These tensions constrain cross-border cooperation. To address that, PathGen applies artificial intelligence (AI) to pathogen genomes alongside clinical and environmental signals but within clear constraints. A federated model keeps raw data such as genomic sequences, clinical records and epidemiological information on national systems while analytical outputs move across borders. This means the intelligence is collective even though the underlying information remains under national control.
“This shows how AI and pathogen genomics can work together to provide actionable intelligence for clinicians and public health authorities. By sharing only essential insights, countries can respond faster to outbreaks while strengthening trust and sovereignty,” says Professor Paul Pronyk, director of Duke-NUS’ Centre for Outbreak Preparedness.
PathGen sits at the centre of the Asia Pathogen Genomics Initiative, led by Duke NUS Medical School’s Centre for Outbreak Preparedness. The effort is backed by the Gates Foundation, Temasek Foundation and Philanthropy Asia Alliance, with technology support from partners including Amazon Web Services, IXO, Sequentia Biotech and the University of Sydney.
More than 50 government and academic institutions across 15 countries are involved, with Singapore acting as the coordinating hub. Several Southeast Asian governments have already committed.
See also: Thales steps up Singapore investments across advanced manufacturing and digital tech
Underpinned by the cloud
Making this sovereignty-first model work in practice requires computing power without centralisation. Three elements of cloud infrastructure underpin PathGen’s approach.
Firstly, Amazon Bedrock provides controlled access to large language models for complex analysis while ensuring processing remains within designated geographic regions. Sensitive data is not logged or replicated across borders, and remains encrypted throughout. If the system is analysing drug-resistant tuberculosis in Thailand, raw patient data never leaves Thai infrastructure.
See also: Local digital asset exchange DigiFT offers tokenised access to Hines’ US$6 bil real estate portfolio
Secondly, Amazon’s Relational Database Service stores and indexes diverse datasets, from genomic sequences to population and environmental information. As intelligence accumulates over time, this supports faster pattern recognition and earlier warning of emerging threats.
Thirdly, Amazon EC2 supplies the high-performance computing needed to run fine-tuned AI models in-region. Training and inference remain local, maintaining the same data-residency and security controls required for real-time threat assessment.
The distinction matters. Systems that require data to be centralised tend to falter when trust is lowest. Federated approaches lower the threshold for participation by aligning technical design with political reality.
What's next
What happens next will determine whether the model holds. PathGen is expected to move from proof of concept to pilots in early 2026, followed by a staged regional rollout through 2027. Participating countries will need to align governance frameworks, integrate the platform with existing surveillance systems and invest in laboratory capacity to generate timely, usable data.
“Every delay between detecting a pathogen and making the right public-health decision costs lives. A shared intelligence system that protects sovereignty, cuts response time, and stops outbreaks before they become crises – that’s the future of health security and preparedness,” says Lee Fook Kay, head of pandemic preparedness at Temasek Foundation.
