Challenges Gap in data ownership/access Gap in digital public infrastructure Key issue Digital technologies can be equalizers, but risk increasing disparities if not managed carefullyA comprehensive DPI approach is crucial for bridging the digital divide and ensuring equitable access to AI-driven health Transverse mitigation There is a need to prioritize and distribute computing resources effectively, including investing in cloud infrastructure that supports both high- performance computing and extensive storage solutions Ensuring equitable access to computational resources is essential for enabling widespread and fair use of AI technologies Transitioning to an open health system, where data is shared for the common good, requires a cross-country approach with harmonized data regulations and standards mandated by governments Specific mitigation Encouraging local ownership and transferring AI technologies to under-represented regions is vital to ensure that benefits are shared globally and to prevent rapid innovation leading to unequal data ownership and deepening inequalitiesComprehensive datasets, including SDOH, can help addressing bias, privacy, accuracy and quality in AI Prioritizing local production and ownership of DPIs will support digital equity Locally controlled but globally federated datasets and a secure sandbox environment for testing AI models are required to facilitate global collaboration while maintaining data privacyPrioritizing digital identities and internet access for the disconnected, along with providing online computing and storage resources, will support digital equity There is a need to validate AI algorithms on local data to make sure that services are accurate, relevant and reflective of the local context; this helps build trust and ensure that AI models are trained on data that truly represents the local population it servesHealth alone might not provide the scale for an effective DPI approach, and should be considered as one service need within a broader DPI strategyTABLE 2 Equity in AI initiatives in the health sector Of course, alongside this training of decision- makers, it is also important to help patients, health professionals and caregivers to understand AI in order to develop greater trust and promote the adoption of public digital health tools. There is a common belief that AI is here to replace people, so building trust involves demonstrating that AI is not here to replace but to complement healthcare professionals. Physicians often view AI as a competitor rather than as a tool, which calls for a stronger focus on education, and unleashing the full potential of AI in health will require addressing this concern to enhance trust among caregivers. Educating the general public about the benefits and purposes of AI-driven health solutions will lead to a higher degree of acceptance and a more effective use of these technologies. Understanding the implications of technical decisions is critical As leaders grapple with the complexities of making strategic technical choices in healthcare, the consequences of disengagement are profound – leading to fragmented systems and missed opportunities for cohesive AI integration. Addressing these challenges is not just about making informed decisions, it is also about bridging the gaps in locally developed digital infrastructure and literacy among decision-makers. This groundwork is essential for building the trust and regulatory frameworks necessary to support AI’s successful implementation. Yet the road ahead is fraught with obstacles, particularly in navigating the fragmented and evolving landscape of AI regulation, which poses significant challenges to establishing trust and ensuring the safe, effective deployment of AI technologies in health. i The Future of AI-Enabled Health: Leading the Way 19

The Future of AI Enabled Health 2025