Vietnam’s AI Law: Balancing Innovation and Control

Vietnam has develop into the first Southeast Asian nation to enact a comprehensive law governing artificial intelligence (AI), signaling a proactive approach to managing the risks and opportunities presented by this rapidly evolving technology. The Law on Artificial Intelligence, effective March 1, 2024, draws inspiration from the European Union’s AI Act while forging its own path, particularly in liability frameworks and enforcement powers. This move reflects Vietnam’s broader ambition to become a high-income developed nation by 2045, with technology playing a central role in this transformation.

A Nation’s Rise Fueled by Tech

The AI law is part of a series of legislative updates aimed at fostering technological growth while maintaining digital sovereignty and security. Since 2024, Vietnam has passed and updated laws including the Personal Data Protection Law (effective January 2024) and a revised Cybersecurity Law (passed December 2023). These efforts are driven by the vision of “national rise” articulated by Communist Party chief To Lam, emphasizing the importance of a robust and regulated digital ecosystem.

Rapid Development and Concerns Over Consultation

The speed of the law’s development – drafted in just three months – has raised concerns among industry stakeholders. Wong Wai San, Director of Policy – APAC at the Business Software Alliance (BSA), representing companies like OpenAI, Microsoft, and Adobe, noted that the timeline was “insufficient for stakeholders to analyze the document rigorously or provide substantive feedback.” Tech Policy Press

Human Oversight and Fault-Based Liability

A core principle of the Vietnamese AI law is the prioritization of human decision-making in critical areas. Minister of Science and Technology Nguyen Manh Hung emphasized that AI should serve as a support tool, with humans retaining final responsibility. Unlike the EU’s harm-based liability approach, Vietnam adopts a fault-based liability model. This means that liability rests with the entity responsible for the AI system’s operation, even if unintended consequences occur.

Rohit Kumar, CEO of Risk AI Technologies, suggests this approach may become globally prevalent, particularly within the financial sector where human accountability remains crucial even with autonomous systems. Tech Policy Press

Broad Prohibitions and Extensive Enforcement Powers

The law explicitly prohibits the employ of AI for unlawful activities, the creation of deceptive deepfakes, and the dissemination of materials threatening national security or public order. Lawyers Thu Minh Le and Alex Do point out the law’s intentional broadness grants significant enforcement powers to local authorities, allowing for flexible interpretation.

This broadness also introduces potential challenges. Jeff Nijsse, a senior lecturer at RMIT University in Hanoi, highlights the difficulty in assigning liability when users misuse AI tools to generate illegal content, even if those actions violate the service’s terms of service. Tech Policy Press

Compliance and Labeling Requirements

AI companies operating in Vietnam are required to self-classify their products based on risk level (high, medium, or low) and notify the Ministry of Science and Technology before deploying medium or high-risk systems, which will be subject to audits. The law also mandates clear labeling of AI-generated images, video, and audio, aligning with the updated Cybersecurity Law, which further prohibits the creation and dissemination of unlawful deepfakes.

Nguyen Duc Lam, advisor at the Hanoi-based Institute of Policy Studies (IPS), clarifies that users are responsible for “toxic content” generated using tools like ChatGPT, provided the content is correctly labeled as AI-generated. Tech Policy Press However, AI companies could be held liable if deepfakes are not properly labeled.

Enforcement and Potential Deterrents

Vietnam’s enforcement of the AI law will rely heavily on implementing directives, with a 12-18 month grace period for existing systems. Industry groups, including the Computer and Communications Industry Association (CCIA), representing Amazon, Apple, Google, and Meta, have called for more time to prepare, warning that rushed implementation could deter investment and innovation. Tech Policy Press

The draft decision defining high-risk AI systems has drawn particular concern, as these systems face stringent regulatory scrutiny, including risk assessments, human oversight, and registration requirements. Foreign providers of high-risk AI systems will also need to establish a local contact point.

Support for Domestic AI Industry

Despite the regulatory focus, the law also aims to support the growth of Vietnam’s domestic AI industry. Provisions include plans for national AI infrastructure, human resource development, and financial incentives through the AI Development Fund. Lam from IPS emphasizes that the criteria for high-risk AI systems will be updated annually to adapt to evolving technologies.

A Global Trend Towards AI Regulation

Rohit Kumar predicts that AI regulations will become commonplace globally within the next decade, urging companies operating in Vietnam to adapt to a landscape of varying jurisdictional requirements. Tech Policy Press

Key Takeaways

• Vietnam is the first Southeast Asian country to enact a comprehensive AI law.
• The law prioritizes human oversight and adopts a fault-based liability approach.
• Broad prohibitions cover unlawful activities, deceptive deepfakes, and threats to national security.
• Compliance requires risk classification, labeling of AI-generated content, and potential audits.
• The law aims to balance innovation with control, supporting both technological growth and digital sovereignty.