Decree 142/2026/ND-CP detailing the Law on Artificial Intelligence

DECREE

Detailing a number of articles, and prescribing measures for implementation, of the Law on Artificial Intelligence

Pursuant to Law No. 63/2025/QH15 on Organization of the Government;

Pursuant to Law No. 134/2025/QH15 on Artificial Intelligence;

At the proposal of the Minister of Science and Technology;

The Government hereby promulgates the Decree detailing a number of articles, and prescribing measures for implementation, of the Law on Artificial Intelligence.

Chapter

GENERAL PROVISIONS

Article 1. Scope of regulation

1. This Decree details a number of articles of the Law on Artificial Intelligence, including:

a) Clause 4, Article 8 on the mechanism for operating, managing and exploiting the single-window artificial intelligence portal and the national database on artificial intelligence systems;

b) Article 9 on classification of risk levels of artificial intelligence systems;

c) Clause 7, Article 10 on the details to be notified, the order and procedures for notification, and technical guidelines on risk classification;

d) Clause 6, Article 11 on the forms of notification and labelling;

dd) Clause 5, Article 12 on reporting, and the responsibilities of relevant agencies, organizations and individuals commensurate with the severity of incidents and the scope of influence of artificial intelligence systems;

e) Article 13 on conformity assessment of high-risk artificial intelligence systems;

g) Article 14 on management of high-risk artificial intelligence systems;

h) Clause 6, Article 16 on the mechanisms for coordinating and sharing, the incentives, and the measures for promoting development of national artificial intelligence infrastructure, appropriate to each stage and to the requirements for ensuring safety and national security;

i) Clause 6, Article 17 on the principles of connection, the mechanisms for sharing and exploiting, and the assurance of safety of, data in databases for artificial intelligence;

k) Clause 5, Article 18 on the mechanisms, criteria and measures for promoting mastery of artificial intelligence technologies, appropriate to each stage of development and to the requirements for ensuring safety and national security;

l) Clause 6, Article 20 on the mechanisms, conditions and procedures for implementing measures to support development of the artificial intelligence ecosystem and market;

m) Article 21 on the regulatory sandbox mechanism for artificial intelligence systems;

n) Clause 4, Article 24 on the criteria, order and procedures for recognition, the operating mechanism of artificial intelligence clusters, and incentive policies;

o) Clause 7, Article 25 on the mechanisms, policies, conditions and procedures for implementing support for enterprises in the field of artificial intelligence;

p) Clause 5, Article 27 on the details, procedures and responsibilities for impact assessment, risk management and supervision of the use of artificial intelligence systems in state governance and the provision of public services.

2. This Decree prescribes measures for implementation of the Law on Artificial Intelligence regarding the transparency obligation and the responsibilities of agencies and organizations.

3. Artificial intelligence activities related to state secrets must also comply with the law on protection of state secrets, the law on cybersecurity and the law on cryptography.

Article 2. Subjects of application

This Decree applies to:

1. Providers, developers, deployers and users of artificial intelligence systems, and persons affected by artificial intelligence systems.

2. Vietnamese authorities, organizations, and individuals; foreign organizations and individuals involved in artificial intelligence activities in Vietnam.

Article 3. Interpretation of terms

In this Decree, the terms below are construed as follows:

1. Technical dossier of a high-risk artificial intelligence system means a set of technical documents commensurate with the risk level of the system, for classification, conformity assessment, supervision and incident handling, including a functional description of the system, a description of training data, testing procedures, risk management measures and other relevant technical information as prescribed by law.

2. Identification code of an artificial intelligence system means an electronic code for dossier management, which is generated automatically on the single-window artificial intelligence portal when the provider notifies the risk classification result.

3. Risk exposure value means the total monetary value of transactions or assets that an organization or individual participating in a sandbox proposes, approves or executes within the scope of the regulatory sandbox.

4. Artificial intelligence model means the algorithmic component of an artificial intelligence system, trained from data to learn, represent and infer the patterns of the data, thereby generating outputs such as predictions, content, recommendations or decisions.

5. Model card means a document accompanying an artificial intelligence model, providing descriptive information on the characteristics of the model, its intended purposes, scope of application, limitations and potential risks, training data, training parameters and conditions, as well as results of performance evaluation of the model, in order to ensure transparency, traceability and reproducibility in the development and use of the model.

6. System card means a document describing an artificial intelligence system, including information on the system architecture, its constituent components, the models used, data used for training or enhancing the system, the purposes and scope of deployment, safety and security assurance measures, and information on relevant incidents, vulnerabilities or remedial measures, in order to ensure transparency, support the assessment and supervision of the system, and promote responsible use of the system.

7. Conditional open data means data that may be accessed and used when an agency, organization or individual meets the conditions on registration, purpose of use and confidentiality requirements as prescribed by law.

8. Commercial data for artificial intelligence means data of organizations, enterprises and individuals that is provided, shared and exploited on the basis of civil and commercial agreements, in compliance with the laws on data, protection of personal data and intellectual property.

Article 4. Single-window artificial intelligence portal and national database on artificial intelligence systems

1. The Ministry of Science and Technology shall manage and operate the single-window artificial intelligence portal to perform the following tasks:

a) To receive notifications of risk level classification results and conformity assessment results; to establish identification codes of artificial intelligence systems; to receive updated information on risk classification dossiers and other information on artificial intelligence systems as prescribed in this Decree;

b) To receive incident reports, periodic reports and information for the management and supervision of artificial intelligence systems as prescribed in this Decree;

c) To support automated risk classification of artificial intelligence systems;

d) To publicly disclose information on artificial intelligence systems, conformity assessment results, the list of recognized artificial intelligence clusters, results of handling of violations, information on the capacity of shared artificial intelligence infrastructure, access conditions, technical regulations, and the methods and procedures for exploiting data, as well as support programs and mechanisms;

dd) To receive requests for guidance; to receive, post for public consultation and publicly disclose impact assessment reports, and to serve the tracking and supervision of the use of artificial intelligence systems in state agencies as prescribed in this Decree;

e) To receive registrations of demand and issue notices on the right to exploit infrastructure for research, development and innovation in the electronic environment.

2. The Ministry of Science and Technology is the data administrator of the national database on artificial intelligence systems, for the purposes of state governance, and of supervision and consolidation of information on artificial intelligence systems.

3. The collection, updating, public disclosure, connection, sharing and exploitation of information on the single-window artificial intelligence portal and the national database on artificial intelligence systems must comply with the law on electronic transactions and must:

a) Be consistent with the purposes of state governance and the scope of information permitted by law;

b) Comply with the laws on cybersecurity, protection of state secrets, business secrets and protection of personal data;

c) Not give rise to administrative procedures, reporting obligations or requests for provision of information for organizations and individuals, other than the cases prescribed in this Decree;

d) Ensure that artificial intelligence systems and databases for the internal operations of Party agencies and political organizations, and artificial intelligence systems on the list of state secrets, national defence and security, are not subject to notification, registration or public data sharing on the single-window artificial intelligence portal.

4. The Ministry of Science and Technology shall issue standards and technical guidelines on data, data models and the regulations on updating, exploiting and sharing data in the national database on artificial intelligence systems, ensuring seamless connection and sharing between the central and local levels.

Chapter II

CLASSIFICATION AND CONFORMITY ASSESSMENT OF ARTIFICIAL INTELLIGENCE SYSTEMS

Article 5. Principles of classification and conformity assessment of artificial intelligence systems

1. The classification of artificial intelligence systems must be based on risk levels as prescribed in the Law on Artificial Intelligence and this Decree.

2. The provider shall classify the artificial intelligence system before putting the system into use.

3. The deployer shall coordinate with the provider in reviewing and reclassifying the artificial intelligence system where the deployment, modification, integration or change of functions or purposes of use gives rise to a new or higher risk of a system that has already been classified.

4. The provider and the deployer shall conduct conformity assessment of high-risk artificial intelligence systems as prescribed in the Law on Artificial Intelligence and this Decree.

5. In case an artificial intelligence system is integrated into a product, goods or service falling within the scope of regulation of sector-specific laws, the law on standards and technical regulations or the law on quality of products and goods, the organization or individual shall comply with such laws for the corresponding product, goods or service, while also fulfilling the specific risk management requirements for the artificial intelligence component as prescribed in the Law on Artificial Intelligence and this Decree.

6. State regulatory agencies in charge of artificial intelligence and sector-specific State regulatory agencies shall ensure the principle of non-duplication in the classification, conformity assessment and inspection of artificial intelligence systems. Results of testing, inspection, certification of conformity with technical regulations, or conformity assessment conducted under sector-specific laws, the law on standards and technical regulations, or the law on product and goods quality may be used to demonstrate compliance with the corresponding requirements prescribed by the law on artificial intelligence; within the scope of the aspects that have been lawfully and fully assessed and remain valid, organizations and individuals shall not be required to re-conduct such assessments.

7. In case the results prescribed in Clause 4 of this Article do not fully cover the specific risk management requirements of the artificial intelligence system as prescribed in the Law on Artificial Intelligence and this Decree, the provider or deployer is only required to perform supplementary assessment of the details not yet assessed or not yet demonstrated.

8. The Ministry of Science and Technology shall provide electronic support tools for self-assessment and classification of artificial intelligence systems on the basis of the criteria prescribed in Article 8 of this Decree. The use of electronic support tools is supportive in nature, is not mandatory, and does not give rise to administrative approval procedures or other obligations beyond the provisions of this Decree.

Article 6. Classification of artificial intelligence systems

1. The provider shall itself classify the artificial intelligence system as prescribed in Clause 1, Article 10 of the Law on Artificial Intelligence before putting the system into use, and shall be legally responsible for the accuracy and truthfulness of the classification result.

2. The classification prescribed in this Article applies only to artificial intelligence systems; it does not apply to artificial intelligence models, except where a model is used as a component of a specific artificial intelligence system.

3. The risk level of an artificial intelligence system is determined as follows:

a) An artificial intelligence system classified as high-risk is a system on the List of high-risk artificial intelligence systems promulgated by the Prime Minister as prescribed in Clause 4, Article 13 of the Law on Artificial Intelligence;

b) An artificial intelligence system classified as medium-risk is a system that does not fall under the case prescribed at Point a of this Clause and falls under the cases prescribed in Article 9 of this Decree;

c) An artificial intelligence system classified as low-risk is a system that does not fall under the cases prescribed at Point a and Point b of this Clause.

4. In case the deployer modifies or integrates the system, or changes its functions or purposes of use compared with the provider's original declaration, thereby giving rise to a new or higher risk, the deployer shall coordinate with the provider in reviewing and reclassifying the risk level.

Article 7. List and procedures for promulgation of high-risk artificial intelligence systems

1. The List of high-risk artificial intelligence systems promulgated by the Prime Minister as prescribed in Clause 4, Article 13 of the Law on Artificial Intelligence includes:

a) The list of artificial intelligence systems determined to be high-risk;

b) The list of high-risk artificial intelligence systems required to obtain conformity certification before being put into use.

2. Responsibilities for reviewing and proposing amendments and supplements to the List prescribed in Clause 1 of this Article are performed as follows:

a) Provincial-level People's Committees, within the scope of their tasks and powers, shall review and evaluate the deployment of artificial intelligence systems in their localities; and propose amendments and supplements to the List and send them to the line ministries for consideration and consolidation;

b) Ministries and ministerial-level agencies, within the scope of their tasks and powers, shall be responsible for reviewing, evaluating and proposing amendments and supplements to the List for artificial intelligence systems in the fields under their management, and send them to the Ministry of Science and Technology for consolidation and appraisal;

c) Central agencies of political organizations and socio-political organizations may propose supplements to the List for high-risk artificial intelligence systems used in the assessment, planning, rotation and appointment of personnel within the political system, for the Prime Minister's consideration and decision;

d) The Ministry of Science and Technology shall assume the prime responsibility for, and coordinate with ministries, ministerial-level agencies and relevant organizations in, consolidating, developing and submitting to the Prime Minister for promulgation, amendment and supplementation of, the List prescribed in Clause 1 of this Article on the basis of the criteria and principles prescribed in Article 8 of this Decree.

