Draft Data Security Law: Categorizing 5 risk levels, limiting administrative procedure generation
At the 2nd Session of the 16th National Assembly, the Government submitted the draft Data Security Law to the National Assembly for consideration and comments.

The draft Data Security Law is structured with 06 chapters and 29 articles
According to the Ministry of Public Security, the drafting agency, the draft Data Security Law consists of 6 chapters and 29 articles, regulating data security protection and the responsibilities of the State, agencies, organizations, enterprises, and individuals in ensuring data security. The focus of the draft is to establish a risk-level management mechanism, determining corresponding protection responsibilities instead of applying the same requirements to all data processing activities.
The draft stipulates 5 levels of data security risk, ranging from level 1, the lowest, to level 5, the highest, determined based on the likelihood of an incident and the potential level of damage caused. Protection obligations are defined by level and applied across digital transformation, digital infrastructure operation, large-scale data processing, the use of new technologies and artificial intelligence, as well as cross-border activities.
On this basis, the draft outlines the responsibilities of each entity. The Government uniformly manages state administration on data security. Agencies, organizations, and enterprises acting as data controllers are responsible for self-determining the risk level of the data they manage, applying proportional protection measures, and bearing legal responsibility. The State primarily conducts ex-post inspections, applying ex-ante inspections only to high-risk data and systems.
The obligations of data controllers are also differentiated based on the scale of operations. For level 1 and 2 data, which account for the majority of routine processing activities, the draft orientation generates virtually no administrative procedures. Household businesses, small and micro enterprises, and innovative startups only need to implement basic protection measures according to general guidelines.
From level 3 and above, data controllers are required to conduct impact assessments and periodic audits. Specifically, level 4 and 5 data must undergo data security assessments along with cybersecurity assessments before systems are officially put into operation. This approach aims to focus management resources on activities with high potential for causing severe damage, while limiting the compliance burden on low-risk activities.
The draft also defines responsibilities and protection mechanisms for individuals in two different capacities. As data subjects, individuals' legal rights and interests continue to be protected under personal data protection laws; if data is processed in high-risk systems, individuals are also entitled to a corresponding additional security protection layer.
For individuals in charge of data security or independent experts, the draft encourages proactively discovering, testing, and reporting data security vulnerabilities in good-faith. Such individuals are directed to have their identity, legal rights, and interests protected, while being exempted from corresponding legal liabilities. This mechanism aims to mobilize additional social resources to participate in discovering and preventing risks, rather than relying solely on the capacity of the state apparatus.
Avoiding procedural overlapping, clearly defining the scope of data management
Another important content is defining the scope of regulation of the Data Security Law in relation to related laws. The principles and conditions for processing personal data and the rights of personal data subjects shall continue to be implemented in accordance with the law on personal data protection. The activities of building, managing, exploiting, sharing, and classifying data according to importance shall fall within the scope of data law; cybersecurity protection for information systems shall be implemented in accordance with cybersecurity law.
Meanwhile, the draft Data Security Law focuses on protecting the security of the data itself throughout its lifecycle. To limit organizations and enterprises from having to perform multiple duplicate procedures for the same management content, the draft establishes the principle that a single subject and a single risk content shall only be evaluated, appraised, and inspected once.
According to this orientation, data must be protected from collection, storage, processing, and transmission until destruction. Cross-border data transfer is also controlled according to risk levels to protect national data sovereignty. The majority of low-risk data is facilitated for smooth circulation with minimal administrative procedures, while high-risk data is subject to stricter protection and control requirements.
The draft adds a mechanism to protect domestic organizations and individuals against demands for direct data provision from abroad that bypass legal channels, in compliance with the principles of respecting national sovereignty, national security, and international practices on mutual legal assistance.
Along with protecting data during processing, the draft emphasizes proactive prevention, monitoring, and response to data security incidents. The State is oriented to have tools for timely response to serious incidents, while the scope and conditions of application must be strictly limited so as not to affect the legitimate rights and interests of organizations and individuals.
Accordingly, restrictive measures on rights shall only be applied when there is a clear basis, prioritizing the least restrictive measures and specifying a fixed timeframe. Their application must be accompanied by reporting and monitoring mechanisms, ensuring the right of organizations and individuals to petition, complain, initiate lawsuits, and receive compensation if measures are implemented unlawfully.
Data protection in AI and supply chain risk management
Regarding new technologies and artificial intelligence (AI), the draft establishes data security protection requirements commensurate with the risk level. Organizations and individuals researching, developing, deploying new technologies or AI systems that process high-risk data must integrate security protection measures right from the design stage; ensuring the legal origin and integrity of the utilized data.
To avoid generating additional procedures, data security assessment for AI systems is expected to be integrated into the risk assessment process under artificial intelligence laws, with no separate dossier and no separate appraisal conducted. The draft also stipulates that competent authorities must not request the handover of source code and model weights, except for the purpose of investigating criminal cases. This approach aims to control risks while still protecting business secrets and technological secrets of enterprises.
The responsibility for ensuring data security is not limited within an organization's internal scope but is extended to the entire supply chain of data products and services, including intermediaries and third parties involved in data processing.
Before connecting with a product or service provider, data controllers must assess risks, limit the scope of data access, and explicitly stipulate confidentiality obligations and penalties for data leaks directly in the contract. During the course of cooperation, data controllers are responsible for proactively monitoring and revoking access rights upon detecting risks or incidents.
For data processing systems with high risk levels, remote access by vendors must be more strictly controlled. This regulation aims to manage risks at intermediary stages, which can become weak points in ensuring data security if responsibilities and access rights are not adequately controlled.
Regarding infrastructure, technology, and human resource development policies, the draft also sets out the principle of not applying measures that create a monopoly position for a specific product or supplier, thereby ensuring a fair competitive environment in the development of the domestic data security industrial ecosystem.
Anh Tho
