Minister Nguyen Manh Hung said the State will build a national AI supercomputing and data center to research, train, and develop Vietnamese models to reduce foreign dependence.
Responding to delegates' opinions at the discussion session at the National Assembly on the afternoon of November 27, Minister of Science and Technology Nguyen Manh Hung said that the Artificial Intelligence Bill was built in the direction of a framework law, inheriting international experience to meet the requirements of both controlling risks and creating space for innovation. Detailed technical regulations will be issued by the Government to ensure flexibility in the context of rapidly changing technology.
He said that AI governance is essentially intelligence governance, similar to the way society regulates people's use of knowledge when impacting the community. Intelligence is formed from education and official information, so input data will be managed by the Data Law and the Personal Data Protection Law ; AI enterprises must publish legal data sources; AI systems must assess risks, independently audit and be accountable.
Regarding output, the State applies safety standards, post-inspection mechanisms and sanctions for violations. "AI governance does not mean inventing a completely new mechanism but inheriting human intelligence management experience and applying it to a stronger and faster tool," the Minister said.
He emphasized the highest principle of putting people at the center; AI serves people, not replaces decision-making power. Transparency is a mandatory requirement and people must know when they are interacting with AI. Legal responsibility when AI systems cause damage will be determined between developers, suppliers, deployers and users.
The State will build a national AI supercomputing and data center to serve research and training; encourage Vietnamese language models and digital data to reduce foreign dependence. AI in public services and sensitive fields must run on Vietnam's AI infrastructure. Priority areas include public administration, health, education, agriculture, industry and transportation.
Minister of Science and Technology Nguyen Manh Hung spoke before the National Assembly on the afternoon of November 27. Photo: Hoang Phong
The Minister agreed to manage AI according to three levels of low, medium and high risk. Regarding the question of whether AI is a source of high risk or not, he said that high risk sources are often inherently dangerous physical means such as cars and explosives. Even if users are careful, they can still cause accidents. AI in this draft law is not a dangerous physical means, but the risk from decision-making.
According to the Minister, AI brings great benefits but also risks and many problems of AI can be solved by AI itself such as human resource training, detecting ethical violations or identifying deepfake. "AI and its problems will always exist. We do not eliminate it but live with it and manage it wisely," he said.
Need for a controlled testing mechanism with AI
Delegate Hoang Minh Hieu (Law and Justice Committee) assessed that the drafting agency had fully explained the opinions of delegates in the group discussion and agreed with the approach of both promoting AI development and controlling risks. The controlled testing mechanism (Sandbox) is necessary because AI is a new field, related to many industries and difficult to apply effectively if only relying on the current legal framework.
The goal of Sandbox is to temporarily remove legal barriers, allowing limited testing of models to collect data to prove effectiveness, thereby perfecting long-term regulations. This approach is consistent with the legislative thinking of "institutions go first", paving the way for innovation. He suggested that the bill should have more specific regulations on this mechanism.
Delegate Thach Phuoc Binh (Vinh Long delegation) agreed with the orientation of promoting Sandbox, open data and computing infrastructure. He proposed prioritizing testing for healthcare, education, finance and public administration; determine criteria for selecting businesses, testing periods and transparency mechanisms.
According to him, the State needs to support small businesses and startups in accessing national computing infrastructure, especially GPUs and supercomputers, through subsidies or free quotas. The delegate also proposed building a national open data warehouse, a fair data sharing mechanism that meets security requirements; and at the same time, implementing specific policies to attract and retain AI talent from university level to international experts. Delegate
Hoang Minh Hieu, in charge of the Law and Justice Committee, spoke at the parliament. Photo: Hoang Phong
Criminal handling of intentional high-risk AI deployment without testing
Delegate Pham Trong Nhan pointed out the fact that China has filed more than 62,000 AI patents in just 10 months, 12 times more than Vietnam's total patents to date; the US attracted more than 40 billion USD in private capital into AI by 2024, while 99% of the platform models Vietnam is using are foreign. This is a strong warning about digital sovereignty and the risk of technological dependence.
In the context of AI being the soft infrastructure of national productivity, can Vietnam accept being just a place to consume other people's data and applications? Mr. Nhan said that if not proactive, Vietnam will be left behind in the global value chain. There are three major existing risks including data inequality and privacy; bias and lack of transparency in AI-supported decisions; and deepfake and cognitive manipulation. In the first 9 months of 2025, the world recorded more than 180,000 political deepfake videos, of which 41% targeted leaders of developing countries.
"Just a few seconds of fake videos can seriously damage the reputation and social security," he said, proposing 6 directions, including applying risk-based management with 4 levels of classification; protecting civil rights; requiring foreign platforms with over 5 million users to establish legal entities in Vietnam; establishing an AI System Safety Committee under the Prime Minister; a real-time update mechanism; and strong enough sanctions, including criminal sanctions for intentionally deploying high-risk AI without testing or transparency.
Delegate Pham Trong Nhan also emphasized that the bill is not just a technical document but a manifesto of Vietnam's development thinking, aiming to turn AI into a new driving force, new strength and new competitiveness of the country.
According to VnExpress