Article 8. Criteria and principles for developing the List of high-risk artificial intelligence systems

1. An artificial intelligence system is determined to be high-risk when it meets one or more of the following criteria:

a) Level of impact: The degree to which it may cause damage to life, health, property, human rights, national interests, public interests or national security; the degree of autonomy of the system; the degree of support for making final decisions; and the capacity for human supervision of, and intervention in, the execution of actions;

b) Field of use: Deployment in the essential fields prescribed in Clause 2, Article 6 of the Law on Artificial Intelligence or fields directly related to public interests;

c) Scope of users and scale of influence: The scope of users, the scale of affected subjects, or the degree of connection to critical technical infrastructure systems.

2. An artificial intelligence system, although falling under the criteria prescribed in Clause 1 of this Article, shall not be proposed for inclusion in the List of high-risk artificial intelligence systems when it meets one of the following criteria:

a) It only performs tasks of collecting, processing, standardizing, classifying, translating or improving the quality of data, and does not directly generate decisions affecting the lawful rights and interests of organizations and individuals;

b) It has technical mechanisms and operating procedures ensuring substantive human supervision, whereby the competent person is able to independently review, intervene in, reject or change the system's decision before such decision takes effect;

c) It is used only for the internal administration and operation of an organization or enterprise, and does not directly affect the legal rights, obligations or lawful interests of organizations and individuals outside the organization;

d) It only performs analysis, forecasting, assessment or recommendation functions of a referential nature; the provider and deployer must not use the output as the sole basis for making a final decision.

3. Ministries and ministerial-level agencies shall, based on the criteria prescribed in Clauses 1 and 2 of this Article, propose artificial intelligence systems for inclusion in the List of high-risk artificial intelligence systems.

4. Ministries and ministerial-level agencies shall, based on the List of high-risk artificial intelligence systems prescribed in Article 7 of this Decree and sectoral management requirements, propose the high-risk artificial intelligence systems that must obtain conformity certification before being put into use.

5. In case an artificial intelligence system falls under the case prescribed in Clause 2 of this Article but there is measurable technical or practical evidence showing that the system has serious risks or vulnerabilities capable of directly causing damage to national defence, security, or social order and safety, the Prime Minister shall consider including it in the List of high-risk artificial intelligence systems and clearly state the grounds in the Decision promulgating the List.

Article 9. Classification of medium-risk artificial intelligence systems

1. The provider shall classify an artificial intelligence system into the medium-risk group when the system concurrently satisfies the following conditions:

a) It is not on the List of high-risk artificial intelligence systems promulgated by the Prime Minister as prescribed in Clause 4, Article 13 of the Law on Artificial Intelligence;

b) It is likely to cause confusion, influence or manipulation of users because users are unable to recognize that the interacting subject is an artificial intelligence system or that content is generated by the system, as prescribed at Point b, Clause 1, Article 9 of the Law on Artificial Intelligence.

2. The deployer shall apply Clause 1 of this Article to reclassify an artificial intelligence system in the case prescribed in Clause 2, Article 10 of the Law on Artificial Intelligence.

3. The provider and deployer shall not classify an artificial intelligence system as a medium-risk system when it falls under one of the following cases:

a) The system only supports technical editing to improve the form of content, does not generate new content and does not change the identity of the subject;

b) The system supports office work, where users can clearly recognize from the context of use that it is functionally an artificial intelligence tool, and it has no simulation or imitation function that may cause confusion about identity or the authenticity of events;

c) The system does not interact with, or directly provide services or content to, the public, including cases where it is not released to the public through third parties or intermediary platforms;

d) The system is used in art, cinema, video games or other creative activities, and the context of publication clearly shows that the content is fictional;

dd) The artificial intelligence system only performs data processing or analysis or optimizes operations within a technical system, does not interact directly with users, does not generate content provided to the public, and does not directly interact with the physical environment as the main control function of the system.

Article 10. Automated risk classification support tool

1. The Ministry of Science and Technology shall establish and maintain the Automated risk classification support tool on the single-window artificial intelligence portal to support providers and deployers in self-assessing and classifying artificial intelligence systems.

2. Providers are not obliged to use the Automated risk classification support tool prescribed in Clause 1 of this Article. Competent state agencies must not require providers to submit classification results exported from this Tool as a mandatory component of the dossier in any administrative procedure.

3. Risk classification results from the Automated risk classification support tool are deemed a valid electronic document of the risk classification dossier and one of the practical bases for competent state agencies to consider and assess the provider's compliance with the law on risk classification.

4. In case the risk level cannot be determined after using the Automated risk classification support tool, the provider has the right to request the Ministry of Science and Technology to provide guidance on classification on the basis of the technical dossier as prescribed in Clause 4, Article 10 of the Law on Artificial Intelligence.

Article 11. Review and updating of risk levels

1. The provider shall review and reclassify the risk level of an artificial intelligence system upon the occurrence of any of the following legal events:

a) The system undergoes a significant change in its functions, purposes of use or deployment context that affects the original risk classification criteria of the system;

b) A serious incident occurs in the system, proving that the actual risk level is higher than the classified level;

c) There is a decision of the Prime Minister amending or supplementing the List of high-risk artificial intelligence systems that changes the risk level of the system;

d) There is a written request from the competent state regulatory agency as prescribed in Clause 6, Article 10 of the Law on Artificial Intelligence.

2. The deployer shall coordinate with the provider in reclassifying the artificial intelligence system where the deployer modifies or integrates the system, or changes its functions, thereby giving rise to a new or higher risk compared with the provider's original classification result.

3. Notification of results after review and reclassification is prescribed as follows:

a) In case the reclassification result determines that the risk level of the system is higher than the classified level, the provider and deployer shall notify the reclassification result to the competent state agency within 15 working days from the date of completing the review and shall immediately apply risk management measures corresponding to the new risk level;

b) In case the reclassification result determines that the risk level of the system is lower than the classified level, the provider and deployer have the right to notify the reclassification result to the competent state agency in order to be subject to the management regime corresponding to the new risk level.

4. The provider and deployer are not required to review and notify reclassification when carrying out upgrades, performance optimization, routine technical bug fixes or periodic data updates that do not change the risk nature of the system.

5. In case an artificial intelligence system must be reclassified as high-risk as prescribed at Point a, Clause 1 of this Article, the provider and deployer shall be given a maximum transitional period of 12 months from the effective date of the Prime Minister's decision amending or supplementing the List of high-risk artificial intelligence systems to complete the dossier and conformity assessment procedures as prescribed in the Law on Artificial Intelligence. During the transitional period, the system is permitted to continue operating, but the provider and deployer shall establish and maintain a mechanism for substantive human supervision. The person performing supervision must be provided with sufficient information and authority to independently assess, intervene in or overrule the results of the system; at the same time, the provider and deployer shall fully retain operation logs and intervention decisions to serve inspection and examination.

Article 12. Risk classification dossier of an artificial intelligence system

1. The provider shall prepare a risk classification dossier for high-risk artificial intelligence systems and medium-risk artificial intelligence systems before putting the system into use as prescribed in Clause 1, Article 10 of the Law on Artificial Intelligence.

2. For high-risk and medium-risk artificial intelligence systems, the provider shall ensure that the risk classification dossier includes the following documents:

a) System identification information, including: the name of the system; the version; the internal identification code or the identification code issued by the competent state agency; and the name, address and contact information of the provider;

b) Description of the system and context of use: the purposes of use; main functions; the system architecture at the functional and business level; the scope of deployment; and the groups of users and affected persons;

c) Information on data: A general description of the main types of input data used to operate the system;

d) Details of risk management: A summary of the risk management measures and measures for ensuring the safety and transparency of the system.

3. In case the provider develops an artificial intelligence system based on a platform, model or artificial intelligence service of a third party, the provider is only required to supply technical information and information on data within the scope of its lawful access and control.

4. The preparation, provision and management of the risk classification dossier does not require the provider to disclose source code, model parameter sets, detailed algorithms, raw training data, or information on the list of state secrets, business secrets or technological secrets as prescribed at Point e, Clause 1, Article 14 of the Law on Artificial Intelligence, unless otherwise provided by law.

5. The provider has the right to use model technical description documents or equivalent technical documents prepared under international standards to satisfy the dossier components prescribed in Clause 2 and Clause 3 of this Article, provided that such documents fully meet the information details required by this Decree.

6. The provider and deployer shall retain the classification dossier throughout the operating life of the system.

7. In case an artificial intelligence system uses personal data, the provider may use the Personal data processing impact assessment dossier prepared under the law on protection of personal data to replace, or integrate as a component of, the Risk classification dossier, in order to reduce administrative compliance costs.

Article 13. Conformity assessment of high-risk artificial intelligence systems

1. The provider of a high-risk artificial intelligence system shall conduct conformity assessment before putting the system into use as prescribed in Clause 1, Article 13 of the Law on Artificial Intelligence.

During use, the provider shall conduct conformity reassessment when there is a significant change affecting the original conformity assessment result, including any of the following cases:

a) A change in the main functions, purposes of use or scope of application of the system that gives rise to a new risk or changes the assessed risk level;

b) A change in the system architecture, the artificial intelligence model or the main technical configuration that may affect the accuracy, reliability, safety level or controllability of the system;

c) A change in the data sources, main types of input data or data processing methods that significantly affects the operating results or risk level of the system;

d) Integration of the artificial intelligence system with another system or a new operating environment that may give rise to new risks or change the assessed operating conditions;

dd) Other changes that significantly affect the satisfaction of the requirements prescribed in Article 14 of the Law on Artificial Intelligence.

Periodic data updates, technical bug fixes, performance optimization or version upgrades that do not change the system's satisfaction of the requirements prescribed in Article 14 of the Law on Artificial Intelligence are not considered significant changes under this Clause.

2. The provider shall conduct conformity assessment against the requirements prescribed in Article 14 of the Law on Artificial Intelligence in the following manners:

a) For high-risk artificial intelligence systems on the List of systems required to obtain conformity certification before being put into use, the provider shall conduct conformity assessment through a conformity assessment organization prescribed in Clause 4 of this Article;

b) For high-risk artificial intelligence systems not falling under the case prescribed at Point a of this Clause, the provider may itself conduct conformity assessment or use the services of a conformity assessment organization prescribed in Clause 4 of this Article. In case the provider itself conducts conformity assessment, the provider shall prepare a technical dossier and be legally responsible for the assessment result.

3. In case a high-risk artificial intelligence system falls within the scope of regulation of sector-specific laws that prescribe standards, technical regulations or conformity assessment, the provider may use assessment results under such sector-specific laws to demonstrate conformity with the corresponding requirements prescribed in Article 14 of the Law on Artificial Intelligence. Competent state agencies must not require the provider to re-conduct assessments that have been certified as satisfactory, so as to ensure that no duplicate administrative procedures arise.

4. A conformity assessment organization for high-risk artificial intelligence systems must concurrently satisfy the following requirements:

a) It is established and has registered its conformity assessment activities as prescribed by the law on standards and technical regulations;

b) It ensures independence and objectivity in conformity assessment activities;

c) It satisfies the operating conditions prescribed by the law on standards and technical regulations that are appropriate to the field of assessing artificial intelligence systems;

d) It has a team of experts, personnel and technical facilities meeting professional requirements on artificial intelligence, cybersecurity and data governance in line with the standards, technical regulations or guidelines issued by the Ministry of Science and Technology;

dd) It is subject to periodic supervision by competent state agencies as prescribed by law.

5. Competent state agencies shall recognize conformity assessment results issued by foreign conformity assessment organizations in accordance with the law on standards and technical regulations and with international treaties to which the Socialist Republic of Vietnam is a party.

