Developing the Law on Artificial Intelligence: A flexible framework law with a human-centric approach

According to the Submission Report presented by Minister of Science and Technology Nguyen Manh Hung, the draft Law on Artificial Intelligence consists of 8 chapters and 36 articles, aiming to institutionalize the policies of the Party and the State on the development and governance of digital technologies. The Law is designed as a “framework law, foundational law” with a high degree of flexibility, sufficiently open to adapt to the rapid pace of technological innovation, while at the same time setting clear “red lines” to protect national interests, human rights, citizens’ rights, and digital sovereignty. The Minister emphasized that the law is not only for management but also for creating a favorable environment “to promote innovation and enhance national competitiveness".

One of the major orientations of the draft law is to regulate AI systems according to their level of risk. Instead of imposing comprehensive restrictions, the Law focuses on stricter requirements for systems with high risk and direct impacts on fundamental rights and interests of citizens, as well as on security, national defense, finance, healthcare, etc. Meanwhile, low-risk applications are encouraged to develop with simplified procedures, minimizing administrative burden for enterprises and researchers. The guiding principle is “not to suppress innovation, but not to loosen management".

Reviewing the draft Law, Chairwoman of the Committee on Science, Technology and Environment Nguyen Thanh Hai stated that in the field of AI, management and promotion are not opposing ends, and that “transparent and effective governance is the very foundation for sustainable development.” The Committee proposed clarifying the spirit of “harmonizing management and promotion,” avoiding the interpretation that it is merely a mechanical “balance.” In addition, while the choice of a “framework law” model is considered appropriate, it requires ensuring consistency with a wide range of existing laws such as the Civil Code, the Law on Cyberinformation Security, the Intellectual Property Law, the Law on Digital Transformation, and with new international treaties such as the United Nations Convention on Countering Cybercrime.

The Committee also noted that AI is penetrating deeply into all sectors and fields, from smart transportation, precision medicine, and digital finance to education, journalism, and culture. Therefore, the Law on Artificial Intelligence cannot “stand alone” but must become a “living law” within the digital legal ecosystem, to be further specified through specialized laws. There is an opinion that each law in key sectors should include specific provisions on AI to clearly define management responsibilities, safety standards, and the rights and obligations of relevant parties.

One issue receiving considerable attention from deputies is the mechanism of ex-ante and ex-post supervision. The current draft contains rather detailed regulations on technical documentation and system operation logs of AI systems before being put into operation. The Committee proposed that the Government review and reduce requirements of a listing nature to avoid turning procedures into barriers. In this direction, the Law should “shift strongly to ex-post supervision,” focusing on developing standards, technical regulations, assessment and inspection procedures, and mechanisms for investigation and supervision based on risk management.

Data is considered the “fuel” of AI, so the draft law allocates a separate chapter to databases on AI systems and data serving AI. The verification viewpoint emphasizes that data must be “accurate – sufficient – clean – live” and be connected and shared on a common platform, avoiding data siloing and fragmentation that hinder research and development. Along with this is the requirement to ensure cybersecurity, data security, and the protection of national AI infrastructure against risks of attacks, takeover, and large-scale data leakage.

Regarding legal liability, the Committee analyzed that any human error can be “replicated” or amplified by AI, while “the legal liability of AI itself is currently a matter of global debate.” If the law does not provide clear provisions, incidents may easily lead to disputes among providers, deployers, integrators, and users, especially with cross-border AI services. Therefore, the draft needs to supplement principles for allocating responsibility among stakeholders, clarifying the boundaries between intentional and unintentional faults and faults arising from technical limitations or unforeseeable factors, so as to ensure fairness while still encouraging innovation.

From the perspective of ethics and social impacts, the verifying body proposed emphasizing accountability regarding the impact of AI on employment, privacy, culture, and ethics, particularly for vulnerable groups. Applications that may lead to mass layoffs or the replacement of low-skilled labor must be accompanied by retraining measures and support for job transitions. In addition, the list of prohibited acts should be expanded, including the use of AI to incite violence, division, infringement upon national security, or election manipulation; and the use of AI to create and disseminate fabricated or deepfake content for fraud, insult, or causing public confusion. These risks “are no longer distant possibilities but have already become present in everyday digital life”.

Another issue attracting attention is the obligation of transparency and labeling for products and services that use AI. The Committee agreed with the policy requiring labeling for image, video, audio, and textual content generated or edited by AI, considering this an “ethical and legal requirement” to enable user identification and allow competent authorities to trace and handle violations. However, criteria, technical standards, and exemptions should be reasonably designed, such as considering a mechanism for applying “invisible labels” to hardware devices integrated with AI - such as televisions, refrigerators, and washing machines - to avoid increasing compliance costs while still ensuring effective management.

With a flexible, risk-based approach that places humans at the center and respects innovation, the Law on Artificial Intelligence is expected to become an important legal foundation, enabling Viet Nam to proactively seize opportunities from AI while minimizing risks and firmly safeguarding digital sovereignty, national security, and human values in the digital era.