Sovereign AI and human oversight key to secure digital government

What principles are key to building inclusive, trusted digital public services?

Manon Lafitte. Photo: Ha Vy

The first principle I consider particularly important is genuinely involving citizens in public decision-making. This is not always easy or straightforward, particularly when a country such as Vietnam is pursuing digital transformation at a very rapid pace. Nevertheless, involving both citizens and public officials in public decision-making and the digital transformation process will be key to future success.

Public officials need to be involved in decision-making, and their needs must be properly communicated and considered. They should have the opportunity to express their needs and identify the difficulties and frustrations they encounter in their work. By ensuring that these needs and concerns are heard, public officials can be genuinely integrated into digital projects and placed at the heart of the digital transformation process.

The second principle is accountability to citizens. This means ensuring transparency not only in the digital strategy itself, but also in the public-sector algorithms used in the digital transformation process.

How has open data supported digital transformation in France, and what lessons can Vietnam draw from this experience?

The opening of public data has been a real success in France. Thanks to this, we have been able to develop a highly effective AI strategy in France and across Europe. Making public data open and facilitating the flow of data between government agencies are therefore at the heart of digital transformation.

This is important because opening public data also unlocks economic value and creates opportunities to develop new services. For example, weather data, as well as mobility and transport data, can be used to create new services. Businesses can access new datasets and use them to develop new services and solutions.

Open data therefore presents a significant economic opportunity. It also plays an important role in enabling the development of effective AI, which depends on high-quality data and the use of open standards. These elements are essential to making data available and usable for the development of new services and AI applications.

What should governments and businesses do to ensure that digital transformation remains inclusive?

The widespread use of freely available commercial AI platforms by public officials for official duties poses significant risks to information security and data confidentiality.

To address this situation, the first step is to conduct a review of actual AI usage across government agencies. This should involve identifying who is using AI, the purposes for which it is being used, and whether the tools in question are sovereign AI systems or unregulated commercial AI tools, thereby establishing a management map of AI usage.

On this basis, research should be conducted into and proposals developed for sovereign AI models designed in-house by the relevant agencies to replace the indiscriminate use of external platforms.

In addition, AI systems can experience hallucinations, generating inaccurate answers or responses containing prejudice and bias. Without stringent oversight mechanisms in place from the outset, these models could lead to serious operational incidents.

France has a dedicated body that serves as an independent laboratory for auditing AI solutions and algorithms before they are deployed.

This body assesses whether AI solutions meet the criteria for transparency, explainability, and traceability, thereby ensuring that government agencies retain full control over their information systems.

The fundamental principle when deploying AI for public service is to maintain human review, oversight, and approval.

Humans must directly verify the sources of data used by AI to prevent hallucinations and information inaccuracies. Depending on the risk level of each AI model, managers should apply corresponding stringent standards and regulatory requirements. The more explainable and transparent an AI model is, the more straightforward it will be to meet the applicable legal requirements for its practical deployment.

The more significant technological advances become, the greater the risk of excluding people who do not have sufficient information about new technologies, how they work, and how they contribute to public decision-making.

One way to make digital transformation more inclusive is to establish common standards and promote a shared ethical framework among citizens. Public officials should also be provided with clear guidelines and charters that can be translated into language accessible to everyone.

In addition, public websites should be made accessible to people with disabilities. This should be supported by legislation on the accessibility of online public services, helping ensure that digital public services are accessible to those who need to use them.

Hanoi to launch portal to ease hurdles for tech businesses Hanoi is plans an online portal and inter-agency task force to streamline case handling and address bottlenecks facing science and technology enterprises, innovative startups and digital technology businesses.

Digital government push focuses on data connectivity and AI Vietnam is accelerating its shift towards digital government by strengthening data connectivity, improving public service delivery, and exploring emerging technologies such as AI and UAVs to make governance more efficient, responsive, and citizen-centric.