On Friday, June 27, 2025, in Tokyo, ACCESS Health International formally launched its newest publication, “Towards Responsible Health AI in Asia Pacific,” at the AI Policy Summit 2025. Represented by Iman Hameed, Associate Director, ACCESS Health contributed key perspectives on the promise and the perils of artificial intelligence (AI) in healthcare settings across the region.
Developed in collaboration with Amazon Web Services (AWS) and informed by six months of desk research, stakeholder interviews with Health AI policy and regulatory experts, and technical reviews, this report offers policymakers, regulators, healthcare providers, and technology developers a rigorous, evidence‑based assessment of Health AI adoption, national policy landscapes, and key initiatives across the Asia Pacific. It outlines current challenges and collaborative pathways to advance safe, equitable, and trusted AI‑driven health innovations.
Framing the Landscape: What Is “Health AI”?
“Health AI” refers to any application of machine learning or other AI techniques to support clinical decision‑making, patient monitoring, administrative workflows, or public health surveillance. In this report, we specifically include:
- Generative AI and Large Language Models (LLMs): Systems like ChatGPT or domain‑specific LLMs that can generate clinical notes, patient communications, or medical literature summaries.
- AI‑Software as a Medical Device (AI‑SaMD): Software algorithms that perform diagnostic or therapeutic functions, such as identifying diabetic retinopathy in retinal images, are subject to medical device regulations.
- Predictive Analytics Tools: Statistical or machine‑learning models that forecast patient risk (e.g., sepsis onset) or population‑level disease trends.
By clarifying these categories, the report ensures that stakeholders speak a common language when setting policies and evaluating use cases.
Six‑Pillar Framework: From Policy to Practice
To benchmark each country’s readiness and identify gaps, we introduce a six‑pillar framework:
- Policy & Strategy: National AI strategies, digital health roadmaps, and cross‑sectoral collaboration bodies that set vision and objectives.
- Legislation & Regulation: Laws, standards, and guidelines – such as data protection acts, medical device regulations, and liability frameworks – that govern AI development and deployment.
- Guidance on Responsible Use: Best‑practice documents, ethical codes, and technical standards that instruct developers and users on topics like algorithmic fairness, informed consent, and explainability.
- Assurance Mechanisms: Processes and institutions – approval bodies, certification schemes, post‑market surveillance systems – that provide ongoing oversight of deployed AI tools.
- Research: Academic and industry research initiatives that generate evidence on AI performance, bias mitigation strategies, and health impact assessments.
- Education & Training: Capacity‑building programs for regulators, clinicians, data scientists, and patients – ensuring everyone understands AI’s capabilities, limitations, and safeguards.
By assessing each country against these interlocking pillars, the report highlights both strengths (e.g., robust data‑privacy legislation in Singapore) and areas for improvement (e.g., limited post‑market monitoring in several Southeast Asian nations).
A Risk‑Based, Use‑Case‑Driven Approach
Not all AI applications carry the same level of risk. To help institutions prioritize their efforts, the report outlines a risk‑management methodology:
- Hazard Identification: Cataloguing potential harms, such as incorrect treatment recommendations, biased patient triage, or data breaches.
- Risk Scoring: Quantifying likelihood and severity on a standard scale, enabling comparison across use cases (e.g., an LLM generating clinical notes versus an AI tool for remote surgery assistance).
- Mitigation Strategies: Technical controls (e.g., bias detection algorithms, encryption), organizational policies (e.g., defined human‑in‑the‑loop checkpoints), and contractual measures (e.g., vendor liability clauses).
- Continuous Monitoring: Establishing post‑deployment surveillance – collecting performance metrics, user feedback, and incident reports to trigger timely recalibration or withdrawal if needed.
Practical templates and real‑world examples guide health systems through each step, ensuring they can tailor the approach to their local context and resource constraints.
In‑Depth Country Profiles
The report presents 12 detailed country profiles, covering:
- Australia & New Zealand: Advanced digital health infrastructure and well‑established medical device regulations, paired with emerging sandboxes for AI innovation.
- India & Indonesia: Rapidly expanding health tech ecosystems but heterogeneous regulatory enforcement and uneven data privacy standards.
- China & Japan: Strong state‑led AI strategies, with China’s national AI plan and Japan’s emphasis on AI‑SaMD guidance under the Pharmaceuticals and Medical Devices Agency (PMDA).
- Southeast Asia (Malaysia, Philippines, Singapore, Thailand, Vietnam): Varied maturity levels – from Singapore’s rigorous governance frameworks to Vietnam’s nascent but ambitious AI roadmaps.
Each profile includes an executive summary of key policies, pending legislative proposals, pilot projects, and capacity‑building initiatives. Comparative tables allow cross‑country learning and identification of best practices.
Five Strategic Recommendations
To catalyse progress, the report issues five core recommendations:
- Strengthen National AI Governance: Develop or update holistic AI strategies that explicitly address health sector needs.
- Harmonize Regulations: Align AI‑SaMD guidelines, data protection rules, and liability frameworks across borders to facilitate safe innovation.
- Build Regulatory Capacity: Invest in training for regulatory staff, accreditation of testing laboratories, and technical advisory panels.
- Streamline Assurance Processes: Implement adaptive, risk‑proportional approval pathways (e.g., regulatory sandboxes, accelerated reviews) coupled with robust post‑market surveillance.
- Foster Sustainable Workforce Development: Launch multi‑stakeholder education programs, from medical curriculum integration to public literacy campaigns, ensuring a broad understanding of AI’s benefits and risks.
These recommendations are accompanied by actionable next steps, potential funding mechanisms, and suggested timelines.
Conclusion
“Towards Responsible Health AI in Asia Pacific” offers a clear, actionable roadmap for harnessing AI’s transformative potential in healthcare while safeguarding patient safety, equity, and public trust. By uniting policymakers, health system leaders, regulators, and technology developers around a shared vision, the Asia Pacific can emerge as a global exemplar of responsible Health AI deployment.