Vietnam’s Sweeping AI Law: Revolutionary Controls or Innovation Killer?

In a landmark move, Vietnam has passed its first thorough artificial intelligence law with overwhelming support from the National Assembly. The legislation, set to take effect on March 1, 2026, was approved alongside amendments to the Intellectual Property Law and revised High Technology Law. The Ministry of Science and Technology will lead central oversight of AI activities throughout the country.

The new law establishes seven core principles for AI development that emphasize human-centered approaches and risk-based management. It creates a flexible framework that balances safeguards with incentives for innovation while focusing on national sovereignty and data autonomy.

Vietnam’s AI Law balances human-centered protections with innovation incentives, prioritizing national data sovereignty through flexible, principled governance.

Vietnam’s AI Law introduces a four-tier risk classification system similar to the EU AI Act. It identifies unacceptable risks that threaten national security or social order, and high-risk applications in sectors like finance, healthcare, and justice that require pre-market approval. The Prime Minister can update the high-risk list in real time to address emerging technologies.

The legislation prohibits several AI practices without proper authorization, including real-time biometric surveillance in public spaces, large-scale facial recognition through data scraping, and deepfakes that could destabilize public opinion. AI systems that manipulate cognitive behavior are also banned.

To promote innovation alongside these controls, the law establishes a National AI Development Fund, sandbox testing environments, and a startup voucher scheme. These measures support Vietnam’s National Strategy on AI through 2030. The regulatory approach aims to create a safe environment for AI growth while encouraging technological advancement.

Transparency requirements form a key part of the law. Users must be informed when interacting with AI systems, and AI-generated audio, images, and video must carry machine-readable markers to distinguish them from authentic content. High-risk AI systems must undergo conformity assessments and be registered in the National AI Database before deployment. The emphasis on transparency aligns with global concerns about AI bias that has been documented in automated analytical systems.

The law clarifies that AI cannot be recognized as a legal entity or hold copyright. Only humans or organizations can possess intellectual property rights related to AI outputs. This approach places liability for AI-generated damages on the owners who treat AI systems as assets, creating new challenges in determining responsibility for content that infringes on existing rights.