Vietnam’s Artificial Intelligence Law 2025: What Businesses Using AI Need to Prepare

Vietnam’s Artificial Intelligence Law 2025 (Law No. 134/2025/QH15) was passed by the National Assembly in December 2025 and took effect on 1 March 2026. It is Vietnam’s first law creating a comprehensive legal framework for researching, developing, providing and using AI.

Vietnam’s risk-based approach to AI management is drawing international attention: at the Vietnam–Australia Digital Cooperation Forum during Vietnam International Digital Week 2026 (late October), the Australian side explored this experience. So what should businesses using chatbots, content-creation tools or AI-powered automation systems keep in mind?

Key points of Vietnam’s Artificial Intelligence Law 2025

The “human-centered” principle

Under Article 4, AI must serve people, not replace human authority and responsibility, and ensure fairness, transparency and freedom from bias. Every AI system must allow humans to supervise, control and intervene when needed.

Clearly defined roles

The law clearly defines “developers”, “providers”, “deployers”, “users” and “affected persons”. Businesses need to identify which role they play, because legal obligations differ by role.

Vietnam’s AI Law 2025 manages AI across 3 risk levels

Article 9(1) classifies AI systems into three levels: high, medium and low risk. Providers classify their own systems before putting them into use; if classification is difficult, they may ask the managing authority for support.

• High risk: systems that can significantly affect health, life, lawful rights and interests, public interests or national security. The official list currently contains 46 systems. Some must undergo a conformity assessment by an independent third party before operation, and are subject to periodic checks or checks when there are signs of violation.
• Medium risk: supervised through reports, sample checks or independent assessments.
• Low risk: checked only when an incident occurs or when safety assurance is requested.

According to a representative of the National Digital Transformation Department, the high-risk list is not fixed: a system can move from low to high risk as its level of automation rises and human supervision decreases.

Obligations for high-risk systems

• Design systems to ensure human supervision is possible.
• Build, update and store technical documentation and operation logs for inspection.
• Users must follow operating procedures and must not intervene on their own in ways that change the system’s functions.

Vietnam’s AI Law 2025 requires labeling of AI-generated content

This is the point with the widest impact on businesses doing marketing and content. Under Article 11(3), deploying organizations must clearly disclose when publishing content generated or edited by AI, especially when the content could cause confusion about its authenticity. Content that simulates humans or recreates real events (deepfakes) must carry an identification label. Artistic and cinematic works are treated flexibly.

For a business website, this means: AI-generated images, videos and voices — especially images of real people or “customers” — need to be clearly noted.

Transition period: how much time is left?

For AI systems already deployed before the law took effect, Article 35 allows a transition period:

• 18 months for sensitive sectors such as healthcare, education and finance.
• 12 months for other sectors.

Counting from 1 March 2026, the 12-month mark falls in early March 2027 — meaning businesses have only about 4 months left to review. Check the official guidance documents to determine the exact deadline that applies to your case.

Support policies for businesses

The law does not only regulate; it also encourages development:

• Small and medium-sized enterprises receive financial support, training and market connections to commercialize AI products.
• The State prioritizes the use of domestic AI products and services (Article 18).
• A National Artificial Intelligence Development Fund is established (Article 22).
• A controlled testing mechanism (sandbox) for new technologies whose risk boundaries are not yet clear.

Compliance checklist for businesses under Vietnam’s AI Law 2025

• Inventory all AI tools in use: chatbots, content creation, customer scoring, recruitment, automation.
• Identify your role: are you a provider, a deployer or only a user?
• Classify the risk for each system, prioritizing systems related to health, finance and customer rights.
• Label AI content on websites, social media and advertising.
• Keep logs and technical documentation for important systems.
• Keep humans in the decision loop for decisions affecting customers.
• Protect data: control the data fed into AI tools, especially personal data.

Conclusion

Vietnam’s Artificial Intelligence Law 2025 is not meant to ban AI; it aims for “fast – safe – humane”. Businesses that review early will both avoid legal risk and build trust with customers. If you need advice on AI deployment and automation in compliance with the regulations, contact Gnurt.

This article is for reference only and does not replace legal advice. Sources: People’s Security Academy, VnExpress.