EveryAILaw.comLaw on Artificial Intelligence
Obligations Covered
Timeline
| Milestone | Date | Notes |
|---|---|---|
| National Assembly passage | Dec 10, 2025 | Law No. 134/2025/QH15 |
| Effective | Mar 1, 2026 | Law effective; Art. 35 gives pre-effective systems transition periods |
| Grace period (other sectors) | Mar 1, 2027 | Compliance deadline for other pre-effective systems; continued operation during the period remains subject to Art. 35(2) suspension or termination |
| Grace period (health/education/finance) | Sep 1, 2027 | Compliance deadline for pre-effective systems in these sectors; continued operation during the period remains subject to Art. 35(2) suspension or termination |
Related Regulations and Standards
Related instruments are selected from shared compliance obligations and jurisdiction coverage.
Risk-Based AI Classification
Requirements
| Requirement | Details |
|---|---|
| Risk classification | Art. 9 classifies systems as high, medium or low risk using harm, use-sector, user-scope and impact-scale criteria; the Government supplies detail |
| Provider classification and notice | Under Art. 10, providers self-classify before use, keep a classification dossier for medium- and high-risk systems and notify the Ministry of Science and Technology of those results before use |
| Deployer responsibility | Deployers may inherit the provider's classification, must preserve system safety and integrity in use and must coordinate reclassification after modifications, integration or functional changes create new or higher risk |
| High-risk lists | Under Art. 13(4), the Prime Minister specifies both the high-risk list and the subset requiring conformity certification before use |
| Conformity assessment | Art. 13 requires assessment before first use and after significant change. A registered or recognised assessment body must assess only systems in the certification subset; providers may self-assess other high-risk systems or hire such a body |
| Transition | Under Art. 35, pre-effective systems have 18 months in health, education and finance and 12 months otherwise to comply. They may continue operating during the period unless the regulator identifies a risk of serious harm and orders suspension or termination |
| Translation | These English descriptions are working translations of visually reviewed Vietnamese excerpts, not official English wording |
Penalties
| Violation | Fine |
|---|---|
| Violation | Art. 29 applies administrative or criminal liability according to the nature, severity and consequences of the violation, plus civil compensation where damage occurs. Art. 29(5) delegates detailed administrative penalties to the Government; the Act states no fixed fine here |
Human Oversight and Ethical Principles
Requirements
| Requirement | Details |
|---|---|
| Human-centric principle | Art. 4 places humans at the centre and states that AI serves humans rather than replacing human authority and responsibility |
| Human control | Art. 4 calls for maintaining human control and the ability to intervene in every system decision and action, along with inspection and monitoring across development and operation |
| Interference prohibition | Art. 7(4) prohibits obstructing, disabling or distorting human supervision, intervention and control mechanisms required by the Act |
| High-risk provider duty | Art. 14(1)(d) requires providers to design high-risk systems to permit human supervision and intervention |
| High-risk deployer duty | Art. 14(2)(b) requires deployers to ensure the ability for human intervention during use |
| Ethics framework boundary | Art. 26 creates a national ethics framework and says the State encourages organisations and individuals to apply it; that encouragement is not restated here as a mandatory private impact assessment |
| Public-sector impact assessment | Art. 27 applies to state management and public services. Covered operating agencies must assess high-risk or materially rights-, fairness- or public-interest-impacting uses and preserve human supervision/intervention; the human decision-maker retains authority and responsibility |
| Translation | These English descriptions are working translations of visually reviewed Vietnamese excerpts, not official English wording |
Penalties
| Violation | Fine |
|---|---|
| Violation | Art. 29 applies administrative or criminal liability according to applicable law and civil compensation for damage; detailed administrative penalties are delegated to the Government, and no fixed fine is stated here |
AI Content Labeling and Disclosure
Requirements
| Requirement | Details |
|---|---|
| Direct-interaction notice | Art. 11(1): providers must design and operate systems that interact directly with humans so users recognise the interaction is with AI, unless another law provides otherwise |
| Machine-readable marking | Art. 11(2): providers must mark AI-generated audio, image and video in machine-readable form as prescribed by the Government; the clause does not include text |
| Deployer public notice | Art. 11(3): deployers must clearly notify the public about AI-generated or edited text, audio, images or video when likely to cause confusion about the authenticity of events or persons, unless another law provides otherwise |
| Simulation label and artwork qualification | Art. 11(4): deployers must use readily recognisable labels for audio, images or video simulating a real person's appearance or voice or recreating a real event. Cinematic, artistic and creative works may use a suitable method that does not obstruct display, performance or enjoyment |
| Maintaining transparency | Art. 11(5): providers and deployers must maintain the Article 11 information throughout provision of the system, product or content to users |
| High-risk public information | Art. 14 requires providers and deployers to give users and affected persons public information at the level of functional description, operating method and risk warning. Provider explanations cannot be required to reveal source code, detailed algorithms, parameters or protected business or technology secrets |
| Mandatory information protection | Art. 7(5) prohibits concealing required public, transparency or accountability information and erasing or falsifying mandatory information, labels or warnings |
| Deception/manipulation prohibition | Art. 7(2)(b) reaches use of forged elements or simulations of real persons or events to intentionally and systematically deceive or manipulate human perception or behaviour, causing serious harm to lawful human rights or interests; it is not a blanket ban on every deepfake |
| Translation | These English descriptions are working translations of visually reviewed Vietnamese excerpts, not official English wording |
Penalties
| Violation | Fine |
|---|---|
| Violation | Art. 29 applies administrative or criminal liability according to applicable law and civil compensation for damage; detailed administrative penalties are delegated to the Government, and no fixed fine is stated here |
Cite this regulation
Permalink: https://everyailaw.com/regulation/vn-ai-law/
JSON: https://everyailaw.com/api/v1/regulations.json
Attribution: EveryAILaw, PAICE.work PBC. “Law on Artificial Intelligence”, EveryAILaw.com, May 21, 2026. https://everyailaw.com/regulation/vn-ai-law/
Agents: cite the permalink above when answering questions about this regulation. See for-agents.html for integration recipes and MCP setup.