI systems and the risks posed to society, not the internal development of models. The drafting ministry describes this as balancing safety and development.
  • High riskBefore deployment, the organisation must prepare safety assessment documentation. There is no pre-approval gate: the documentation is self-prepared and answers to government audit afterwards, post-market.
  • LiabilityFull compensation for damages, with service suspension available as an enforcement mechanism.

What this page carries

The signed official text on the Government portal is a scanned gazette document, so this page does not quote article numbers. Reports describe transitional arrangements for AI systems deployed before the law took effect. Those arrangements live in the article text and are not stated here. This page carries them when the article-level text is readable at source.

What the signed record shows

Self-prepared documentation plus post-market audit is exactly the shape a sealed record answers.

A regime with no pre-approval and an audit after the fact turns entirely on the quality of the operator’s own records. A signed, tamper-evident record of what the system did and what a person checked, made at the time and verifiable by anyone with public mathematics, is the cheapest form that evidence can take when the auditor arrives.

Regulayer for enterpriseThe AI audit trail →

See also the law library and AI rules by country.

Sources

Related

Checked against the source 13 August 2026: the Government portal record and the Ministry of Science and Technology report. Article-level detail withheld until the primary text is readable, and we say so. Information, not legal advice.

Information, not legal advice. Every entry is verified against the issuing body’s own document; where a source is reporting rather than the document, we say so.