📋 Tender Overview
The Minister of Science and Technology has just issued Decision No. 2847/QD-BKHCN on a set of criteria for evaluating Artificial Intelligence (AI) Platforms supporting public affairs on a national scale, creating a basis for selecting, testing and deploying AI platforms to serve officials, civil servants and public employees in state management activities. Download Illustration ImageDecision No. 2847/QD-BKHCN was issued to specify the requirements for developing and applying artificial intelligence in the public sector, aiming to build a shared AI infrastructure on a national scale, supporting the improvement of efficiency in public service performance. Accordingly, the set of evaluation criteria is built as a unified basis to evaluate the level of response of AI platforms before being tested or officially deployed. AI supports, does not replace the responsibilities of officials and civil servants. According to the Ministry of Science and Technology, the Artificial Intelligence Platform supporting public affairs on a national scale is defined as a shared artificial intelligence infrastructure, serving cadres, civil servants and public employees in the process of performing tasks in the fields of state management. An important principle in the set of criteria is to ensure control of public affairs. Artificial intelligence only plays a supporting role in consulting, synthesizing, and analyzing; Competent officials, civil servants and public employees are still the ones who review, decide, approve and bear the highest legal responsibility in state management activities. The results provided by the AI platform are for reference only and do not replace the administrative decisions of competent persons. The set of criteria also requires the platform to only use data sources that are legal documents that have been publicly announced; All outputs must have a legal reference to the Article, Clause, and Point level. The scope of support includes central legal documents and documents issued by the Peoples Council and Provincial Peoples Committee to serve local civil servants. At the same time, officials, civil servants and public employees do not use the platform to process records, data and information that are state secrets. AI activities related to state secrets, national defense, and security must comply with regulations on protecting state secrets, network security, and ciphers. Evaluation on a 100-point scale The set of criteria includes two main groups with a maximum total score of 100 points. Of which, Group A - general criteria has a total score of 35, focusing on evaluating requirements for legal compliance, source references, data boundaries, compliance with AI laws, infrastructure capacity, AI security, personal data protection and AI ethics. Group B - public service support criteria according to the field of state management has a total score of 65, evaluating the platforms specialized support capacity such as synthesizing, analyzing, explaining, and providing professional guidance based on documents. published version of each field. The set of applied criteria evaluates the ability to support public affairs in the fields of state management of 14 ministries, 3 ministerial-level agencies and 34 localities. To be recognized as eligible for official implementation, the platform must achieve 70 points or more, and the total score of Group A must reach at least 50% and there are no criteria for being paralyzed. Platforms that score from 50 to less than 70 points are eligible to participate in testing and must upgrade for re-evaluation after 2 months. If you score less than 50 points, you are not eligible to participate in the test. One of the outstanding requirements of the set of criteria is to ensure autonomy in core AI technology. The platform will be judged ineligible to participate in testing if it does not use a domestic Vietnamese large language model owned by a Vietnamese enterprise, along with training, inference and service provision infrastructure in Vietnam. The set of criteria requires the platform to meet the contents related to mastering the Vietnamese large language model, including model architecture and weights designed and owned by Vietnamese organizations and enterprises; Vietnamese training data is collected and processed legally; suitable Vietnamese encoder; training and inference infrastructure located in Vietnam and a core AI research and development team belonging to Vietnamese organizations and businesses. In addition to technological capacity, the platform must also have a process for reporting serious incidents and meet wide-area service requirements, with the capacity to serve at least 5,000 concurrent users in the testing phase and at least 50,000 users when officially deployed. In the group of public service support criteria, the AI platform is evaluated through its ability to Respond, synthesize, and analyze legal information in each field; Requires the results to have references to articles, clauses, and points of legal documents and have a warning when the question exceeds the range of data allowed to support. The criteria also require the platform not to fabricate legal bases, not process specific records and data within a limited scope, and must be able to detect and warn of cases of overlap, contradiction, or inconsistency between legal documents in each field.Thu GiangReferences moreIssue a Decree detailing a number of articles and measures to implement the Law on Artificial IntelligenceSee moreNoteworthy points of the Law on Artificial Intelligence about to take effectSee moreVietnam officially has a Law on Artificial Intelligence (AI) Refer to the Guidance on civil servants and public employees using virtual assistants to serve their work