The provider may use recognized conformity assessment results to demonstrate conformity for the corresponding details that have been assessed. For details that have not been assessed or recognized, the provider shall continue to conduct conformity assessment as prescribed in this Article.

6. A provider of a high-risk artificial intelligence system that has undergone conformity assessment shall maintain conformity and publicly disclose information on the conformity assessment results on the single-window artificial intelligence portal before putting the system into use and when updating reassessment results. The publicly disclosed details include:

a) The name of the artificial intelligence system, the system version and the identification code of the artificial intelligence system, if any;

b) The name, address and contact information of the provider;

c) The conformity assessment method applied;

d) The name of the conformity assessment organization, where the provider conducts conformity assessment through a conformity assessment organization;

dd) The conclusion on the conformity of the artificial intelligence system;

e) The time of completion of the conformity assessment or the time at which the conformity assessment results were recognized;

g) The time of the latest update of information.

Article 14. Notification of risk classification results of artificial intelligence systems

1. The provider of a medium-risk artificial intelligence system or a high-risk artificial intelligence system shall notify the risk classification result to the Ministry of Science and Technology through the single-window artificial intelligence portal before putting the system into use. The provider shall self-declare and be legally responsible for the details of the notification.

2. The provider shall ensure that the details of the risk classification result notification comply with Article 12 of this Decree.

3. The provider shall make the notification electronically on the single-window artificial intelligence portal in one of the following forms:

a) Direct declaration using the electronic form on the single-window artificial intelligence portal;

b) Automatic transmission of information through an application programming interface or another appropriate electronic method as prescribed by law.

4. The Ministry of Science and Technology shall, through the single-window artificial intelligence portal, automatically record the information, issue the identification code of the artificial intelligence system and send an electronic confirmation to the provider immediately after the provider completes sending the notification. In the course of receiving notifications, competent state agencies must not require the provider to submit additional documents or supplement information components beyond the details prescribed in Clause 2 of this Article. Competent state agencies shall conduct inspection, examination and post-examination as prescribed by law.

Article 15. Management of high-risk artificial intelligence systems

1. The provider of a high-risk artificial intelligence system shall establish and maintain a risk management system for the system it provides, appropriate to the purposes of use, scope of deployment and risk level of the system.

2. The risk management system prescribed in Clause 1 of this Article shall include at least the following aspects:

a) Identifying and assessing risks that may arise to human rights, safety, security or public interests in the process of designing, developing and providing the system;

b) Ensuring the quality, suitability and representativeness of training data, testing data and assessment data to the extent necessary to limit risks arising from data;

c) Designing and maintaining a mechanism for human supervision and intervention appropriate to the risk level of the system;

d) Applying technical or management measures to prevent, limit or control the identified risks;

dd) Reviewing and updating risk management measures when the system undergoes a significant change in its model, data, operating method or purpose of use.

3. The provider shall supply the deployer with necessary information on the purposes of use of the system, safe operating conditions, identified risks and corresponding risk management measures, in order to ensure that the system is deployed and used for its intended purposes.

4. The deployer of a high-risk artificial intelligence system shall organize risk management in the course of deploying and operating the system in a manner appropriate to the purposes of use, scope of deployment and risk level of the system and to the provider's technical guidelines.

5. In the course of deployment and operation, the deployer shall implement the following risk management measures:

a) Operating the system in accordance with the purposes of use, conditions of use and limits of use determined by the provider;

b) Organizing supervision of the operation of the system in order to detect deviations, risks or incidents that may arise;

c) Establishing and maintaining a mechanism for human supervision and intervention in the course of using the system as prescribed by law;

d) Applying measures to limit or control risks within the scope of its control when detecting that the system is operating contrary to its purposes of use or that a new risk has arisen;

dd) Coordinating with the provider in reviewing and updating risk management measures when the system undergoes a significant change or when an incident occurs.

6. In case it detects that the system is likely to cause serious damage to human life or health, human rights, property, cybersecurity, social order and safety or public interests, the deployer shall promptly apply risk limitation measures within the scope of its control and notify the provider and the competent state agency as prescribed by law.

Chapter III

TRANSPARENCY, ACCOUNTABILITY AND INCIDENT RESPONSE OF ARTIFICIAL INTELLIGENCE SYSTEMS

Article 16. Principles of transparency and accountability

1. Providers and deployers of artificial intelligence systems shall perform the transparency obligation, notification, technical marking and visible labelling for artificial intelligence systems and content generated by the systems as prescribed in Article 11 of the Law on Artificial Intelligence. The performance of the transparency obligation must be appropriate to the purposes of use, deployment context and risk level of the system.

2. Providers and deployers shall perform the following obligations:

a) To notify, technically mark and visibly label artificial intelligence systems and content generated by the systems as prescribed by law;

b) To provide information on the purposes of use, scope of application, conditions of use and limitations of the artificial intelligence system;

c) To retain information and documents for inspection and supervision as prescribed by law.

3. Providers and deployers shall give explanations upon a written request of the competent state agency as prescribed in Articles 14 and 15 of the Law on Artificial Intelligence. The scope, details and time limit of explanations must be appropriate to the legal status of each subject and the risk level of the system.

4. When performing the responsibility of accountability and providing information as prescribed by law, providers and deployers are not required to disclose source code, detailed algorithms, training data, parameter sets or information within the scope of business secrets or technological secrets, unless otherwise provided by law.

5. The provision of information in the course of transparency and accountability must comply with the law on protection of personal data and the law on data.

6. In case an artificial intelligence system integrates an artificial intelligence model of another organization or individual, the provider shall reach an agreement with the organization or individual providing the model to ensure coordination in providing the necessary technical information for performing the transparency obligation and accountability as prescribed by law.

Article 17. Technical marking of content generated by artificial intelligence systems

1. The provider of an artificial intelligence system shall apply technical solutions so that output content that is audio, images or video is marked in a machine-readable format as prescribed in Clause 2, Article 11 of the Law on Artificial Intelligence. The provider is not required to perform technical marking in a machine-readable format for output content that is text, unless otherwise provided by law.

2. The technical marking prescribed in Clause 1 of this Article must ensure machine-readable identifiability and may be performed by one of the following measures:

a) Integrating identifying signs into the file structure or content data;

b) Integrating identifying signs into the metadata of the file;

c) Using digital signatures, electronic signatures or equivalent technical authentication methods;

d) Applying other technical solutions that ensure the ability to identify content generated or edited by an artificial intelligence system.

3. The identifying sign in machine-readable format must show information confirming that the content was generated or edited by an artificial intelligence system. The provider may additionally integrate information on the provider, the artificial intelligence system or the time of generating or editing the content, if the technical solution permits, in order to serve the authentication of origin and traceability of the content.

4. The provider shall ensure that the technical marking function is maintained in the course of generating, exporting and providing content within the scope of the functions controlled by the system.

5. In case an artificial intelligence system is provided as open source or free of charge, the provider is deemed to have performed the obligation prescribed in Clause 1 of this Article when:

a) The technical marking function is built in; or

b) Tools, standard configurations, application programming interfaces or technical documents enabling the deployer to configure and operate the marking function are publicly disclosed.

When using such a system to provide content to the public, the deployer shall apply technical solutions to mark the output content.

6. The Ministry of Science and Technology shall publicly disclose and update the list of reference standards, technical regulations or technical guidelines for the performance of technical marking prescribed in this Article. The public disclosure prescribed in this Clause does not give rise to administrative procedures, business investment conditions or obligations beyond the provisions of this Decree.

Article 18. Notification and visible labeling of content generated by artificial intelligence systems

1. The deployer shall clearly notify when providing to the public content generated or edited by an artificial intelligence system that is likely to cause confusion about the authenticity of events, characters or the origin of the content, as prescribed in Clause 3, Article 11 of the Law on Artificial Intelligence.

2. The deployer shall apply easily recognizable labels to audio, images or video generated or edited by an artificial intelligence system in the following cases:

a) Simulating or imitating the appearance or voice of a real person;

b) Recreating real events so as to distinguish them from authentic content, unless otherwise provided by law.

3. The notification and visible labelling must satisfy the following requirements:

a) Be clear, comprehensible and easily recognizable to the recipient;

b) Be made before or at the time the recipient accesses the content;

c) Not be designed in a manner that conceals or reduces the ability to recognize the nature of the content;

d) Be appropriate to the type of content and the method of providing the content;

dd) Not significantly obstruct the display, presentation or use of the content.

4. The deployer is not required to perform the obligation of visible labelling in the following cases:

a) The content is technically edited to improve the quality of audio, images or video without changing the nature or main context of the content;

b) Text is processed with tools supporting correction of spelling or grammar, summarization, paraphrasing or translation without distorting the basic content of the original text;

c) The content is used within an agency, organization or enterprise and is not provided to the public;

d) The content is generated in the course of research, development or testing in a controlled environment and is not provided to the public.

5. The deployer has the right to select the form of presenting the notification and visible label appropriate to the type of content and the method of provision, including:

a) Displaying directly on the content;

b) Displaying in the title, description or caption accompanying the content;

c) Displaying on the interface of the platform providing the content;

d) Playing an audio announcement or another appropriate form.

6. For cinematographic works, art programs or creative content, the notification and visible labelling shall be performed by the deployer that directly puts the content into public circulation and may be performed in the opening section, closing section, credits, description or documents accompanying the work, appropriate to the specifics of each type. The notification and labelling must ensure that recipients can clearly recognize that the content was generated or edited by an artificial intelligence system, without causing confusion about the origin of the content.

In case content is generated or edited by an artificial intelligence system in the course of production, the party generating or editing the content shall provide the information necessary for performing the obligation prescribed in this Clause; the party directly putting the content into public circulation shall perform the obligation on the basis of the information provided.

7. The Ministry of Science and Technology shall publicly disclose and update reference technical guidelines on the form of notification and visible labeling prescribed in this Article. The public disclosure prescribed in this Clause does not give rise to administrative procedures, business investment conditions or obligations beyond the provisions of this Decree.

Article 19. Reporting of and response to serious incidents

1. A serious incident of an artificial intelligence system, as prescribed in Clause 8, Article 3 of the Law on Artificial Intelligence, is an event occurring during the operation of an artificial intelligence system that causes one of the following consequences:

a) Loss of human life or serious harm to human health;

b) Significant property damage or serious impact on the operation of an organization;

c) Serious infringement of human rights, or of the lawful rights and interests of agencies, organizations or individuals;

d) Serious disruption of the provision of public services or essential services as prescribed by law, or impact on national security, social order and safety.

2. When a serious incident of an artificial intelligence system occurs, the following subjects are responsible:

a) The deployer and the user shall promptly record the incident, take the necessary measures to limit consequences and notify the provider for coordinated remediation;

b) The provider shall apply the necessary technical measures to prevent, limit and remedy the consequences of the incident;

c) The provider and the deployer shall coordinate in providing information and taking the necessary measures to respond to the incident as prescribed by law.

3. The provider or deployer shall submit a preliminary report on a serious incident to the competent state agency, using Form AI01a for organizations or Form AI01b for individuals in the Appendix issued with this Decree, through the single-window artificial intelligence portal within the following time limits:

a) For urgent serious incidents prescribed at Points a and d, Clause 1 of this Article, and uncontrollable cases prescribed at Point c, Clause 1 of this Article: the preliminary report shall be submitted within 72 hours from the time of incident confirmation;

b) For the remaining serious incidents: the preliminary report shall be submitted within 05 working days from the time of incident confirmation;

c) The time of incident confirmation prescribed in this Clause runs from the moment the organization or individual possesses sufficient initial information to establish that the incident has actually occurred and bears a high likelihood of originating from a malfunction of the artificial intelligence system, without waiting for the completion of a comprehensive investigation into the technical cause. The submission of a preliminary report within the prescribed time limit is not deemed an admission of technical fault or legal liability of the reporting organization or individual.

