© 2026 NervNow™. All rights reserved.
ByNN Desk
Google AI is Helping Taiwan Predict & Prevent Diabetes at Scale
Google and Taiwan's National Health Insurance Administration uses Gemini and 20 years of patient data to cut diabetes risk assessments from 20 minutes to 25 seconds, and the two sides say the model is designed to be replicated globally.
A collaboration between Google and Taiwan’s National Health Insurance Administration uses Gemini and 20 years of patient data to cut diabetes risk assessments from 20 minutes to 25 seconds, and the two sides say the model is designed to be replicated globally.
Google and Taiwan’s National Health Insurance Administration have announced a collaboration that uses Gemini and two decades of national health data to bring predictive diabetes care to millions of patients, in what the two parties are calling a blueprint for preventative, predictive, and proactive health systems everywhere. The announcement was published on Google’s official blog by Amy McDonough, Managing Director of Strategic Health Solutions at Google Health.
The first major output of the collaboration is AI-on-DM, Artificial Intelligence on Diabetes Mellitus, an AI model developed by the NHIA under Taiwan’s Ministry of Health and Welfare that assesses diabetes risk at population scale. Previously, a single patient’s risk assessment took an average of 20 minutes. Screening 20,000 patients would have required 40 professionals working non-stop for three weeks. The AI-on-DM model, built on Google Cloud’s concurrency infrastructure, completes the same assessment in 25 seconds per case — a 14,400x efficiency improvement that allows 20,000 evaluations to be completed in under 90 minutes. The model digitizes established clinical logic and flags patterns in patient data for clinical review, assisting doctors in identifying and intervening before complications arise.
Taiwan’s national health system is uniquely positioned for a deployment of this kind. The country operates a single-payer universal healthcare system that has covered almost its entire population for over 20 years, accumulating a unified population-scale database of patient records, test results, imaging, and claims data. That depth of longitudinal data is precisely what makes the AI model viable, and what Google says makes Taiwan an ideal proving ground for a model it intends to scale globally.
Beyond the clinical system, the NHIA is this month launching a Gemini-powered health assistant inside its government app, used by 10 million people in Taiwan. The tool will generate personalized health insights grounded in clinical guidelines, providing secure, daily care support directly to patients. The AI-on-DM project also builds on a series of earlier hospital-level AI collaborations in Taiwan, including China Medical University Hospital’s adoption of MedLM for cancer care planning, Chang Gung Memorial Hospital’s AI-enhanced ultrasound diagnostics for breast cancer, and Taipei Medical University Hospital’s use of automated clinical workflows to address staff shortages. The NHIA has also used MedGemma to process over 30,000 pathology reports.
“Taiwan is transforming public health with Google…the partnership turns 20+ years of securely-aggregated data into proactive care for millions of patients.”
Amy McDonough, Managing Director, Strategic Health Solutions, Google Health
To extend access beyond urban centers, Google.org — Google’s philanthropic arm — has awarded a $1 million grant to the Digital Humanitarian Association, which will use the funding to bring diabetes management services and digital health training to 300 community centers across Taiwan, support 240,000 health check-ins, and train 200 local caregivers of diabetes patients. The NHIA has stated it plans to apply the same AI framework to hypertension and hyperlipidemia in subsequent phases.
ALSO READ: Google Pledges $15 Billion to Build India’s First Full-Stack AI Hub
The announcement frames the Taiwan deployment not as a local initiative but as a template. With population-scale coverage, 20 years of unified data, and government backing, Taiwan offers conditions that few health systems can match, but Google’s stated intent is to demonstrate what is achievable and make those methods transferable. Whether that blueprint travels will depend in part on how well AI predictions hold up under real clinical conditions and whether doctors in other markets trust and act on AI-flagged risk assessments in the way the Taiwan system is designed to encourage.
This article is based entirely on the post published by Amy McDonough on Google’s official blog March 4, 2026. All figures are sourced directly from that post. NervNow has not independently verified clinical performance claims.
MORE ON GOOGLE
Meet Gemini 3.1 Flash-Lite, Google’s Most Cost-Efficient AI Model Yet
Nano Banana 2 is Google’s Fastest Image AI Yet
Google Launches AI Professional Certificate on Coursera
Alphabet’s Intrinsic Robotics Unit Officially Joins Google
Google Publishes 2026 Responsible AI Report Amid Growing Scrutiny
