Building AI readiness for health in Southeast Asia

The concept of AI readiness–a framework for assessing how prepared governments are to adopt artificial intelligence (AI) ethically and responsibly in public services—has gained global recognition, with endorsement from intergovernmental organizations and forums.¹ Multiple benchmarking tools now exist,2, 3, 4 converging on three pillars: strategic vision and governance capability, maturity of the national innovation ecosystem, and availability and quality of data and infrastructure.

According to the Oxford Insights 2024 Government AI Readiness Index, which draws on hundreds of indicators across these pillars, Singapore and Malaysia lead in AI readiness in Southeast Asia, followed by Thailand, Indonesia, Viet Nam, Brunei and the Philippines as middle performers with strengths in one or two pillars, while Cambodia and the Lao People's Democratic Republic have yet to catch up across all three pillars.⁵ A common strength across the region is the inclusion of AI in national digital strategies and small-scale pilots, as well as recognition of the urgency to introduce governance frameworks for responsible AI use. However, local innovation ecosystems remain underdeveloped–talent, funding, homegrown AI technologies, and cross-sectoral collaboration to sustain and scale AI in public services are limited. Data and infrastructure gaps persist, with availability, quality, and interoperability remaining uneven.⁶

Although these assessments are cross-sectoral, they have critical implications for health systems, where both the urgency and risk of AI adoption are high. While high-income countries can pursue self-sufficient approaches, building AI readiness in the health systems of lower-and-middle-income countries (LMICs) requires regional cooperation.⁷ Drawing on insights from policymakers, regulators, academia, civil society, private sector actors, and funding agencies across eight countries at a recent multi-stakeholder roundtable, this commentary outlines three priorities for advancing health AI readiness in the region.

Adopt AI calibrated to the local readiness and needs today for immediate benefits

Rather than waiting for ideal conditions, countries can adopt AI solutions that match current readiness levels and strengthen health systems incrementally. Ministries of Health and national regulators can set phased priorities tied to existing governance rules; technical expertise within the region can supply implementation guidance and benchmarking tools; universities, civil society, and industry supply adaptable solutions and evaluation capacity.

Countries with high AI readiness can trial higher-stakes applications backed by secure cloud and formal clinical governance–such as large language model-based clinical decision support tools and AI assistants embedded in healthcare systems⁸ In lower-readiness settings, the focus should be on AI solutions with proven clinical effectiveness, minimal connectivity requirements, and low integration complexity. For example, AI-assisted chest X-ray screening for TB has improved program performance in the Philippines and is being scaled with standardized implementation frameworks in Viet Nam, demonstrating practical gains when paired with confirmatory testing and calibrated operating thresholds.^9,10 Likewise, the AI-based clinical decision support chatbot, localized for Timor-Leste and embedded in a popular messenger app, shows how generative AI can be tailored to national guidelines and languages to support frontline workers without heavy system integration and upfront cost.¹¹ Finally, machine-learning guidance for antimicrobial stewardship in Cambodia illustrates that valuable decision support can be built from routinely collected hospital data even in resource-limited settings.¹²

Organize regional public goods to elevate AI readiness

Barriers to health AI adoption in Southeast Asia can be significantly lowered by taking a regional approach and through the development of regional public goods, guided by the collective commitment from health system leadership:

First, open-source foundation AI models that are regionally representative. Current mainstream models underrepresent Southeast Asian populations, risking bias and poor performance. Governments should support the curation of regionally representative clinical and public health datasets with clear access terms for public interest. Recognizing the sensitivity of cross-border or multi-institutional data sharing, a consortium anchored in academia, co-governed and supported by intergovernmental agencies such as WHO, can be a suitable host.

Second, regionally shared computing to lower the startup cost of implementation for less AI-ready countries. This requires a different, and significantly more rigorous, set of data governance and safeguards compared to data kept within a single, local entity. Even though there has not been such a regional initiative thus far, applications of federated learning in health AI can offer highly relevant lessons on strengthening data governance, especially regarding privacy, security, and trust.

Third, harmonised principles for health AI regulation. While national laws remain sovereign, convergence on core principles–safety, transparency, accountability-can provide a clear pathway for proven AI solutions to be accessible across borders. While waiting for national innovation ecosystems and homegrown technologies to mature, this approach allows health systems in LMICs to adopt or subscribe to proven, affordable AI technologies from anywhere in the region.

Step up a regional platform to coordinate efforts

Countries in Southeast Asia would benefit from a dedicated platform to coordinate strategies for building health AI readiness and organizing regional public goods. For one, the platform is crucial for governments to negotiate and reach consensus on data sharing and the development of regionally representative AI models. It could also facilitate the alignment of AI solutions with specific health system needs, support local adaptation and adoption of validated tools, foster dialogue on governance principles and regulatory frameworks, and provide guidance on implementation and best practices. Neutral partners such as the WHO Regional Office for the Western Pacific at this forum, could serve as the secretariat for such a platform.

Conclusion

There will never be a perfect time and condition to unlock the benefits of AI, and countries should not wait for ideal readiness. With the right approach, countries can implement fit-for-purpose solutions today, while working collectively to organize regional public goods and create a platform for sustained partnerships.

Declaration of interests

MC and KP were funded by the Ministry of Health and Welfare of the Republic of Korea to attend the Precision Public Health Asia 2025 Conference.