On 10 December 2025, the National Assembly passed Vietnam’s first Law on Artificial Intelligence (“AI Law”) at its 10th session. With this step, Vietnam joins a small group of jurisdictions most notably the EU (AI Act 2024) and South Korea (AI Basic Act 2024) - that have adopted a dedicated legal framework for artificial intelligence. The AI Law will take effect from 01 March 2026.

The AI Law is principle-based and risk-oriented, and is broadly similar to the approach taken under the EU AI Act; that said, it also introduces several features that go a step further, as briefly outlined below. It sets out the core principles and structure, while deliberately leaving room for the Government and line ministries to develop detailed guidance. In this sense, forthcoming implementing decrees, sector-specific regulations, and national planning instruments are expected to progressively complete and refine Vietnam’s AI legal framework.

This article highlights the main features of the AI Law and outlines areas where further regulatory clarification can be expected.

This article covers:

(1) Key principles and prohibited uses

(2) Risk classification and administrative requirements

(3) State policies to promote AI

(4) Anticipated legal implications

1. Key Principles and Prohibited Uses

Like the EU AI Act, Vietnam’s AI Law is built around a limited set of core principles that apply across all AI activities, including research, development, deployment, and use. These principles include:

• Human-centric and lawful use

• Human authority, responsibility and control

• Fairness, transparency, and non-discrimination

• Sustainable and responsible development

One of the most distinctive features of the Vietnamese AI Law is its express affirmation that AI does not replace human authority or responsibility. This goes further than the EU approach, which focuses primarily on human oversight and control. Under the Vietnamese formulation, the message is clear: where AI is used, a human decision-maker must remain legally responsible.

Non-discrimination

Non-discrimination is another central pillar of the AI Law. Substantively, it targets unlawful bias based on protected characteristics such as age, gender, ethnicity, or similar grounds.

For example, an AI-driven recruitment system must not automatically exclude candidates based on age or ethnicity without lawful justification. Vietnamese law already addresses discrimination in sector-specific legislation, notably the Labor Code. The AI Law does not replace these rules; instead, it extends and reinforces them in the AI context.

In practice, whether an AI system is considered “discriminatory” will still depend on substantive legal standards. Not every differential outcome is unlawful. The decisive question will be whether the discrimination is legally unjustified, not merely whether differences in treatment exist.

2. Risk Classification and Administrative Requirements

Vietnam adopts a risk-based regulatory model, closely aligned with the EU AI Act.

Risk classification

Before deployment, AI providers must classify their systems into one of three categories:

• High risk: systems capable of causing significant harm to life, health, legal rights, national security, or public interests;

• Medium risk: systems that may mislead, influence, or manipulate users who are unaware they are interacting with AI;

• Low risk: all other systems.

Where classification is unclear, providers may request guidance from the Ministry of Science and Technology (“MOST”) based on a technical dossier.

Notification and supervision

• Medium- and high-risk systems must be notified to MOST via the national AI one-stop portal before deployment.

• Low-risk systems are exempt from notification, although voluntary disclosure is encouraged.

Regulatory oversight

• High-risk systems are subject to periodic or ad hoc inspections;

• Medium-risk systems are monitored through reporting, sampling, or independent assessments;

• Low-risk systems are generally subject only to reactive supervision.

Conformity assessment

• High-risk AI systems must undergo conformity assessment prior to deployment.

• Certain high-risk systems will require mandatory third-party certification.

• Other high-risk systems may be self-assessed or assessed by a registered conformity assessment agency.

Successful conformity assessment is a condition precedent to lawful deployment, and compliance must be maintained throughout the AI system’s lifecycle.

Transparency and labeling

Users must be informed when they are interacting with AI rather than a human. AI-generated or AI-modified content that could reasonably be mistaken for real must be clearly labeled.

While the policy objective is sound, technical guidance will be essential. The key challenge will be ensuring transparency without materially disrupting user experience, particularly for digital and platform-based services.

Foreign providers

Foreign providers offering high-risk AI systems in Vietnam must:

• appoint a lawful local contact point; and

• establish a commercial presence or authorized representative where certification is required.

This confirms Vietnam’s intention to regulate cross-border AI services with meaningful enforcement mechanisms.

3. State Policies to Promote AI Development

The AI Law is not solely restrictive. It is also clearly pro-development. Key policy directions include:

National AI infrastructure

The State will invest in and operate strategic AI infrastructure—including computing capacity, data resources, AI platforms, and Vietnamese-language models—as public services.

National AI Strategy

The Prime Minister is required to issue and update a National AI Strategy at least every three years, covering technology development, infrastructure, human resources, and priority sectors.

Core technology development

Vietnam prioritizes domestic research, development, and mastery of core AI technologies.

Incentives and funding

AI projects may benefit from tax incentives, investment incentives, and financial support under existing laws. A National AI Innovation Fund will support AI research, development, and commercialization. Overall, the policy message is clear: Vietnam aims to regulate AI without deterring innovation or investment.

4. Anticipated Legal Implications

Duty of care and Tort liability

The AI Law confirms that using AI does not eliminate a professional’s duty of care. On the flip side, global regulatory discussions are increasingly asking whether, once AI becomes industry standard, failing to use it could itself amount to negligence. Such a theoretical issue is unlikely to be imposed directly by statute, but may gradually emerge through civil liability and professional standards. For now, the expectation under the AI Law is clear on one side: use AI carefully, supervise it actively, and never rely on it blindly.

Liability allocation – the “AI triangle”

One of the most significant features of the Vietnam AI Law is its approach to civil liability. Concerning that issue, the AI Law sets out the three roles: the developer, the deployer and the injured party (typically, customer). Where a properly managed high-risk AI system nevertheless causes damage:

• the deployer is responsible for compensating the injured party;

• the deployer may seek reimbursement from the provider or developer if contractually agreed.

This reflects familiar product-liability and consumer-protection logic: the party closest to the harm pays first. Developers are not automatically liable unless there is fault, contractual allocation, or another legal basis.

In practice, this places strong emphasis on:

• contractual risk allocation;

• indemnities;

• insurance coverage.

Conclusion

Vietnam’s AI Law is not the final word, but it is a decisive starting point. It establishes responsibility, human control, and risk management as the governing themes of AI regulation. The true impact will depend on implementing decrees, sectoral regulations, and enforcement practice, all of which will require close and ongoing attention from businesses deploying or supplying AI in Vietnam.