1. Introduction
The transformative potential of artificial intelligence (AI) in public health is increasingly recognized, offering opportunities for enhanced disease surveillance, personalized medicine and improved healthcare delivery systems. AI technologies can analyze vast amounts of health data, providing insights that can significantly influence public health policy and practice.Reference Al-Hwsali 1 However, the integration of AI into public health raises complex ethical, legal and cultural challenges that necessitate careful consideration. These challenges include issues of data privacy, algorithmic bias and the need for transparency in AI decision-making processes.Reference Murphy 2 The ethical implications of AI in healthcare are profound, as they touch upon fundamental principles like autonomy, confidentiality and the trust that underpins patient-provider relationships.Reference Cudjoe 3 As AI systems become more prevalent, the necessity for ethical governance frameworks that prioritize human-centric values becomes increasingly urgent.
Despite the recognized potential of AI in public health, there exists a significant gap in cohesive international legal frameworks to regulate these technologies effectively. Current AI governance structures are often fragmented, with varying interpretations and implementations of ethical principles across different jurisdictions.Reference Jobin, Ienca and Vayena 4 This lack of a unified approach can lead to disparities in the application of AI technologies, potentially exacerbating existing inequalities in healthcare access and outcomes.Reference O’Shaughnessy 5 Furthermore, the lack of comprehensive guidelines for AI use in public health may impede these technologies’ widespread adoption by practitioners and policymakers. Hence, there is a pressing need for an adaptive public international law framework that can address these challenges while promoting ethical and inclusive AI regulation in public health.
This study seeks to establish a human-centric public international law framework for AI in public health. Grounded in three core pillars — ethical accountability, regulatory adaptability, and transparency — the framework directly addresses critical gaps in existing AI governance models, like the World Health Organization’s (WHO’s) lack of enforceability and the General Data Protection Regulation’s (GDPR’s) rigidity.Reference Farhat 6 Ethical accountability responds to the WHO’s broad ethical guidelines, which, while comprehensive, fail to provide enforceable mechanisms to rectify biases or ensure equitable AI deployment in healthcare. 7 Regulatory adaptability counters the inflexibility of frameworks like the GDPR, whose stringent data localization requirements hinder cross-border collaboration essential for global health initiatives. 8 Finally, transparency addresses the opacity of AI systems, a limitation observed in both the GDPR’s vague “right to explanation” clause 9 and the WHO’s insufficient focus on explainable AI (XAI) standards. 10 Integrating these pillars bridges existing gaps while promoting innovation and protecting public health through inclusive, pragmatic, and globally relevant policies. The framework ensures ethical AI accountability beyond voluntary guidelines, establishes adaptable regulatory mechanisms for diverse legal and healthcare contexts, and enhances transparency to facilitate international collaboration without compromising sovereignty. Methodologically, the study employs an interdisciplinary approach, combining legal doctrinal analysis, case studies and a scoping review of international treaties, conventions, regional regulations and scholarly works. This scoping approach allows for a broad mapping of existing AI governance frameworks to identify key themes, gaps and jurisdictional variations.
This study emphasizes the critical need for adaptive legal frameworks that reconcile ethical principles with technological innovation in AI-driven public health systems. Prioritizing inclusivity and accountability, the study proposes actionable strategies to harness AI’s potential for enhancing health outcomes and advancing global dialogue on ethical governance. Fostering collaborative regulatory environments that balance innovation with public health safeguards is paramount. Besides, developing a cohesive and adaptive public international law framework that prioritizes ethical considerations and inclusivity is essential for harnessing the full potential of AI in improving public health outcomes. This research contributes to the ongoing discourse on AI governance by providing actionable insights for stakeholders across the public health landscape.
The remainder of this manuscript is organized as follows: Section 2 provides a scoping review of the theoretical foundations and existing literature on human-centric AI principles, public health ethics, and international law. Section 3 details the research methodology. Section 4 presents the results, identifying key challenges in AI governance, like data sovereignty conflicts and cultural fragmentation. Section 5 presents three case studies, applying the framework to real-world AI governance challenges. Finally, Section 6 highlights the framework’s innovative features, including its adaptability, inclusivity, and cross-border applicability, and provides a comparative analysis with existing models. Limitations and future research directions are explored while underscoring the framework’s contributions to fostering ethical and effective global AI governance in public health and its implications for policymakers and practitioners.
