 |
| A woman experiences an AI-powered robot that applies AI technology to assist with administrative procedures at Cua Nam ward, Hanoi__Photo: VNA |
The Ministry of Science and Technology (MOST) is gathering public opinions on the draft Law on Artificial Intelligence (AI), which is scheduled to be submitted to the National Assembly at the upcoming 10th session. If passed, the legislation would position Vietnam among the few countries worldwide with a comprehensive legal framework governing AI.
Seven fundamental principles
All AI-related activities in Vietnam must follow seven fundamental principles.
The first principle states that AI technology must be human-centered, meaning that AI technology must serve and support humans rather than replace them in making critical decisions. AI must always remain under human control, supervision and ultimate responsibility.
The second principle emphasizes safety, fairness, transparency and accountability, requiring that AI systems be developed and operated in a secure, reliable and non-discriminatory manner, with clear identification of legal liability for any damage caused.
The third principle focuses on national autonomy and international integration, calling for the development of domestic capabilities in AI technology, infrastructure, data and models while ensuring active participation in international cooperation based on alignment with global principles and best practices.
The fourth principle helps promote inclusive and sustainable development, ensuring that AI contributes to the sustainable national socio-economic growth, environmental protection, and preservation of cultural identity while providing fair and equal opportunities for all citizens.
The fifth principle highlights the importance of balance and harmony in both policymaking and implementation.
The sixth principle introduces a risk-based management approach, under which the State applies regulatory measures proportionate to the level of risk posed by AI systems, imposing mandatory controls only on systems that present clear and significant potential harm.
The seventh principle facilitates innovation, with the State responsible for creating a safe and supportive legal and policy environment that promotes AI research, and commercialization.
Four levels of AI risk
Under Article 9 of the draft Law, AI systems would be arranged for management according to four levels of risk.
The first level, labeled as “unacceptable risk,” includes systems that pose serious threats to human rights or national security. Such systems are strictly prohibited under Article 11.
The second level, defined as “high risk,” covers AI systems that may significantly affect human life, health, safety, security or fundamental rights. These systems must submit to rigorous registration, assessment, transparency and supervision regulations.
The third level, identified as “medium risk,” applies to systems that may interact directly with humans or produce contents that must be described as AI-generated to users.
The fourth level, referred to as “low risk,” applies to all other systems that pose minimal risk. Developers of these systems are encouraged to voluntarily apply technical standards, and they would only be subject to post-inspection if risks arise.
The Government would issue a list of high-risk sectors and criteria for identifying each risk level to ensure that risk management remains flexible and adaptable to real-world conditions.
Incentives for small businesses and startups
To promote innovation and participation in the AI sector, the draft Law provides small enterprises and startups with raft of incentives.
Accordingly, they would be entitled to exemption from or reduction of certain administrative procedures when participating in regulatory sandboxes, such as registration and conformity assessments.
Small enterprises and startups would also be allowed to use sandbox data and reports to simplify subsequent registration processes. They would receive priority access to national AI infrastructure, computing resources, and open data. In addition, they would be provided with free compliance templates and standardized documentation to help reduce legal complexity. Businesses that meet specific conditions might also receive partial or full exemption from conformity assessment charges.
In addition, the State would further support small- and medium-sized enterprises (SMEs) through the issuance of AI vouchers to help them access to AI products, services and resources at reduced costs, thereby promoting technological adoption and development.
Strict penalties for violations
Under the draft Law, the MOST proposes strict penalties for violations, particularly for organizations and enterprises whose activities cause serious harm to individuals, society and national security. In some cases, fines may be calculated as a percentage of global revenue, following the model of the European Union’s AI Act.
According to Articles 63 and 64, violations include developing or using prohibited AI systems, such as harmful deepfakes, social scoring systems or emotion recognition technologies in workplaces. Violations also include failing to comply with high-risk system obligations, such as registration, certification or impact assessment; providing false information or concealing incidents; and infringing upon human rights, privacy, trade secrets or national security.
Penalties might include administrative fines amounting to the percentage of global revenue, for cross-border businesses, revocation of certifications, suspension of operations, bans on the distribution of AI systems, or even criminal prosecution, for acts causing serious consequences or being committed intentionally.
Expert opinions
At a recent workshop gathering expert opinions on the draft Law on AI, Dau Anh Tuan, VCCI Vice Secretary-General and Director of the Legal Affairs Department highly appreciated the open-minded, scientific and innovative approach taken in the draft Law.
However, to ensure that the Law is effectively implemented and truly becomes a driver of development, Dau Anh Tuan stresses the need to establish a flexible and adaptive legal framework that enables rapid policy updates, experimentation, and feedback. He also emphasizes minimizing administrative barriers for businesses as much as possible, prioritizing investment in data and computing infrastructure, and clearly defining legal responsibilities across the AI value chain from developers and providers to end-users.
Deputy Chairman of the National Assembly’s Committee on Science, Technology and Environment Tran Van Khai emphasized that Vietnam considers AI as a national priority technology that plays a key role in the digital transformation, the development of a knowledge-based economy, and the improvement of the national competitiveness. He said that this was Vietnam’s first law in the field of AI, which would help establish a legal foundation to promote the research, development, application and governance of AI in a safe, responsible and human-centered manner.
According to Deputy Minister of Science and Technology Pham Duc Long, alongside the drafting of the AI Law, the Government and the MOST are scheduled to implement policies to support the development of a national AI ecosystem, including initiatives to strengthen data and computing infrastructure, train human resources, and assist SMEs in adopting AI technologies.- (VLLF)
