VietnamLaw on Artificial Intelligence, risk classification and conformity assessment
Law No. 134/2025/QH15, arts. 9-10, 13-14
In force since .
An AI risk obligations rule binding public and private bodies.
- How the hook was established
- express
- What makes it apply
- Place of effect, Market targeting
- Obligation class
- Governance, DPIA, Reporting
- Audit expectation
- continuous
- Who audits it
- Self, Registered or designated auditor
- Where the report goes
- Kept, Produced on request
What it requires
- Classify your AI system's risk level (high, medium, or low) before putting it into service, based on its potential impact on rights, safety, security, and public interest, and the scale and context of its use.NIST AI RMFMAP 3.3
- For a medium-risk or high-risk AI system, prepare a classification dossier and notify the classification result to the Ministry of Science and Technology through the one-stop AI portal before putting the system into service.NIST AI RMFMAP 3.3NIST AI RMFGOVERN 1.1
- For a high-risk AI system, complete a conformity assessment before putting it into service or after a significant change, and maintain that conformity throughout operation.NIST AI RMFGOVERN 1.1
- Re-classify an AI system's risk level when a modification, integration, or functional change gives rise to a new or higher risk.NIST AI RMFMAP 3.3
These lines are LexLint's reading of the law's text.
In the frameworks
- NIST AI RMF
- GOVERN 1.1Legal and regulatory requirements involving AI are understood, managed, and documented.
- MAP 3.3Targeted application scope is specified and documented based on the system’s capability, established context, and AI system categorization.
- NIST AI 600-1
- GAI-RISK-07Human-AI Configuration
- GAI-RISK-09Information Security
- NIST AI 200-2 Draft
- OWASP LLM Top 10
- LLM03:2026Excessive Agency
- OWASP Agentic Top 10
- ASI09Human-Agent Trust Exploitation
- MIT mitigations Draft
- 1.1Board Structure & Oversight
- 1.2Risk Management
- 3.5Post-deployment Monitoring
- 4.1System Documentation
- 4.2Risk Disclosure
- 4.3Incident Reporting
1 more
- 4.5Third-Party System Access
- FINOS AIGF
- AIR-PREV-005System Acceptance Testing
- AIR-PREV-006Data Quality & Classification/Sensitivity
- AIR-PREV-024Human-in-the-Loop Action Approval Gate
- AIR-DET-004AI System Observability
- AIR-DET-011Human Feedback Loop for AI Systems
- AIR-DET-021Agent Decision Audit and Explainability
- ETSI EN 304 223
- 5.1.4-1Capability for human oversight designed in and maintained.
- 5.2.1-1Full inventory of assets, including their dependencies and connections.
- 5.2.2-5AI-specific incident management and recovery plans that are created, tested and maintained.
- 5.2.4-1Developers keep an audit trail of system design and post-deployment maintenance plans and offer it downstream.
- 5.4.2-1Operators log system and user actions.
Our reading of where this law's duties sit in each framework, and in the NIST AI RMF, line by line beside each duty above. It never says that running a control meets the law. How we read a framework against the law
Identifiers refer to ETSI EN 304 223 V2.1.1 (2025-12), © ETSI 2025, all rights reserved. ETSI owns the standard; no ETSI text (titles or sentences) is reproduced here. The standard is free to download from ETSI.
If you get it wrong
Criminal exposureYes
Private right of actionYes
Criminal exposure note
Article 29(1) provides that a violation of the Law or other relevant AI regulations is, depending on its nature, severity, and consequences, sanctioned as an administrative offence or examined for penal liability, with civil compensation where damage is caused; the Law does not itself define an AI-specific criminal offence or state an administrative fine amount, and Article 29(5) leaves the sanctioning regulations to a Government decree not yet issued as of this review.
Who enforces it
Enforcement body
The Ministry of Science and Technology, the focal agency responsible to the Government for state management of artificial intelligence nationwide, with the Government performing unified state management and provincial People's Committees managing artificial intelligence locally (art. 30(2)).
Settledness
Decree 142/2026/ND-CP details Articles 9, 10, 13 and 14 of the Law, and Decision 33/2026/QD-TTg of the Prime Minister, in force from , issues the list of high-risk AI systems under Article 13(4). Both are binding implementing instruments and are recorded here rather than as guidance.
- As of
- Open questions
- Does an AI system that is not on the Prime Minister's list of high-risk AI systems but could cause significant harm under Article 9(1)(a) of the Law count as high-risk, when Article 6(3)(a) of Decree 142/2026/ND-CP classifies as high-risk only the systems on that list?
- Which administrative sanctions apply to a breach of Articles 10, 13 and 14 of the Law, given that Article 29(5) calls for a Government decree on administrative sanctions for violations caused by artificial intelligence systems and Decree 142/2026/ND-CP contains no sanctions provision?
- Can a person who suffers damage from a high-risk AI system claim directly against the deployer under Article 29(2) of the Law, or only through the civil-law compensation rules to which Article 29(1) refers?
What it makes you log
Who may demand the log
Regulator, Auditor or notified body
What the log must hold
Data lineage, Input reference
Logging duty
Article 14(1)(c) names the record directly: a provider of a high-risk system must prepare, update and archive technical dossiers and operation logs. Article 10(1)'s classification dossier is a related but separate, more implicit-shaped duty (a dossier accompanying a medium- or high-risk system, never called a log or record); it is not the basis for this explicit logging duty.
- Kind
- Explicit
- As of
- Provision
- art. 14(1)(c)
- Trigger
- high_risk_systems
What this law does
Article 9 classifies an AI system as high, medium, or low risk based on its potential to cause significant harm to rights, safety, security, or public interest, its capacity to confuse users about whether they are dealing with an AI system, and criteria the Government is to detail further, including the field of use and the scale and range of its impact.
Article 10 requires a provider to classify its own system before putting it into service, to prepare a classification dossier for a medium-risk or high-risk system, and to notify the classification result to the Ministry of Science and Technology through the one-stop AI portal before deployment; a deployer inheriting a provider's classification must re-classify if a modification creates a new or higher risk.
Articles 13 and 14 require a high-risk system to undergo conformity assessment, by a registered assessment organization or by the provider itself depending on the system, before being put into service or after a significant change, and require the assessment result to be maintained throughout operation as a condition of continued use.
Decision No. 367/QD-TTg, the government's implementation plan for the Law, is preparing a Prime Ministerial decision listing high-risk AI systems by five criteria, but this list had not been issued as of this review.
When LexLint raises it
When your app profile says your app makes high-risk automated decisions, deploys a chatbot, generates content with AI, trains models or crawls the web.
Back to the example · Lint your app
Page updated .