On the afternoon of November 27, under the direction of Vice Chairman of the National Assembly Le Minh Hoan, the National Assembly discussed in the Hall the draft Law on Artificial Intelligence.
Lack of regulations on compensation and insurance liability
Regarding Article 20 of the draft Law on controlled testing mechanism for the development of artificial intelligence, National Assembly member Hoang Minh Hieu ( Nghe An ) stated that the main purpose of the controlled testing mechanism for artificial intelligence (AI) is to remove legal barriers that are hindering AI models, allowing limited testing to obtain data to prove effectiveness, thereby promoting long-term regulatory reform. This approach is also very consistent with our current legislative thinking, which is that institutions must go first to pave the way for development.
The delegate emphasized that regulations on controlled testing mechanisms must be the focus of the legal framework on AI, so the drafting agency needs to research and regulate this issue more specifically in the draft Law.
Accordingly, the delegate proposed to review the regulations on organizations and individuals participating in controlled trials being exempted or having some compliance obligations reduced as prescribed in Clause 2, Article 20 of the draft Law. Such regulations lead to the scope of exemption being limited to the provisions of this Law, while AI systems are often related to many other relevant legal provisions.
Delegates cited the mechanism of testing AI in self-driving passenger aircraft in Japan, which is exempted from some rules on ensuring aviation safety. Or the testing of self-driving cars in China and Japan is exempted from the provisions of the Personal Data Protection Law, which is to use image data without having to blur the faces of traffic participants to train and train AI vision.
From the experience of other countries, delegates proposed that the agency in charge of drafting the study should only stipulate in Clause 2, Article 20 in the direction of clearly defining that organizations and individuals participating in controlled trials are exempted or reduced from some compliance obligations prescribed in this Law and other relevant laws.
Clause 5, Article 20 assigns the state management agency on AI to organize the reception, appraisal and processing of dossiers to participate in controlled testing activities according to the appraisal and quick response process.
Delegate Hoang Minh Hieu noted that AI systems are not simply issues related to technology, algorithms... but also multi-disciplinary and interdisciplinary issues. Therefore, the provisions in Clause 5 are probably only suitable for general-purpose AI systems, such as generative AI (ChatGPT, Gemini)... As for specialized AI systems, the interdisciplinary nature is very high.
Citing the DrAid system, an AI product supporting medical image diagnosis developed by VinBrain, delegate Hoang Minh Hieu said that if it is registered for testing, it is impossible not to have the monitoring and management role of the Ministry of Health. Therefore, in some countries, the assessment and licensing of controlled testing of AI systems is assigned to an interdisciplinary council.
Delegates proposed to revise the provisions in Clause 5, Article 20 of the draft Law in the direction of "the state management agency on artificial intelligence presides over and coordinates with relevant state management agencies to organize the reception, appraisal and processing of dossiers to participate in controlled testing activities".
The delegate also said that the provisions in Article 20 of the draft Law still lack provisions on compensation and insurance responsibilities. And it does not mention who is responsible if the test causes damage to a third party. According to the general principle, the testing organization or individual will have to pay civil compensation if causing damage. Therefore, delegate Hoang Minh Hieu suggested considering adding this provision to the draft Law.
All AI software operating in Vietnam must be equipped with a " standard human-like foundation database "
National Assembly Deputy Tran Van Lam (Bac Ninh) highly appreciated the draft Law for defining the principle of putting people at the center; classifying and controlling risks; inspecting high-risk AI systems; and developing national computing and data infrastructure. This is a progressive legal framework, in line with international trends.
However, to make the Law truly complete and have a long-term vision, the delegate proposed a key addition: Consider AI as a "newborn human being". An "intelligent but innocent entity": not knowing the law, not understanding ethics and etiquette, having no cultural values, and not distinguishing right from wrong if not taught.
“If we consider AI as an emerging entity, our responsibility is not only to control it, but also to teach it from the beginning, like teaching a child about standard human values. And this should be a mandatory provision in this Law, not just a recommendation,” said delegate Tran Van Lam.
Delegates proposed that the National Assembly add a legal principle in the Law: "All AI software operating in Vietnam, whether developed by domestic or foreign organizations, must be equipped with a standard human database ."
According to the delegate, the “standard human foundation database” must contain sufficient information to teach AI to understand Vietnamese law; teach AI ethical standards, culture, and good traditions of the nation; teach AI the ability to identify right and wrong, standard and unstandard; behavioral limits that must not be exceeded, and rules for identifying risks to humans. And most importantly, ensure that this foundation database must always take absolute priority in governing AI behavior; must not be overwritten, replaced, or disabled by any other algorithm or training data. This is the legal and technical “golden ring” that ensures AI does not go astray.
To effectively implement this orientation, delegates suggested that the Ministry of Science and Technology take the lead in building a unified, standardized, and continuously updated “Being Vietnamese” database platform; providing it to AI developers as a mandatory requirement. The Government should develop a strong national computing infrastructure to ensure that Vietnamese AI is trained on Vietnamese values, not dependent on foreign models.
At the same time, build a National AI Testing Center to promptly assess deviant behavior, control high-risk models, and promptly intervene when there are signs of going beyond limits. Promote AI education throughout society, focusing on helping people, businesses, and especially the public sector to have the capacity to properly identify, use, and monitor AI.
Delegate Tran Van Lam emphasized, "Today's Law on Artificial Intelligence is not just a technology law, but a law of the future; a commitment of the National Assembly to the People, we develop technology but do not lose control; promote innovation but do not let risks go beyond our reach. And nurture AI with Vietnamese values to serve the Vietnamese people."
Source: https://daibieunhandan.vn/phat-trien-cong-nghe-thuc-day-doi-moi-nhung-khong-de-rui-ro-vuot-tam-10397378.html
Comment (0)