In case the deployer is unable to contact the provider, the deployer shall make the report as prescribed in this Clause.

4. The provider and deployer shall maintain and retain system logs, data and information related to the incident for verification, assessment and remediation of the incident; and shall also send an official report on the remediation results to the competent state agency within 15 days from the date of submitting the preliminary report.

5. In case a serious incident is concurrently subject to reporting obligations under the laws on cybersecurity, personal data protection, or other sector-specific laws, the reporting shall be carried out in accordance with such laws.

Competent state agencies shall coordinate and share information, and shall not require re-provision of information and documents already available in information systems or databases connected and shared as prescribed by law, or in the artificial intelligence systems and databases prescribed at Point d, Clause 3, Article 4 of this Decree.

6. The competent state agency shall receive reports, organize verification and provide guidance on responding to serious incidents; when necessary, it has the right to request suspension of operation, recall or reassessment of the artificial intelligence system as prescribed in Clause 2, Article 12 of the Law on Artificial Intelligence.

7. In case an incident shows signs of affecting national security, social order and safety, or signs of violating criminal law, the Ministry of Science and Technology shall coordinate with the Ministry of Public Security and relevant agencies to conduct verification and handling in accordance with law.

8. The reporting and receipt of reports of serious incidents shall be carried out through the single-window artificial intelligence portal.

Article 20. Impact assessment when using artificial intelligence systems in state agencies

1. State agencies shall conduct impact assessment of artificial intelligence systems that fall under one of the following cases:

a) The system belongs to the group of high-risk artificial intelligence systems as prescribed by the Law;

b) The artificial intelligence system prescribed in Clause 7 of this Article whose output is used as a direct basis for the competent person to consider and decide on issuing an administrative decision.

2. In case an artificial intelligence system prescribed in Clause 1 of this Article undergoes a change in purpose of use, main functions, main input data sources or scope of subjects of application that gives rise to a new risk or changes the risk level of the system, the deploying agency or unit shall conduct a supplementary impact assessment before continuing to use the system.

3. The deploying agency or unit shall prepare an impact assessment report using Form AI02 in the Appendix issued with this Decree. The impact assessment report must satisfy Clause 3, Article 27 of the Law on Artificial Intelligence and include the following details:

a) Information describing the artificial intelligence system and its purposes of use;

b) Risk identification and assessment;

c) Risk control and mitigation measures;

d) The mechanism for ensuring the capacity for human supervision and intervention in the course of operating the system.

4. The head of the deploying agency or unit shall organize the preparation and approval of the impact assessment report before putting the artificial intelligence system into use, and shall be legally responsible for the details, truthfulness and completeness of the report.

5. The deploying agency or unit shall publicly disclose the impact assessment report as prescribed in Clause 4, Article 27 of the Law on Artificial Intelligence, except for details constituting state secrets, business secrets or personal data as prescribed by law.

6. The use of artificial intelligence systems in state agencies must comply with the national framework of artificial intelligence ethics and must not replace the decision-making authority and responsibility of the competent person as prescribed in Clauses 1 and 2, Article 27 of the Law on Artificial Intelligence.

7. An artificial intelligence system deployed by a state agency is determined to be related to human rights, social justice or public interests when it falls under one of the following cases:

a) The output of the system is used as a basis in the course of consideration and decision-making in state governance activities or the provision of public services;

b) It is used to classify, score, assess or rank organizations or individuals;

c) It is used to allocate budgets or public resources, or to determine the entitlement levels and beneficiaries of policies and regimes;

d) It is used to detect, select, screen or monitor organizations or individuals for state governance activities.

8. The Ministry of Science and Technology shall track and supervise the performance of impact assessment for the use of artificial intelligence systems in state agencies; and consolidate and report to the Prime Minister when necessary.

9. Funds for conducting impact assessment shall be ensured from the state budget in accordance with budget decentralization, lawful revenues of public non-business units and other lawful funding sources as prescribed by law.

Chapter IV

REGULATORY SANDBOX MECHANISM

Article 21. Principles of implementing the regulatory sandbox mechanism for artificial intelligence systems

1. The implementation of the regulatory sandbox mechanism for artificial intelligence systems shall be applied in accordance with Article 21 of the Law on Artificial Intelligence, the law on science, technology and innovation and the provisions of this Decree.

2. Participation in the regulatory sandbox mechanism for artificial intelligence systems shall be on a voluntary basis. Organizations and individuals participating in testing shall comply with the following principles:

a) To clearly determine and comply with the scope, scale, subjects, duration and conditions of testing and the risk management measures as required by the competent state agency;

b) To protect the lawful rights and interests of organizations and individuals participating in testing;

c) To ensure transparency of information to testing participants regarding the objectives, scope, conditions and risks that may arise in the course of testing;

d) To comply with the laws on national security, social order and safety, data, protection of personal data and cybersecurity and other relevant laws;

dd) To be subject to tracking and supervision by the competent state agency throughout the testing process.

3. Research, development or testing of artificial intelligence systems in simulated environments, closed environments or internal testing without real testing participants and without producing impacts outside the organization does not fall within the scope of application of the regulatory sandbox mechanism prescribed in this Decree.

4. Organizations and individuals participating in testing shall:

a) Develop and implement a risk management plan;

b) Publicly disclose risks that may arise and establish a mechanism for receiving and handling complaints of testing participants.

5. The results of completing a regulatory sandbox are one of the bases for the competent state agency to consider and decide on:

a) Recognizing all or part of the testing results for conformity assessment of the artificial intelligence system;

b) Exempting, reducing or adjusting the application of compliance obligations under the law on artificial intelligence.

6. Organizations and individuals participating in testing are exempted or excluded from, and shall bear, civil, criminal and administrative liability arising in the course of testing as prescribed in Articles 22 and 23 of the Law on Science, Technology and Innovation.

7. Within the scope, time and conditions of testing recorded in the Certificate of participation in testing, organizations and individuals participating in testing may be subject to the following support mechanisms as decided by the competent state agency:

a) Application of alternative notification or transparency measures for certain obligations on risk classification, technical marking or visible labelling under the Law on Artificial Intelligence, if it is ensured that testing participants are fully and clearly informed;

b) In case an artificial intelligence system is subject to conformity assessment under the Law on Artificial Intelligence, the conformity assessment may be conducted after completing the testing phase;

c) Priority access to support programs, resources and policies supporting the development of artificial intelligence as prescribed by law.

8. In case sector-specific laws already prescribe a regulatory sandbox mechanism for products, services or business models in a sector-specific management field, the testing of artificial intelligence systems in that field shall be carried out under the sandbox mechanism of the sector-specific laws. State regulatory agencies in charge of artificial intelligence shall coordinate with sector-specific regulatory agencies in assessing, supervising and managing risks of the artificial intelligence component of the tested system.

9. Party agencies, the Vietnam Fatherland Front Committee and socio-political organizations are encouraged to voluntarily apply the provisions on regulatory sandboxes in this Chapter to artificial intelligence systems for internal operations. Competent agencies of the Party shall decide the scope and scale and organize the appraisal and assessment of testing results independently of the provisions on administrative procedures in this Decree.

Article 22. Classification of regulatory sandbox levels

1. Artificial intelligence systems participating in the regulatory sandbox mechanism are classified into 03 testing levels on the basis of a comprehensive assessment of the following factors:

a) The risk level of the artificial intelligence system as prescribed in the Law on Artificial Intelligence;

b) The nature of the data used, including personal data, sensitive personal data, children's personal data or data on the restricted list under the law on data and protection of personal data;

c) The scope and scale of testing deployment, including the deployment area, the number of testing sites, testing participants and the degree of connection to data systems or information systems;

d) The degree of impact on national security, social order and safety, and the lawful rights and interests of organizations and individuals.

2. In case an artificial intelligence system concurrently meets the criteria of several different testing levels, the system shall be classified at the highest testing level.

The classification of testing levels shall be determined on the basis of the functions, purposes of use and overall impact of the artificial intelligence system, including where the system is deployed through multiple components, modules or services combined together.

3. An artificial intelligence system is classified as testing level 3 when its participants, deployment location and testing duration are specifically determined and it falls under one of the following cases:

a) It processes sensitive personal data, core data, important data or children's personal data in cases subject to a personal data processing impact assessment under the law on protection of personal data;

b) The system processes data on the list of state secrets as prescribed by law;

c) The system is deployed in, or directly connected to, an information system on the list of information systems critical to national security as prescribed by the law on cybersecurity.

4. An artificial intelligence system is classified as testing level 2 when its participants, deployment location and testing duration are specifically determined and it meets the following conditions:

a) Real users or persons directly affected by the system participate;

b) The output of the system is used in actual operations, including where the output is integrated into the process of providing services or handling work, is provided to users or is used as a reference basis in the decision-making process, but does not directly give rise to, change or terminate the rights or obligations of organizations or individuals;

c) It does not fall under the case prescribed in Clause 3 of this Article.

5. An artificial intelligence system is classified as testing level 1 when its deployment location and testing duration are specifically determined and it meets the following conditions:

a) Testing participants are determined according to specific criteria;

b) The output of the system is not used in actual operations, is not integrated into the process of providing services or handling work, is not provided to users and is not used as a reference basis in the decision-making process;

c) It does not fall under the cases prescribed in Clauses 3 and 4 of this Article.

6. The competent state agency shall determine the regulatory sandbox level for an artificial intelligence system on the basis of the criteria prescribed in this Article upon receiving and processing the dossier for participation in the regulatory sandbox mechanism.

Article 23. Competence and procedures for approving participation in the regulatory sandbox mechanism

1. The competent state agency shall approve participation in the regulatory sandbox mechanism for artificial intelligence systems as prescribed in Article 21 of the Law on Artificial Intelligence and this Decree.

The certificate is the basis for organizations and individuals to carry out testing within the approved scope. Within the scope of testing, organizations and individuals shall be subject to the mechanism of exemption, reduction or adjustment of compliance with a number of relevant legal provisions as prescribed in Article 22 of the Law on Artificial Intelligence and the details of the certificate.

2. Organizations and individuals shall submit 01 set of dossiers by electronic means through the National Public Service Portal for receipt and processing as prescribed by law. The dossier shall be interconnected and synchronized to the single-window artificial intelligence portal to perform other state governance tasks as prescribed in this Decree.

3. The competence to receive, appraise and issue certificates is determined as follows:

a) Provincial-level People's Committees have the competence to receive, appraise and issue certificates for level 1 or level 2 artificial intelligence systems deployed within the scope of one province or centrally run city;

b) Ministries and ministerial-level agencies have the competence to receive, appraise and issue certificates only for level 1 or level 2 artificial intelligence systems when such systems fall within the state governance scope of the ministry or ministerial-level agency and are deployed in two or more provinces or centrally run cities; or are deployed by agencies or organizations under the direct management of the ministry or ministerial-level agency;

c) The Ministry of Public Security has the competence to receive, appraise and issue certificates for level 3 artificial intelligence systems.

In case an artificial intelligence system falls within the management scope of multiple ministries and ministerial-level agencies, the ministry or ministerial-level agency managing the sector or field that has direct management over the main use of the system shall have the competence to receive, appraise and issue the certificate.

4. Within 03 working days from the date of receiving the dossier, the competent agency shall check the validity of the dossier and notify once, in an electronic written document, any request for amendment or supplementation of the dossier.

5. For a valid dossier, the competent agency shall organize appraisal and issue the certificate or a written refusal within the following time limits:

a) No more than 10 working days in the case prescribed at Point a of Clause 3 of this Article;

b) No more than 20 working days in the case prescribed at Point b, Clause 3 of this Article;

c) No more than 30 working days in the case prescribed at Point c, Clause 3 of this Article.

During appraisal, the competent agency may consult relevant state regulatory agencies on details within the management scope of those agencies when such details affect the appraisal result.

