The Ministry of Information and Communication ('MIC') requested, on 20 April 2023, public comments on the draft National Standard on Artificial Intelligence and Big Data. In particular, the draft Standard on artificial intelligence ('AI') is separated into two documents on Artificial Intelligence Lifecycle Process and Requirements, namely one on quality and another on robustness. Nonetheless, the draft Standard on AI establishes the concept of an AI module lifecycle, consisting of the conception, development, deployment, operation, and decommissioning of AI modules.

Notably, the draft Standard on AI details the aim to establish quality assurance and transparency of AI modules, proposing quality requirements for the safety, privacy, and ethics of AI. In addition, the draft Standard on AI outlines that the first step in evaluating AI is whether the AI module in question presents a high or low risk, and the aspects that should be considered when undertaking a risk assessment of AI modules. Although the draft Standard on AI does not establish a rigorous risk assessment process or an ethical design framework, the draft Standard on AI does identify expectations for AI systems. Specifically, the draft Standard on AI identifies confidentiality, integrity, and availability as key considerations, alongside privacy and the related data subject rights throughout the AI module lifecycle. Likewise, the draft Standard on AI notes that unintentional bias of AI modules should be considered in the development of AI modules.

In addition, the draft Standard on AI considers sustainability a fundamental pillar of AI quality, being formed of resistance to adversarial disturbances, such as changes in data distribution and bias. Regarding sustainability, the draft Standard on AI stipulates that risk assessments should be carried out to identify adversarial disturbance and bias risks. Equally, the draft Standard on AI details general risk management processes to be applied when determining risks to sustainability, including identification of the likelihood of risks, risk scenarios, and acceptable ranges of such risks.

Public comments may be submitted here until 20 June 2023.

You can read the documents forming the draft Standard on AI here and here, both only available in Vietnamese.