5 Key Points from Law on Artificial Intelligence 2025: What Businesses Need to Know

Artificial Intelligence Systems Must Be Classified by Risk Level

The Law on Artificial Intelligence 2025 requires all organizations operating artificial intelligence systems in Vietnam to classify those systems according to the level of risk they pose. This risk based approach is central to the broader artificial intelligence regulatory framework in Vietnam and establishes four tiers summarized below.

Risk Level	Description	Key Obligations
Unacceptable	Systems such as extreme social scoring or uncontrolled public facial recognition	Absolute prohibition; no deployment permitted
High	Systems used in healthcare, recruitment, financial services, or law enforcement,	Human oversight, technical documentation, conformity assessment, and incident reporting
Limited	Systems with moderate potential impact on users	Transparency and user notification obligations
Minimal	Systems posing negligible risk to individuals or society	Voluntary codes of conduct; no mandatory requirements

Generative Artificial Intelligence Faces New Disclosure and Labeling Rules

The Law on Artificial Intelligence 2025 introduces specific obligations for providers of generative artificial intelligence, including systems that produce text, images, audio, or video. Providers must disclose the data sources used to train their models, directly addressing concerns around copyright issues with artificial intelligence generated content in Vietnam.

In addition, all content generated by artificial intelligence must carry a digital watermark. This technical measure helps users and platforms identify whether content was created by a human or produced by a machine, supporting efforts to combat misinformation.

Users Have the Right to Request an Explanation of Automated Decisions

The Law on Artificial Intelligence 2025 grants individuals a formal right to explanation. When an artificial intelligence system makes a decision that affects a person’s interests, such as rejecting a loan application or filtering out a job application, the affected individual may request the organization to explain how and why that decision was made.

Organizations must also inform users when they are interacting with an artificial intelligence system rather than a human. These measures are designed to build public trust and ensure individuals can make informed choices when using services powered by artificial intelligence.

Foreign Providers Must Comply with Local Requirements and Face Revenue Based Penalties

The Law on Artificial Intelligence 2025 applies to foreign technology companies providing artificial intelligence services to users in Vietnam. These companies may be required to appoint a legal representative in Vietnam or store certain categories of data locally.

Penalties for violations are calculated as a percentage of global revenue, following a model similar to Vietnam’s personal data protection regulations. This ensures that sanctions remain proportionate regardless of the size of the organization.

A Regulatory Sandbox Supports Innovation While Managing Compliance Challenges

The Law on Artificial Intelligence 2025 establishes a regulatory sandbox that allows startups and technology companies to test high risk artificial intelligence products in a controlled environment. Participants operate under defined conditions with supervision from the competent authorities and must transition to full compliance upon exiting. This mechanism builds on the artificial intelligence development guidelines in Vietnam issued by the Ministry of Science and Technology in 2024.

This mechanism addresses a practical concern: Vietnam currently has a limited number of organizations capable of conducting artificial intelligence safety audits, and compliance costs may challenge smaller enterprises. The sandbox provides a pathway for these businesses to develop their products before facing the full weight of regulatory obligations.

Practical Implications

Businesses developing or using artificial intelligence systems in Vietnam should assess whether their systems fall within the scope of the Law. Organizations should review their products against the risk classification criteria, evaluate disclosure obligations, and confirm whether cross border compliance measures apply. The Law on Artificial Intelligence 2025 takes effect on March 1st, 2026.

About ANT Lawyers, a Law Firm in Vietnam

We help clients overcome cultural barriers and achieve their strategic and financial outcomes, while ensuring the best interest protection, risk mitigation and regulatory compliance. ANT Lawyers has lawyers in Ho Chi Minh city, Hanoi, and Danang, and will help customers in doing business in Vietnam.