6. In case of refusal of approval, the competent agency shall reply in writing and clearly state the reasons.

7. The Certificate of participation in the regulatory sandbox shall be made using Form AI09a for organizations or Form AI09b for individuals in the Appendix issued with this Decree and must clearly specify the following details:

a) The scope, area and duration of testing;

b) The scale and limits of testing;

c) Technical requirements, and requirements for ensuring safety, security, data protection and risk management;

d) Reporting responsibilities and supervision mechanism;

dd) Cases of suspension or termination of testing;

e) The details of exemption, reduction or adjustment of compliance with legal provisions applicable within the scope of testing.

8. The certificate is valid only within the scope of testing and does not give rise to the right to officially deploy the artificial intelligence system outside such scope.

Article 24. Conditions and dossiers for participation in testing

1. An organization or individual registering to participate in the regulatory sandbox shall satisfy the following conditions:

a) He/she/it must have an artificial intelligence system or artificial intelligence solution proposed for testing that involves innovation, application of new technology or a new deployment model;

b) He/she/it must have a testing plan that clearly specifies the objectives, scope, testing duration, participants and risk control measures appropriate to the testing level prescribed in Article 22 of this Decree;

c) He/she/it must have measures to protect the lawful rights and interests of organizations and individuals affected in the course of testing;

d) In case the artificial intelligence system is capable of directly causing damage to human life or health or large-scale property damage, the organization or individual participating in testing shall obtain civil liability insurance or an equivalent financial guarantee measure appropriate to the scope of testing.

2. The dossier for registration to participate in the regulatory sandbox mechanism includes the following documents:

a) The application for participation in the regulatory sandbox mechanism, using Form AI03a for organizations or Form AI03b for individuals in the Appendix issued with this Decree;

b) The testing plan, describing the artificial intelligence system proposed for testing, the objectives, scope and duration of testing, and the risk management plan;

c) A document describing measures to protect the lawful rights and interests of organizations and individuals affected in the course of testing;

d) A document describing technical capacity, personnel or infrastructure for the deployment of testing that is directly related to the proposed scope of testing.

3. Innovative startups, small and medium-sized enterprises, science and technology organizations and research groups registering to participate in a regulatory sandbox for an artificial intelligence system at testing level 1 shall apply a simplified dossier, including the following documents:

a) The application for participation in the regulatory sandbox mechanism, using Form AI03a for organizations or Form AI03b for individuals in the Appendix issued with this Decree;

b) A document giving a general description of the artificial intelligence system proposed for testing, the main risks that may arise and risk mitigation measures in the course of testing.

4. The dossiers prescribed in Clauses 2 and 3 of this Article shall be submitted electronically as prescribed in Article 23 of this Decree.

5. Competent state agencies must not require organizations and individuals to provide additional dossier components beyond those prescribed in this Article, except where sector-specific laws directly prescribe the dossier components for testing activities in the respective field.

Article 25. Supervision and reporting in the course of the regulatory sandbox

1. During the testing period under the certificate of participation in the regulatory sandbox mechanism, the agency that issued the certificate shall track and supervise testing activities through the following measures:

a) Receiving and assessing periodic reports, incident reports, limit-exceedance reports and testing summary reports as prescribed in this Decree;

b) Requiring organizations and individuals participating in testing to supplement risk control measures, or proposing adjustment of the scope of testing for consideration within its competence, when risks arise outside the scope of testing;

c) Deciding to suspend testing, terminate testing or terminate the validity of the certificate of participation in the regulatory sandbox mechanism when the organization or individual violates the testing limits or fails to remedy an incident as required by the competent state agency.

2. Organizations and individuals participating in testing shall submit periodic reports on the testing situation through the single-window artificial intelligence portal, using Form AI05a for organizations or Form AI05b for individuals in the Appendix issued with this Decree, at the following intervals:

a) Levels 1 and 2: once every 06 months;

b) Level 3: once every 03 months.

3. Upon the occurrence of any of the following cases, the organization or individual participating in testing shall immediately apply remedial measures and submit an ad hoc report within 72 hours through the single-window artificial intelligence portal:

a) The artificial intelligence system has a serious incident under Clause 8, Article 3 of the Law on Artificial Intelligence;

b) The artificial intelligence system exceeds the testing limits determined in the certificate.

An incident report shall be made using Form AI06a for organizations or Form AI06b for individuals; a report on exceeding testing limits shall be made using Form AI07a for organizations or Form AI07b for individuals in the Appendix issued with this Decree.

4. No later than 15 days before the end of the testing period, the organization or individual participating in testing shall submit a summary report on testing results through the single-window artificial intelligence portal, using Form AI08a for organizations or Form AI08b for individuals in the Appendix issued with this Decree.

Article 26. Financial support for the regulatory sandbox

1. Enterprises participating in the regulatory sandbox mechanism, and innovative startups, small and medium-sized enterprises, science and technology organizations and research groups with feasible innovation projects, shall be considered for the issuance of Artificial Intelligence Development Support Vouchers as prescribed in this Decree.

2. Financial support for regulatory sandbox activities must comply with the following principles:

a) It must be implemented under a co-payment mechanism between the State and the organizations and individuals participating in testing;

b) It must ensure public disclosure, transparency, and use for the right purposes, and must avoid duplication with forms of support from the state budget and state financial funds outside the budget;

c) It must support only costs directly for regulatory sandbox activities within the scope of testing approved by the competent agency;

d) It must be implemented within the resources of the National Artificial Intelligence Development Fund and other lawful financial sources as prescribed by relevant laws.

3. Artificial Intelligence Development Support Vouchers are used to pay part of the costs of services directly for regulatory sandbox activities, including:

a) Renting and using computing and storage infrastructure and platform services for the training, testing or operation of artificial intelligence systems;

b) Renting the right to use shared data and artificial intelligence models, including large language models for Vietnamese and ethnic minority languages;

c) Using technical consulting services, safety testing, security assessment, risk assessment or other technical services for regulatory sandbox activities.

The maximum level of support shall not exceed 50% of the total actual eligible costs incurred for the services prescribed in this Clause.

4. Eligible costs are costs incurred directly to serve regulatory sandbox activities, determined on the basis of:

a) Service provision contracts;

b) Lawful invoices and vouchers;

c) The laws on accounting, taxation and the state budget.

5. Payment of Artificial Intelligence Development Support Vouchers shall be made directly to the service-providing organization on the basis of:

a) The actual volume of services provided;

b) Contracts, invoices and documents confirming the provision of services;

c) Reconciliation data on the single-window artificial intelligence portal as prescribed in this Decree.

6. The issuance, use, payment and supervision of Artificial Intelligence Development Support Vouchers shall comply with this Decree and relevant laws, and must not give rise to administrative procedures, business investment conditions or obligations for organizations and individuals beyond the provisions of the Law on Artificial Intelligence and this Decree.

Article 27. Completion of testing and post-testing transition

1. Within 15 working days from the date of receiving the summary report as prescribed in this Decree, the agency that issued the certificate of participation in the regulatory sandbox mechanism shall organize assessment of the testing results and issue the Certificate of completion of the regulatory sandbox, using Form AI04 in the Appendix issued with this Decree. In case it refuses to issue the Certificate, the competent agency shall reply in writing and clearly state the reasons.

In case an extension of the testing period is needed, no later than 15 working days before the end of the period, the organization or individual shall submit an Application for extension using the form in Appendix AI03c or AI03d and a Summary report on the testing process using the form in Appendix AI08a or AI08b. Within 10 working days, the competent agency shall appraise, decide on the extension and issue the extension of the Certificate of participation in testing; where it does not approve or does not extend the Certificate of participation in testing, it shall reply in writing and clearly state the reasons.

2. The Certificate of completion of the regulatory sandbox is the basis for the competent state agency to consider and decide, in accordance with the Law on Artificial Intelligence, on:

a) Recognizing all or part of the testing results for conformity assessment of the artificial intelligence system;

b) Applying measures of exemption, reduction or adjustment of compliance obligations.

3. Organizations and individuals participating in testing may continue to operate the artificial intelligence system during a transitional period of no more than 12 months from the date of issuance of the Certificate of completion of the regulatory sandbox, within the scope, conditions, limits and risk management measures determined in the certificate of participation in the regulatory sandbox mechanism and the Certificate of completion of the regulatory sandbox.

4. During the transitional operation period, organizations and individuals may adjust organizational operating measures, technical parameters and the technical manner of operating the system within the determined scope, conditions, limits and risk management measures, on the basis of the approved risk management plan and ensuring the ability to control risks, provided that the main purpose of use and the risk level of the system under the approved dossier are not changed.

5. In case an adjustment exceeds the determined scope, conditions or limits, it is deemed an expansion of the scope of operation, and consideration and approval shall be carried out as prescribed in Article 23 of this Decree. In case an expansion changes the geographical scope, the management field or the testing level, leading to a change in competence under Article 23 of this Decree, the organization or individual shall request the corresponding competent agency to consider and approve before carrying it out.

6. In case an extension of the transitional operation period is needed, no later than 15 days before the expiry of the period prescribed in Clause 3 of this Article, the organization or individual shall send a request to the agency that issued the Certificate of completion of the regulatory sandbox. Within 07 working days from the date of receiving a valid request, that agency shall consider and decide on a single extension of no more than 06 months when the following conditions are met:

a) The testing has been completed and incidents have been remedied as required by the competent state agency;

b) The approved risk management measures continue to be maintained.

7. Upon expiry of the transitional operation period, organizations and individuals shall complete their compliance obligations under the Law on Artificial Intelligence, this Decree and relevant laws before continuing to provide products or services or to operate the artificial intelligence system.

8. The Certificate of completion of the regulatory sandbox does not replace licenses, certificates, approval documents or business investment conditions under the law.

9. Failure to comply with Clause 7 of this Article shall be handled in accordance with the law on handling of administrative violations and relevant laws.

Chapter V

INFRASTRUCTURE, DATA AND ENSURING SELF-RELIANCE CAPACITY IN ARTIFICIAL INTELLIGENCE DEVELOPMENT

Article 28. National artificial intelligence infrastructure

1. National artificial intelligence infrastructure is organized under a combined model of State-invested infrastructure, including infrastructure of ministries, ministerial-level agencies and provincial-level People's Committees, and infrastructure invested in and developed by enterprises and social organizations.

2. State-invested artificial intelligence infrastructure is organized and assigned for management as follows:

a) For infrastructure used for research, development and innovation, the Ministry of Science and Technology shall assume the prime responsibility for organizing the construction, operation and coordination of connection, sharing and exploitation of the components of this infrastructure in the national artificial intelligence infrastructure network;

b) For artificial intelligence infrastructure for the purpose of state governance, it is organized under a combined, unified, synchronized model with connection and sharing between artificial intelligence infrastructure that the Ministry of Public Security assumes the prime responsibility for organizing the construction and centralized operation of at national data centers, and artificial intelligence infrastructure that ministries, ministerial-level agencies and provincial-level People's Committees build and manage in line with the demands and actual conditions of each sector, field and locality.

3. Artificial intelligence infrastructure for research, development and innovation includes components invested in by the State, commissioned for development or leased as services, including:

a) Computing capacity, storage and shared data for the training, testing and assessment of artificial intelligence systems;

b) Platforms and environments for the development, training, testing and assessment of artificial intelligence systems;

c) Shared artificial intelligence models, including foundation models, general-purpose artificial intelligence models and language models for Vietnamese and ethnic minority languages;

d) Other infrastructure components for research, development and innovation as decided by the competent level.

4. Artificial intelligence infrastructure for state governance is one of the key and core components for the management and development of the national artificial intelligence system, including:

a) Computing capacity, storage and data;

b) Specialized artificial intelligence platforms for the provision of public services and ensuring social order and safety;

c) Other infrastructure components for state governance as decided by the competent level.

5. The Ministry of Science and Technology shall assume the prime responsibility for organizing the construction, operation and exploitation of the infrastructure components prescribed in Clause 3 of this Article; the provision of access and exploitation rights shall be carried out under the mechanism of providing public services as prescribed by law.

