AI being ever more embedded

Vietnam’s AI adoption rate reached 26.5 per cent in the first quarter of 2026, according to Microsoft’s Global AI Diffusion Report, up from 23.5 per cent in 2025 and representing a 3-percentage-point increase - the highest in the region. While there remains a significant gap with Singapore, which ranks second globally with an adoption rate of 63.4 per cent, Vietnam still leads many other Southeast Asian countries, including Malaysia (21.8 per cent), the Philippines (20.1 per cent), and Thailand (12.4 per cent).

Economic impact

At the recent CRIF Forum Vietnam 2026, Mr. Do Tien Thinh, Deputy Director of the National Innovation Center (NIC) under the Ministry of Finance, emphasized that Vietnam is among the countries with the fastest rates of AI adoption and innovation, with growth in some areas outpacing the economy and exceeding the global average. “Among government officials and civil servants, more than 90 per cent are currently using AI on a daily basis,” he told the Forum. “We divide businesses into two groups - micro, small, and medium-sized enterprises, and large enterprises. Within the startup community, almost all have adopted AI.”

Vietnam is therefore at a stage where the adoption and use of AI to drive economic development is accelerating rapidly. Some studies by the NIC in collaboration with Google and Boston Consulting Group (BCG) show that AI is already making a significant contribution to Vietnam’s economy.

The report estimates that nearly 30 per cent of Vietnam’s GDP growth is linked to digital technology. Within digital technology, AI is one of the core technologies, alongside blockchain, the internet, and others. Various forecasts estimate AI’s current contribution to Vietnam’s economy at around $40-$50 billion, spanning agriculture, industry, and services.

The law classifies AI systems as high, medium, or low-risk, while also establishing mechanisms to promote infrastructure, data, human resources, and the innovation ecosystem, support startups and small businesses, and establish a national AI development fund.

However, alongside the opportunities, AI is also beginning to create negative impacts on the economy and workforce. In practice, AI is already exhibiting certain adverse effects on the labor market. 

Vietnam trains a large number of information technology workers, but among the corporations working with the NIC, most have hired relatively few additional IT employees over the past one or two years. “The reason is that they have started using AI,” Mr. Thinh said, noting that he has encountered several corporations with thousands of highly-skilled IT employees that, after effectively adopting AI, now need only a small number of employees to check and validate AI-generated results.

Overall, Vietnam has a major opportunity to adopt AI and already has a relatively open policy framework. At the same time, however, there are risks related to technology and trust, AI’s impact on the economy and trade, and changes in the structure of employment.

This is also one of the reasons a Law on AI is needed - to put people at the center and establish regulations to protect workers. The Ministry of Science and Technology is also developing regulations and guidelines on AI ethics to operate alongside the legal framework. In addition to AI ethics, another important issue in AI development - both within businesses, in credit processes, and in the development of new products - is the use of sandboxes, or controlled testing environments.

Vietnam has officially incorporated the sandbox mechanism into its legal framework, under which organizations may be permitted to conduct pilot programs for one to three years, with the possibility of a further three-year extension. “The mechanism can be applied flexibly in terms of location, participating entities, and the technologies being tested,” Mr. Thinh emphasized, adding that Vietnam’s AI development path will combine sandbox mechanisms and AI ethics while continuing to refine policies and mechanisms for governing and monitoring AI adoption.

We divide businesses into two groups - micro, small, and medium-sized enterprises, and large enterprises. Within the startup community, almost all have adopted AI. 

Mr. Do Tien Thinh, Deputy Director of the National Innovation Center (NIC), Ministry of Finance

Controlling input data 

Meanwhile, Mr. Sachin B Singh, Head of Enterprise Solutions, Asia-Pacific, at Dow Jones, said the biggest challenge for AI is not simply the model itself but first and foremost the quality and verifiability of the data.

A recent study, he continued, found that more than 40 per cent of the information or answers users receive from AI come from sources such as Reddit and Wikipedia. At the individual level, using such information to decide which restaurant to eat at tonight may not be a major concern. But at the organizational level, using the same type of data to make important business decisions is an entirely different matter.

Organizations often view this as a model problem, but Mr. Singh emphasized that no matter how sophisticated the technology is, it cannot resolve the underlying data problem. Organizations therefore need to be able to trace the source of an answer. For example, if AI says electric vehicle production in Vietnam increased 10 per cent, users should be able to trace that information back to its source and determine how reliable it is.

This is also why many leading organizations, when adopting AI, do not allow models unrestricted access to the entire internet. Rather, they have AI operate using verified internal data sources, ranging from policy documents and research to databases that the organization considers reliable.

According to Mr. Singh, the journey toward building trust in AI must therefore start with the data source, followed by the model, and ultimately the people involved. This is especially important as organizations move from generative AI into the era of agentic AI, in which systems begin making decisions autonomously and the level of risk is significantly higher. AI needs to be continuously learned and trained, not only by checking whether an answer is right or wrong, but also by assessing the degree of error, identifying the cause, and determining the level of confidence. These insights can then be used to recalibrate and retrain the system so the quality of its decisions continues to improve.

Vietnam, however, is also facing a reality in which data is already abundant across the economy, but finding reliable signals on which to base accurate decisions remains a challenge. In particular, determining which data can be trusted, verified, and effectively leveraged remains far from straightforward.

As AI becomes increasingly embedded in economic activity, the challenge is no longer simply to develop a smarter model. More importantly, it is to build a reliable data system, verification mechanisms, clear human accountability, and a sufficiently flexible governance framework that allows the technology to develop without undermining trust.