At the 2025 Smart Healthcare Forum held in Kaohsiung, Tu highlighted that “Kaohsiung’s AI strengths are not just in the exhibition halls, but all around us.” With support from the Kaohsiung City Government, the city has partnered with Taiwan AI Labs, Kaohsiung Medical University Chung-Ho Memorial Hospital, Kaohsiung Chang Gung Memorial Hospital, Kaohsiung Veterans General Hospital, and E-DA Hospital to form the world’s first Federated Learning Smart Healthcare Alliance, establishing a complete AI governance system covering software, hardware, and institutional frameworks.

Since 2021, Taiwan AI Labs has collaborated with the National Development Council to promote federated learning, building a nationwide, cross-center data governance platform. “Data doesn’t leave the hospital or institution. Each unit can train AI locally without moving raw data, protecting privacy and improving efficiency,” Tu explained.

Currently, the Taiwan Federated Learning Medical Alliance has created 161 cross-hospital federated learning models across specialties such as orthopedics, cardiology, and nephrology. These models provide clinical solutions in areas like disease diagnosis, medical image analysis, voice-driven forms, drug development, mental health, and workflow automation. “These results weren’t developed in labs—they were validated day in and day out in real clinical settings,” Tu emphasized.

AI is Not Omnipotent — It Needs Medical Expertise

“Hospitals don’t need a GPT that chats. They need an AI assistant that understands surgeries, medical images, and patient records,” said Tu. While large language models like ChatGPT perform well in various fields, the vast amount of non-professional training data makes them unsuitable for direct clinical use.

Tu introduced FedGPT as a future solution tailored to healthcare and other high-sensitivity fields like finance. These mission-specific models will be capable of autonomous learning and updating, all within hospital environments. “It’s like a trained professional employee—dedicated to you, not leaking information, and not shared with others,” he said.

From Taiwan to Europe: A Global Benchmark for Federated AI

In contrast to the U.S.’s focus on data centralization, Tu proposed the “ReGenerative AI” framework, emphasizing trust, responsibility, and sustainability—a concept that has garnered attention from several European countries. He was invited to the AI for Action Global Summit in Paris, where he showcased Taiwan’s leadership in medical AI data governance.

“The EU AI Act and GDPR are currently the world’s strictest regulations on data and AI. Our federated learning model aligns perfectly with these rules,” Tu noted. “AI is not just a tech race—it’s a global battleground for ethical governance and trust. Taiwan’s approach is not only viable, but export-ready.”

Facing Demographic Challenges with AI Collaboration

Dr. Chao-Yuan Wang, Superintendent of Kaohsiung Medical University Chung-Ho Memorial Hospital, pointed out that Taiwan is facing dual challenges of low birthrate and an aging population. With shrinking newborn numbers and limited medical workforce expansion, future healthcare must rely on AI-human collaboration to reduce clinical errors and improve efficiency. Taiwan’s robust National Health Insurance and medical infrastructure provide a solid foundation for implementing smart healthcare solutions. By combining these strengths with public-private collaboration, Taiwan is poised to export smart healthcare globally and expand its reach into international markets.

Dr. Yi-Shan Tsai, Director of the Clinical Innovation Center at National Cheng Kung University Hospital, also emphasized the need for clearer policy incentives and regulatory flexibility to support AI development. While Taiwan has strict data protection laws, she advocated for accelerated creation of guidelines for data use and de-identification standards.

The Next Step: AI in Every Hospital

“We’re not just theorizing—we’re deploying and experimenting right here,” said Tu. The Kaohsiung Smart Healthcare Alliance is gathering real performance data to help reduce costs for the National Health Insurance, streamline hospital workflows, and improve patient care. He also revealed that the collaboration between AI Labs and the four major medical centers in Kaohsiung has attracted the attention of international organizations such as OECD, Elekta, and NYU Langone Health.

Dr. Huang-Chi Chen, Medical Secretary of Kaohsiung Municipal Siaogang Hospital, stressed that smart healthcare isn’t about showcasing tech—it’s about reducing manpower burdens and improving care quality. He advised that companies should start small, collaborate with local hospitals to build evidence, and scale into larger healthcare systems. “If you want to solve clinical problems, start by working backward from real needs,” he advised.

During the panel discussion, Dr. Wang also noted that deploying AI in healthcare currently requires significant time and manpower. More funding and resources will be needed in the future. Most of the current investments come directly from hospitals, so including smart healthcare in government incentive programs would greatly accelerate development.

Tu’s Final Message: “Keep Data at Home, Learn Nearby, Decide Locally.”

“This is what smart healthcare should look like,” Tu concluded. “And it’s also Taiwan’s greatest competitive edge in the global AI race.”

BOX: What is a Federated Platform?

A Federated Platform utilizes distributed machine learning techniques, allowing data to remain under local control while sharing model parameters for collaborative global model training. This enables alliance members to strengthen anti-fraud capabilities significantly while protecting customer data privacy—providing the public with a secure digital shield for their assets.

Source: Taiwan AI Labs