6. The Ministry of Public Security shall assume the prime responsibility for organizing the construction, operation and provision of artificial intelligence infrastructure for state governance and other purposes assigned by the Government, concentrated at the national data centers prescribed in Clause 4 of this Article; the provision of access and exploitation rights shall be carried out in accordance with the law on artificial intelligence, the law on data and relevant laws.

7. Ministries, ministerial-level agencies and provincial-level People's Committees shall assume the prime responsibility for building and operating artificial intelligence infrastructure for state governance according to their functions and tasks, on the basis of agreement with the Ministry of Science and Technology and the Ministry of Public Security, ensuring no duplicated or wasteful investment and compliance with the principles in Article 29 of this Decree.

Article 29. Principles of developing, coordinating and mobilizing resources for national artificial intelligence infrastructure

1. National artificial intelligence infrastructure shall be developed in a unified, open and safe orientation, capable of connection, sharing and expansion, in line with national strategies, plans and programs on science, technology, innovation and digital transformation, while meeting requirements of national defence, security and safety.

2. Agencies, organizations and individuals participating in investing in, building, managing, operating or providing services of artificial intelligence infrastructure in the national artificial intelligence infrastructure network shall comply with the following principles:

a) To ensure connectivity, sharing and technical compatibility under the technical regulations and technical connection requirements issued or publicly disclosed by competent state agencies; the application of standards shall comply with the law on standards and technical regulations;

b) Not to impose technical or commercial conditions that restrict the right of agencies, organizations and individuals using the infrastructure to choose technological solutions, artificial intelligence models or deployment methods, unless otherwise provided by law;

c) To ensure cybersecurity, national security, data protection, protection of state secrets and business secrets, and the lawful rights and interests of agencies, organizations and individuals as prescribed by law;

d) To connect, share and effectively exploit existing infrastructure resources in order to limit duplicated investment;

dd) To comply with the law on economical and efficient use of energy. Agencies, organizations and individuals shall implement solutions to reduce greenhouse gas emissions. Operating activities must not threaten the safety of the national power system.

3. The Ministry of Public Security shall assume the prime responsibility for, and coordinate with the Ministry of Science and Technology, ministries, ministerial-level agencies and relevant agencies in, coordinating national artificial intelligence infrastructure on the basis of strategies, plans and orientations for artificial intelligence infrastructure development, including the following details:

a) Coordinating the connection, sharing and exploitation of artificial intelligence infrastructure in the national artificial intelligence infrastructure network;

b) Reviewing and consolidating demand; assuming the prime responsibility for developing and proposing plans for investment in, use, sharing and exploitation of infrastructure in order to ensure efficiency and avoid duplicated investment;

c) Organizing the public disclosure of information on the capacity of shared artificial intelligence infrastructure.

4. The Ministry of Public Security shall assume the prime responsibility for organizing the coordination, operation and exploitation of artificial intelligence infrastructure for state governance activities, deployed in a centralized manner at national data centers as prescribed in Clause 2 of this Article and relevant laws.

5. The State encourages the mobilization of social resources to invest in, build and develop artificial intelligence infrastructure through investment under the public-private partnership method, leasing of services and other forms of cooperation in accordance with the law on investment under the public-private partnership method and relevant laws.

Article 30. Connection, sharing and exploitation of national artificial intelligence infrastructure

1. Agencies, organizations and individuals managing and operating artificial intelligence infrastructure in the national artificial intelligence infrastructure network shall connect, share and exploit the infrastructure in order to effectively use computing capacity, data and artificial intelligence platforms, for state governance, research, development and innovation in the field of artificial intelligence, according to the following principles:

a) For State-invested artificial intelligence infrastructure, the agency managing and operating the infrastructure shall connect, and publicly disclose information on infrastructure capacity, scope of provision, methods of exploitation and technical requirements, within the scope of its functions, tasks and powers and in accordance with relevant laws;

b) For artificial intelligence infrastructure invested in by organizations and enterprises, participation in connecting, sharing and exploiting the infrastructure shall be voluntary, through agreements or contracts in accordance with law;

c) The connection, sharing and exploitation of infrastructure shall not change the ownership regime or the lawful management rights and exploitation rights over the infrastructure and related resources, unless otherwise provided by law.

2. Agencies, organizations and individuals that need to use artificial intelligence infrastructure in the national artificial intelligence infrastructure network shall access and use the infrastructure by the methods publicly disclosed by the agency managing and operating the infrastructure or through the single-window artificial intelligence portal.

The provision and use of infrastructure shall comply with the laws on provision of public services; science, technology and innovation; the state budget; public investment; investment under the public-private partnership method; and relevant laws; and shall not give rise to new administrative procedures.

3. Agencies, organizations and enterprises managing and operating artificial intelligence infrastructure, when participating in connecting, sharing and exploiting infrastructure in the national artificial intelligence infrastructure network, shall:

a) Comply with the standards, technical regulations and reference technical specifications issued or publicly disclosed by competent state agencies;

b) Ensure cybersecurity, data protection, state secrets, business secrets and the lawful rights and interests of agencies, organizations and individuals as prescribed by law;

c) Apply appropriate technical and management measures to ensure system safety and control access to, and exploitation of, the infrastructure.

Article 31. Standards, technical regulations and inspection of national artificial intelligence infrastructure

1. The Ministry of Science and Technology is the point of contact and shall assume the prime responsibility for, and coordinate with the Ministry of Public Security and relevant state agencies in, within the scope of its functions, tasks and powers, issuing or publicly disclosing in accordance with the law on standards and technical regulations:

a) National technical regulations mandatorily applicable to artificial intelligence infrastructure participating in the national artificial intelligence infrastructure network;

b) Technical connection requirements and reference technical specifications for the connection, sharing and exploitation of artificial intelligence infrastructure;

c) Technical criteria on cybersecurity for artificial intelligence infrastructure in accordance with the law on cybersecurity;

d) Standards applicable to artificial intelligence infrastructure in accordance with the law on standards and technical regulations.

2. Agencies, organizations and enterprises managing and operating artificial intelligence infrastructure, when participating in the national artificial intelligence infrastructure network, shall comply with the mandatory national technical regulations, technical criteria on cybersecurity and technical connection requirements issued or publicly disclosed under Clause 1 of this Article. The application of standards to artificial intelligence infrastructure shall be carried out in accordance with the law on standards and technical regulations.

3. Inspection, examination and handling of violations in connecting, sharing and exploiting artificial intelligence infrastructure shall be carried out in accordance with the law on inspection, the law on handling of administrative violations and relevant sector-specific laws.

Article 32. Databases for artificial intelligence

1. Databases for artificial intelligence shall be organized on the principles of openness, safety and controlled access to serve research, training, testing, assessment and development of artificial intelligence applications; and include the following types of data:

a) Open data;

b) Conditional open data;

c) Commercial data of organizations and enterprises as prescribed by law.

2. The Ministry of Public Security is the managing agency of the national database on artificial intelligence and has the following responsibilities:

a) To be responsible for building, managing and operating the national database on artificial intelligence at the National Data Centre;

b) To assume the prime responsibility for developing standards on connection and sharing and technical requirements ensuring security and safety for the national database on artificial intelligence;

c) To assume the prime responsibility for, and coordinate with ministries and ministerial-level agencies in, submitting to the Prime Minister for promulgation, and updating, the List of datasets for the development of artificial intelligence in essential fields.

3. Ministries and ministerial-level agencies shall organize the development and promulgation, or submit for promulgation, of standards and technical regulations for data for the development of artificial intelligence within the scope of their sectoral and field management as prescribed by law.

4. The National Data Centre shall provide technical infrastructure and storage capacity and ensure the technical conditions for operating the national database on artificial intelligence as prescribed by law.

5. Ministries, ministerial-level agencies and provincial-level People's Committees shall:

a) Develop and update open data, conditional open data and commercial data within their management scope;

b) Organize the collection, creation, updating, quality assurance, labelling, annotation and standardization of datasets on the List of datasets for the development of artificial intelligence in essential fields;

c) Carry out unified connection and data sharing and exploit the national database on artificial intelligence as prescribed by law.

6. The state budget shall be used to carry out activities of collecting, creating, ensuring the quality of, standardizing, labelling and annotating datasets and databases for artificial intelligence through the following methods:

a) Implementing information technology application investment projects;

b) Performing science, technology and innovation tasks;

c) Implementing the mechanisms of State ordering, assignment of tasks and bidding for the provision of products and services, and other methods as prescribed by law.

Article 33. Mechanism for supporting artificial intelligence development through voluntary sharing

1. Organizations, enterprises and individuals that voluntarily share data, artificial intelligence models, tools, software or research results for artificial intelligence development, in compliance with the laws on data, protection of personal data and intellectual property, shall be considered for support under the criteria prescribed in Clause 3 of this Article and within the support resources prescribed by law.

2. The support prescribed in this Article is provided in non-financial forms, including:

a) Priority access to and use of computing infrastructure and data storage infrastructure under the mechanism for exploiting and using artificial intelligence infrastructure prescribed in this Decree and relevant laws;

b) Priority access to and use of technical services for data processing, training, testing or assessment of artificial intelligence systems;

c) Other lawful non-financial forms of support as prescribed by law, excluding payment in money or in forms of equivalent monetary value.

3. The Ministry of Science and Technology shall consider and decide on the level of support based on the following criteria:

a) The volume, scale and scope of application of the shared data, models, tools or research results;

b) The degree of completeness and standardization, and the direct usability, of the data, models or research results;

c) The availability and substitutability of the data, models, tools or research results among existing resources;

d) The extent to which they serve the development of artificial intelligence models in Vietnamese and ethnic minority languages, serve public interests or state governance tasks, or promote innovation.

4. The support prescribed in this Article must ensure the following principles:

a) It shall not change the ownership, intellectual property rights and lawful exploitation rights of organizations, enterprises and individuals over the shared resources;

b) It shall not give rise to transactions of sale and purchase of data, models, tools or research results in violation of law;

c) Non-financial support shall not be converted into money or in-kind items of equivalent value.

5. The Minister of Science and Technology shall prescribe the electronic dossier form and guide the organization of implementation of this Article. The guidance must not give rise to administrative procedures, business investment conditions or dossier components beyond the provisions of this Decree.

Article 34. Organization of non-financial support

1. Organizations, enterprises and individuals that wish to receive non-financial support as prescribed in Article 33 of this Decree shall provide information on the resources they wish to share and their demand for support through the single-window artificial intelligence portal.

The provision of information under this Clause is voluntary and not mandatory, does not give rise to administrative procedures, and is not a condition for being considered for support.

2. The information details include:

a) A description of the data, artificial intelligence models, tools, software or research results to be shared, including basic technical information and scope of use;

b) Information on lawful ownership or right of use over the shared resources, and a commitment to comply with the laws on data, protection of personal data, cybersecurity and intellectual property;

c) Demand for use of the forms of support prescribed in Clause 2, Article 33 of this Decree, including the type of resources requested, purposes of use, expected scale and duration of use.

3. The Ministry of Science and Technology shall organize the receipt, consolidation and assessment of information on the basis of the criteria prescribed in Clause 3, Article 33 of this Decree and the capacity of the publicly disclosed support resources to determine the appropriate level and form of support.

The determination of support shall be based on publicly disclosed, transparent and non-discriminatory criteria, ensuring a match between the proposed demand for use and the supply capacity of the infrastructure, data and support services.

4. The result of determining support shall be notified to the organization, enterprise or individual by electronic means, clearly stating the form of support, scope of use, duration of use and conditions of exploitation, if any.

5. Organizations, enterprises and individuals receiving support shall use the support for the right purposes and coordinate in providing information on the results of use where necessary to serve the assessment of the effectiveness of support.

The provision of information under this Clause does not give rise to periodic reporting regimes or regular administrative obligations.

