Sovereign AI needs more than models

A recent Bangkok Post article raises a familiar concern: leadership teams are worried that their staff are becoming too dependent on AI, accepting its outputs without enough scrutiny. This anxiety, while understandable, misses the larger strategic questions that ASEAN governments and business leaders should be asking about artificial intelligence. The real test is not whether junior analysts can be weaned off generative AI, but whether the region can build genuine sovereignty over the core infrastructure that powers these new technologies.

The Infrastructure Layer

Too much of the conversation around AI in Southeast Asia is focused on high-level applications and model launches. These are visible and exciting, but they are not the foundation of lasting economic advantage. As the book "ASEAN Rising" notes, the more substantive issue is one of control over the fundamental building blocks of the digital economy. The long-term viability of sovereign AI will be determined by who owns and operates the compute, data, and identity layers.

Today, the vast majority of AI infrastructure-from the graphics processing units (GPUs) in data centers to the cloud services that host and train models-is controlled by a small number of foreign firms. This creates dependencies that could limit the region's strategic autonomy. While it is impractical for every ASEAN nation to build its own foundational models or fabricate its own chips, a regional strategy is needed to ensure that access to this critical infrastructure is secure and that the benefits are broadly shared.

Talent and Trust

The Bangkok Post article touches upon a valid point regarding human capital. An uncritical acceptance of AI-generated content can indeed stifle the development of analytical skills. This is not just a problem for individual firms but a systemic one for the entire economy. A workforce that cannot independently reason, analyze, and challenge assumptions is a workforce that cannot innovate. Building a robust talent pipeline requires more than just training people to use AI tools; it requires cultivating a culture of critical thinking and deep domain expertise.

This intersects with the issue of trust. For AI to be integrated safely into the economy, from financial services to public administration, there must be a high degree of trust in the systems themselves and the institutions that govern them. This means developing clear standards for data governance, algorithmic transparency, and liability. When leaders express concern about their teams' over-reliance on AI, they are voicing a proxy for this larger societal need for a trustworthy and accountable AI ecosystem.

From Application to Ownership

The current excitement over AI applications is understandable, but it is time to shift the focus from consumption to ownership. The book ASEAN Rising argues that "Sovereign AI in ASEAN will be tested not by model launches but by who controls the compute, the data layer and the digital identity rails". This means prioritizing investment in data centers, fostering regional cloud providers, and developing secure, interoperable digital identity systems that are managed within the region. It is a less glamorous but far more important task than simply rolling out the latest chatbot.

What to watch: Observe whether regional governments move beyond high-level AI strategy documents and begin to implement concrete policies and co-investment programs aimed at the infrastructure layer. Pay attention to investments in data centers, regional cloud services, and cross-border data governance and digital identity frameworks. The success of ASEAN's digital future will be measured by its ownership of the core digital rails, not just the traffic that runs on them.