6. The Ministry of Science and Technology shall issue electronic information forms for the provision of information and guide the organization of implementation of this Article.

The use of electronic information forms is supportive in nature, is not mandatory, is not a condition for being considered for support, and does not give rise to administrative procedures, business investment conditions or obligations to provide information beyond the provisions of this Decree.

Article 35. Exploitation and use of data for artificial intelligence development

1. Organizations and individuals may exploit and use data in databases for artificial intelligence development when they satisfy the access conditions, scope of exploitation and purposes of use under the laws on data, protection of personal data and intellectual property, and the access conditions publicly disclosed by the data administering agency or data owner under Clause 2 of this Article.

The exploitation and use of data under this Article shall be carried out independently of the support mechanism prescribed in Articles 33 and 34 of this Decree.

The determination of access conditions and the organization of data exploitation must ensure public disclosure, transparency and non-discrimination among organizations and individuals satisfying the same access conditions as prescribed by law.

2. Agencies and organizations managing and operating databases invested in and built by the State shall publicly disclose on the single-window artificial intelligence portal the following details:

a) The list of data to be shared and exploited;

b) Access conditions, scope of exploitation and purposes of use of the data;

c) Standards, technical regulations and technical connection requirements applicable as prescribed by law;

d) The method of accessing the data and the technical support contact point.

The details prescribed in this Clause shall be interconnected and synchronized to the data information portal. The public disclosure of information under this Clause is intended to ensure the ability to access and exploit data and does not give rise to administrative procedures.

3. The provision of data for artificial intelligence development shall be carried out through one or more of the following forms:

a) Allowing direct access to the data system or access through an application programming interface;

b) Providing datasets for the training, testing or assessment of artificial intelligence systems;

c) Establishing a secure data exploitation environment so that organizations and individuals can process data or train artificial intelligence models on the infrastructure of the data management agency without changing data ownership.

4. Agencies and organizations managing and operating databases shall, when providing data:

a) Establish an access management mechanism appropriate to the type of data and the scope of exploitation;

b) Record and monitor data access and exploitation activities as prescribed by law;

c) Control data exploitation for the right purposes and within the scope of the publicly disclosed access conditions or the established agreement;

d) Apply technical and management measures to ensure data safety and cybersecurity and compliance with the laws on data, protection of personal data and relevant laws.

5. Organizations and individuals exploiting and using data shall:

a) Use data for the right purposes and within the scope and access conditions publicly disclosed or agreed;

b) Not alter or falsify data, or use data in violation of law;

c) Not infringe data ownership, intellectual property rights and the lawful rights and interests of relevant agencies, organizations and individuals as prescribed by law.

6. The collection of fees for exploiting and using data in the national database on artificial intelligence and other artificial intelligence databases managed by state agencies shall be carried out in accordance with the law on data and the law on fees and charges.

Article 36. Ensuring security and safety of infrastructure and data for artificial intelligence development

1. Agencies and organizations managing and operating artificial intelligence infrastructure or databases for artificial intelligence development shall apply technical and management measures appropriate to the purposes of use, scale of deployment and risk level to ensure the security and safety of infrastructure and data as prescribed by law and in line with relevant national standards and technical regulations.

2. Organizations and individuals exploiting and using artificial intelligence infrastructure or data for artificial intelligence development shall comply with the publicly disclosed or agreed access conditions and exploitation conditions, within the scope of use of the infrastructure and data and in accordance with relevant laws.

3. The measures prescribed in Clause 1 of this Article include:

a) Ensuring the confidentiality, integrity and availability of infrastructure and data as prescribed by law;

b) Preventing, detecting and handling cybersecurity risks and incidents and data safety incidents as prescribed by law;

c) Preventing and controlling risks specific to artificial intelligence systems, including the risk of attacks, exploitation of weaknesses of artificial intelligence models, and the risk of re-identification of de-identified personal data in the course of data processing and exploitation.

4. Notification, reporting and coordination in responding to incidents related to personal data and cybersecurity shall be carried out in accordance with the law on protection of personal data, the law on cybersecurity and relevant laws.

Article 37. Mechanisms and measures for promoting mastery of artificial intelligence technologies

1. The Ministry of Science and Technology is the focal agency coordinating the promotion of research, development and mastery of artificial intelligence technologies; it shall consolidate demand, determine priority orientations, track and assess implementation, and ensure coordination among ministries, ministerial-level agencies and provincial-level People's Committees as prescribed by law.

2. Core artificial intelligence technologies prioritized for promotion of mastery include:

a) General-purpose artificial intelligence models, foundation models and large language models for Vietnamese and ethnic minority languages;

b) Technologies for processing Vietnamese knowledge, data and language;

c) High-performance artificial intelligence training technologies and computing capacity for the development and deployment of artificial intelligence systems;

d) Hardware, semiconductor chips and computing technologies for artificial intelligence;

dd) Open-source solutions and foundational technologies for the development of the domestic artificial intelligence ecosystem;

e) Other artificial intelligence technologies on the list promulgated by the Prime Minister for each stage.

3. Ministries, ministerial-level agencies and provincial-level People's Committees, within the scope of their functions, tasks and powers and in accordance with law, shall implement the following measures:

a) Commissioning and assigning science and technology tasks for the technologies prescribed in Clause 2 of this Article;

b) Supporting research, development, testing, refinement and application of artificial intelligence technologies;

c) Organizing connection and cooperation among enterprises, science and technology organizations and higher education institutions;

d) Giving priority to arranging and using artificial intelligence products and solutions in public procurement in accordance with the law on bidding and relevant laws.

4. Organizations and individuals conducting research, development and mastery of the artificial intelligence technologies prescribed in Clause 2 of this Article may access and use support mechanisms and incentive policies in accordance with the laws on science, technology and innovation, investment, taxation and relevant laws.

5. The measures specified in this Article shall be implemented openly and transparently, within the proper authority, and for their intended objectives, without duplicating support from the state budget and in compliance with applicable laws.

Chapter VI

DEVELOPMENT OF THE INNOVATION ECOSYSTEM AND ARTIFICIAL INTELLIGENCE MARKET

Article 38. Development of the artificial intelligence ecosystem and market

1. The Ministry of Science and Technology is the focal agency coordinating the development of the artificial intelligence ecosystem and market; it shall publicly disclose information on infrastructure capacity, data, programs, tasks and demand for artificial intelligence applications in the public sector in order to connect supply and demand and improve the efficiency of resource use as prescribed by law.

2. The development of the artificial intelligence ecosystem and market shall be carried out under the following policies:

a) Developing the supply of artificial intelligence technology by facilitating access to and exploitation of computing infrastructure, data and regulatory sandbox environments; promoting the development of commercial sable artificial intelligence products and services;

b) Stimulating market demand and expanding the application of artificial intelligence by giving priority to the use of artificial intelligence products and services within information technology application tasks and projects, digital transformation, development of digital government and the digital economy, and provision of public services as prescribed by law; encouraging commissioning models, innovative procurement and the application of artificial intelligence in state governance, production, business and social life; recurrent expenditure and public investment funds shall be considered for allocation in accordance with the law on the state budget, the law on bidding and relevant laws;

c) Connecting the market and promoting the commercialization of artificial intelligence products and services by developing technology supply-demand connection platforms; promoting cooperation among enterprises, science and technology organizations, higher education institutions and investors.

3. Technology enterprises, small and medium-sized enterprises, innovative startups, science and technology organizations and higher education institutions participating in developing, providing or applying artificial intelligence products and services shall be given priority in implementing the policies prescribed in Clause 2 of this Article in accordance with relevant laws.

4. The implementation of the policies prescribed in this Article shall be carried out through programs, schemes, science and technology tasks and mechanisms as prescribed by law; and shall not give rise to new administrative procedures or business investment conditions.

Article 39. Artificial intelligence clusters

1. The Ministry of Science and Technology shall decide on recognition of artificial intelligence clusters as prescribed in Clause 1, Article 24 of the Law on Artificial Intelligence when the following criteria are met:

a) There is cooperative participation of at least two of the following three groups of subjects: enterprises; scientific research institutions or higher education institutions; innovation support organizations, intermediary organizations of the science and technology market, state agencies or public non-business units;

b) The objectives, scope and details of activities focus on research, development, testing, application or commercialization of artificial intelligence technologies;

c) There is a cooperation agreement, operating regulation or linkage scheme agreed in writing by the members, clearly specifying the coordination mechanism, benefit sharing and responsibilities of participating parties;

d) There is a mechanism for coordinating the activities of the cluster, including a coordinating focal point and the method of coordination among members.

2. The organization representing the cluster shall submit 01 set of electronic dossiers requesting recognition of the artificial intelligence cluster through the National Public Service Portal for receipt and processing as prescribed by law. The dossier shall be interconnected and synchronized to the single-window artificial intelligence portal to perform other state governance tasks as prescribed in this Decree. The dossier comprises:

a) A document requesting recognition of the artificial intelligence cluster;

b) The list of members participating in the cluster;

c) The cooperation agreement, operating regulation or linkage scheme;

d) Documents proving satisfaction of the criteria prescribed in Clause 1 of this Article.

3. Within 15 working days from the date of receiving a complete and valid dossier, the Ministry of Science and Technology shall receive, appraise and decide on recognition of the artificial intelligence cluster. The list of recognized artificial intelligence clusters shall be publicly disclosed on the single-window artificial intelligence portal. Competent agencies and persons receiving and settling dossiers must not require additional dossier components, conditions or procedures beyond the provisions of this Article.

4. Artificial intelligence clusters operate on the basis of voluntariness, autonomy and cooperation; participation in a cluster does not change the legal status or independent rights and obligations of participating members and does not form a state governance organization within the cluster.

5. After being recognized, artificial intelligence clusters shall be given priority in accessing support policies as prescribed by law, including:

a) Priority access to and exploitation of State-invested artificial intelligence infrastructure, including computing capacity, artificial intelligence platforms and regulatory sandbox environments, under the service provision mechanism and relevant laws;

b) Priority access to and exploitation of shared data and datasets for the training, testing and assessment of artificial intelligence systems in accordance with the law on data and relevant laws;

c) Priority consideration for participation in science and technology programs and tasks, innovation programs, artificial intelligence development programs and digital transformation programs using the state budget in accordance with the law on the state budget, the law on science and technology and relevant laws;

d) Consideration for support through Artificial Intelligence Development Support Vouchers and other forms of support as prescribed in Article 40 of this Decree and relevant laws.

6. The Ministry of Science and Technology shall act as the focal point supporting artificial intelligence clusters in connecting with programs, infrastructure, databases and policies supporting artificial intelligence development; such support shall not interfere with the internal operations of the cluster.

Article 40. Mechanism for supporting artificial intelligence development through Support Vouchers

1. An Artificial Intelligence Development Support Voucher is an electronic voucher recording a funding support limit from the National Artificial Intelligence Development Fund and other lawful sources to pay the costs of using computing infrastructure, shared data, large language models for Vietnamese and ethnic minority languages, training and testing platforms and technical consulting services for research, development and deployment of artificial intelligence applications.

2. Subjects eligible for consideration for a Support Voucher include:

a) Innovative startups, and small and medium-sized enterprises;

b) Science and technology organizations and research groups with feasible innovation projects;

c) Organizations and individuals participating in the regulatory sandbox mechanism under this Decree;

d) Other organizations and individuals with artificial intelligence research, development or application projects consistent with programs, schemes or support mechanisms approved by competent agencies as prescribed by law.

3. The issuance, management and use of Support Vouchers shall comply with the following principles:

a) Public disclosure, transparency, objectivity and non-discrimination;

b) Consistency with policy objectives and the capacity to allocate support resources;

c) A Support Voucher has no monetary conversion value, is non-transferable and may only be used within the scope, purposes, limit, duration and conditions stated on the Voucher;

d) No duplicate support for the same expenditure item already supported by the state budget or other lawful financial sources, unless otherwise provided by law.

4. Organizations and individuals prescribed in Clause 2 of this Article shall send a request for a Support Voucher through the National Public Service Portal for receipt and processing as prescribed by law. The dossier shall be interconnected and synchronized to the single-window artificial intelligence portal to perform other state governance tasks as prescribed in this Decree. The request dossier includes:

a) A document requesting the issuance of a Support Voucher;

b) An explanation of the demand for use of services and the expected objectives to be achieved;

c) Documents proving that the applicant falls within the subjects prescribed in Clause 2 of this Article.

Within 05 working days from the date of receiving a complete and valid dossier, the Ministry of Science and Technology shall consider and decide on issuing the Support Voucher on the basis of public criteria and the capacity to allocate resources; it shall not require dossier components beyond those prescribed in this Clause.

5. Support Vouchers are used to pay the costs of services directly for artificial intelligence activities, including:

a) Computing and storage infrastructure services and technical platforms;

b) Data and artificial intelligence model services within the scope permitted by law;

c) Services for data processing, training, testing and assessment of artificial intelligence systems;

d) Technical consulting services for safety, security and risk governance.

6. Organizations and enterprises providing the services prescribed in Clause 5 of this Article may participate in providing services paid for by Support Vouchers when they:

a) Operate lawfully in accordance with law;

b) Publicly disclose information on services, conditions of provision and service prices;

c) Comply with the laws on data, protection of personal data, cybersecurity, intellectual property and relevant laws.

The public disclosure of information prescribed at Point b of this Clause is not a business condition and does not give rise to administrative procedures.

7. Organizations and individuals issued with Support Vouchers may select the organization or enterprise providing the services prescribed in Clause 6 of this Article to use services suited to their needs. Such selection is a civil transaction between the parties; the payment of the value of the Support Voucher does not give rise to a contractor selection relationship of a state agency under the law on bidding.

8. Payment of the value of Support Vouchers shall be made directly to the service-providing organization or enterprise on the basis of:

a) The issued Support Voucher;

b) The service provision contract or agreement;

c) Invoices and accounting vouchers proving completion of service provision as prescribed by law.

Payment shall not be made through the subject issued with the Support Voucher.

9. The Ministry of Science and Technology shall:

a) Organize the issuance, management and use of Support Vouchers;

b) Publicly disclose the criteria for considering the issuance of Support Vouchers, the list of services and information on service-providing organizations;

c) Track and inspect the use of Support Vouchers as prescribed by law;

d) Publicly disclose information on the implementation of the Support Voucher mechanism at an appropriate level, ensuring no infringement of state secrets, business secrets, personal data and the lawful rights and interests of organizations and individuals.

10. Organizations and individuals that commit acts of fraud, collusion, inflating service prices, untruthful declaration, falsely confirming the volume or results of service provision, using unlawful invoices or vouchers, using Support Vouchers for improper purposes, transferring Support Vouchers, or abusing the mechanism to obtain duplicate support shall refund the entire value of the support received and shall be handled in accordance with law.

Chapter VII

RESPONSIBILITIES OF AUTHORITIES AND ORGANIZATIONS

Article 41. Responsibilities of the Ministry of Science and Technology

The Ministry of Science and Technology is the focal agency, is accountable to the Government for performing state governance of artificial intelligence nationwide, and has the following tasks and powers:

1. To organize the construction, management and operation of the single-window artificial intelligence portal and the national database on artificial intelligence systems.

2. To assume the prime responsibility for, and coordinate with ministries and ministerial-level agencies in, consolidating, developing and submitting to the Prime Minister for promulgation, and updating, the following lists:

a) The List of high-risk artificial intelligence systems;

b) The List of important artificial intelligence applications in essential fields.

3. To organize the receipt and consolidation of information for state governance of artificial intelligence through the single-window artificial intelligence portal, including:

a) Risk classification notifications;

b) Serious incident notifications;

c) Feedback and recommendations of organizations and individuals;

d) Other details as prescribed in this Decree.

The receipt and consolidation of information prescribed in this Clause do not give rise to reporting obligations beyond those prescribed by law.

4. To issue, organize guidance on, and inspect the implementation of the national framework of artificial intelligence ethics in accordance with law.

5. To perform state governance of conformity assessment activities for artificial intelligence systems in accordance with law.

6. To assume the prime responsibility for managing and coordinating the connection and exploitation of State-invested shared artificial intelligence infrastructure for research, development and innovation; to coordinate with the Ministry of Public Security in giving unified opinions on investment projects in artificial intelligence infrastructure of ministries, sectors and localities in order to ensure that duplicated and wasteful investment is avoided.

7. To organize the implementation of the regulatory sandbox mechanism for artificial intelligence systems within its competence under this Decree and relevant laws.

8. To organize the implementation of mechanisms supporting artificial intelligence development under this Decree and relevant laws.

9. To assume the prime responsibility for developing, and submitting to the Government for promulgation, regulations on the financial mechanism, organization and operation of the National Artificial Intelligence Development Fund in accordance with law.

10. To organize inspection, settlement of complaints and denunciations, and handling of violations in accordance with law.

11. To issue general standards, regulations and guidelines for artificial intelligence systems.

12. To provide guidance on implementing the regime of information and periodic reporting on the situation of development, application and management of artificial intelligence.

Article 42. Responsibilities of the Ministry of Public Security

Within the scope of its functions, duties, and powers, the Ministry of Public Security shall be responsible for coordinating in the state governance of artificial intelligence with respect to matters relating to safeguarding national security, public order and social safety, cybersecurity, and data protection, including the following duties and powers:

1. To assume the prime responsibility for, and coordinate with ministries, ministerial-level agencies, provincial-level People's Committees and relevant agencies and organizations in, detecting, preventing, investigating and handling in accordance with law acts of abusing artificial intelligence systems to infringe national security, social order and safety, and the lawful rights and interests of organizations and individuals.

2. To assume the prime responsibility for organizing the construction, operation and provision of artificial intelligence infrastructure for state governance, concentrated at national data centers as prescribed in Clause 6, Article 28 of this Decree; to coordinate with the Ministry of Science and Technology in giving unified opinions on investment and construction projects of artificial intelligence infrastructure for state governance of ministries, ministerial-level agencies and provincial-level People's Committees in order to ensure cybersecurity requirements and avoid duplicated and wasteful investment.

3. To be the managing agency and be responsible for building, managing and operating the national database on artificial intelligence at the National Data Centre in accordance with this Decree and the law on data.

4. To assume the prime responsibility for organizing the development of cybersecurity standards and issuing technical regulations applicable to infrastructure deploying important artificial intelligence applications in essential fields as assigned by the Government.

5. To assume the prime responsibility for verifying and handling in accordance with law serious artificial intelligence incidents showing signs of affecting national security, social order and safety; to coordinate with the Ministry of Science and Technology and relevant agencies in determining causes and providing guidance on remedial measures within its competence.

6. To review, assess and send to the Ministry of Science and Technology proposals on artificial intelligence systems in the field of security and order that should be considered for inclusion in the List of high-risk artificial intelligence systems.

7. To coordinate with the Ministry of Science and Technology in implementing details on the regulatory sandbox mechanism and other details prescribed in this Decree.

8. To assume the prime responsibility for, and coordinate with the Ministry of Finance in, issuing economic and technical norms for activities of collecting, creating, ensuring the quality of, standardizing, labelling and annotating data for artificial intelligence, as a basis for managing and using the state budget as prescribed in Clause 6, Article 32 of this Decree.

9. To assume the prime responsibility for, and coordinate with ministries and ministerial-level agencies in, consolidating, developing and submitting to the Prime Minister for promulgation, amendment and supplementation of the lists of datasets for the development of artificial intelligence in essential fields.

Article 43. Responsibilities of ministries, ministerial-level agencies and other agencies under the Government

Ministries, ministerial-level agencies and other agencies under the Government shall, in the course of performing their functions, tasks and powers and based on state governance needs, have the following responsibilities:

1. To review, organize the development and promulgation within their competence of, or submit to competent agencies or persons for promulgation of, legal documents and national standards and technical regulations detailing requirements on safety, risk management and deployment conditions for the application of artificial intelligence systems in the sectors and fields under their management in accordance with the Law on Artificial Intelligence.

2. To organize inspection within their competence, examination and supervision of the deployment and application of artificial intelligence systems in the sectors and fields under their management in accordance with the Law on Artificial Intelligence, this Decree and relevant laws.

3. To receive, appraise and issue approval documents for participation in the regulatory sandbox mechanism for level 1 and level 2 artificial intelligence systems within their competence; to coordinate with the Ministry of Science and Technology in appraising the sector-specific suitability of dossiers requesting participation in the regulatory sandbox mechanism for artificial intelligence systems in the sectors and fields under their management as prescribed in this Decree.

4. To organize the collection, creation, quality assurance, labelling, annotation, standardization, management and exploitation of databases for artificial intelligence within their management scope; to assume the prime responsibility for building and operating artificial intelligence infrastructure for state governance within their competence; to ensure connection and sharing with the national database on artificial intelligence in accordance with standards, technical regulations and cybersecurity requirements as prescribed by law.

5. To integrate artificial intelligence development objectives and tasks into sectoral and field development strategies and plans; to give priority to allocating resources and using artificial intelligence products and services in state governance activities and the provision of public services in accordance with the Law on Artificial Intelligence.

Article 44. Responsibilities of provincial-level People’s Committees

Provincial-level People's Committees shall perform state governance of artificial intelligence in their localities and have the following responsibilities:

1. To promulgate and organize the implementation of programs and plans for the development and application of artificial intelligence appropriate to local conditions in accordance with law.

2. To organize the collection, creation, quality assurance, labelling, annotation, standardization, management and exploitation of local databases for artificial intelligence; to assume the prime responsibility for building and operating artificial intelligence infrastructure for state governance in the locality; to ensure connection and sharing with the national database on artificial intelligence in accordance with standards, technical regulations and cybersecurity requirements as prescribed by law.

3. To integrate artificial intelligence development objectives and tasks into local development strategies and plans; to give priority to allocating resources and using artificial intelligence products and services in state governance activities and the provision of public services; to direct agencies and units under their management to conduct impact assessment and meet ethical requirements when using artificial intelligence systems in accordance with law.

4. To coordinate with the Ministry of Science and Technology in organizing the implementation of regulations on risk classification, incident reporting, conformity assessment and the regulatory sandbox mechanism for artificial intelligence systems in their localities.

5. To receive, appraise and issue approval documents for participation in the regulatory sandbox mechanism for level 1 and level 2 artificial intelligence systems within their competence; to coordinate with the Ministry of Science and Technology in organizing the implementation of regulations on risk classification, incident reporting, conformity assessment and the regulatory sandbox mechanism for artificial intelligence systems in their localities.

6. To organize examination and handling of violations and implement the regime of information and periodic reporting on the situation of development, application and risk management of artificial intelligence systems in their localities in accordance with law.

Chapter VIII

IMPLEMENTATION PROVISIONS

Article 45. Effect

This Decree takes effect from May 01, 2026.

Article 46. Transitional provisions

1. In case the single-window artificial intelligence portal has not yet been officially publicly disclosed by the Ministry of Science and Technology as being in operation, notification and reporting procedures under this Decree shall be carried out through the electronic information system, online public service or receiving method publicly disclosed by the Ministry of Science and Technology, ensuring uniformity nationwide.

Notification and reporting by the methods prescribed in this Clause have legal validity equivalent to performance through the single-window artificial intelligence portal.

2. During the period prescribed in Clause 1 of this Article, public disclosure of information under this Decree shall be carried out on the official portal or website of the agency, organization or unit responsible for public disclosure under this Decree, unless otherwise provided by law.

3. Dossiers and information received before the time prescribed in Clause 3 of this Article shall continue to be handled in accordance with the provisions in force at the time of receipt.

* All Appendices are not translated herein.