2. Literature Review
2.1. Theoretical Foundations
Human-centric AI (HCAI) principles are crucial for establishing an ethical and inclusive framework for AI governance, particularly in public health. HCAI emphasizes inclusivity, accountability, and transparency, ensuring that diverse perspectives are integrated into AI systems, thereby enhancing their relevance and effectiveness across different populations. This is particularly significant in public health, where disparities in health outcomes can be exacerbated by technology that does not consider the needs of marginalized groups.Reference Jae Moon 11 Furthermore, participatory governance in AI development is essential to ensure that all stakeholders, including patients, healthcare providers, and policymakers, have a voice in designing and implementing AI technologies. 12 Accountability in HCAI refers to the mechanisms that hold AI systems and their developers responsible for their outcomes. This is particularly relevant in public health, where AI can significantly influence health decisions and policies. The need for accountability is underscored by the potential for AI to perpetuate biases or make erroneous decisions that could adversely affect health outcomes. 13 Transparency complements accountability by requiring that AI systems operate in an understandable manner interpretable by users and stakeholders. This transparency is vital for fostering trust in AI applications, especially in sensitive areas like healthcare, where decisions can have profound implications for individual well-being.Reference Camilleri 14
Meanwhile, public international law plays a pivotal role in global health governance, particularly in the context of emerging health challenges and the regulation of AI technologies. The 2005 International Health Regulations (IHR) serve as a legal framework that obligates countries to develop national laws and regulations that align with international health standards.Reference Indriati, Yuliantiningsih and Wismaningsih 15 This framework emphasizes the need for countries to cooperate and coordinate their responses to public health emergencies, thereby enhancing global health security.Reference Katz 16 Human rights law’s (HRL’s) integration into public health governance further strengthens this framework by establishing universal standards that guide government obligations and facilitate accountability.Reference Gostin 17 HRL not only frames health-related issues as rights violations but also provides a normative basis for evaluating health practices and policies. 18
Drawing further, the intersection of public international law and health governance has become increasingly significant in managing global health crises, like the COVID-19 pandemic. The pandemic underscored the need for adaptive legal frameworks that ensure equitable access to health resources while responding to rapidly evolving health threats.Reference Zhang 19 The Global Health Security Agenda (GHSA) exemplifies efforts to strengthen legal and regulatory capacities for preventing, detecting and responding to public health emergencies, highlighting the critical role of legal frameworks in addressing global health challenges. 20 Overall, HCAI governance is grounded in the principles of inclusivity, accountability and transparency, while public international law provides a structured approach to global health governance. Together, these elements form a comprehensive framework for ethical and inclusive AI regulation in public health.
2.2. AI Applications in Public Health
AI technologies hold transformative potential in public health, particularly in diagnostics, epidemiology, and healthcare management. In diagnostics, AI enhances accuracy and reduces errors, as demonstrated in systems detecting diabetic retinopathy with precision rivalling specialists.Reference Abràmoff 21 Epidemiologically, AI enables real-time disease tracking and predictive modelling, as seen during the COVID-19 pandemic.Reference Vayena 22 AI technologies can optimize resource allocation and operational efficiency within healthcare management, aligning with the WHO’s priorities for digital health innovation.Reference Andreychenko 23 However, these advancements are hindered by the following systemic gaps in governance:
Ethical accountability: Algorithmic bias persists due to unrepresentative training data, leading to disparities in diagnostic accuracy across demographic groups. For example, racially biased referral algorithms disproportionately underestimate the care needs of Black patients.Reference Morgenstern 24 Weak accountability frameworks fail to mandate corrective measures or redress mechanisms.
Regulatory adaptability: Fragmented data-sharing protocols exemplified by conflicts between the GDPR’s strict localization requirementsReference Grunhut, Marques and Adam 25 and low-resource nations’ nascent regulationsReference Brand 26 impede cross-border collaboration critical for pandemic response. Rigid frameworks struggle to accommodate evolving technologies like generative AI.
Transparency: Public distrust stems from opaque AI decision-making, particularly in high-stakes applications (e.g., triage systems). While Explainable AI (XAI) principles are advocated,Reference Tang and Cai 27 few governance models enforce transparency audits or patient-facing interpretability standards. These gaps exacerbate risks like biased outcomes, inequitable access and eroded public trust. For instance, AI-driven mental health chatbots risk cultural insensitivity when deployed without localized training, 28 while diagnostic tools in low-resource settings often lack accessibility infrastructure. 29 Centering ethical accountability, adaptability and transparency helps governance frameworks mitigate these challenges while harnessing AI’s benefits.
2.3. Ethical Challenges in AI Governance: Accountability and Transparency
The integration of AI in healthcare raises significant ethical challenges that demand robust accountability mechanisms and transparency standards to ensure equitable and responsible use. Through a systematic review of AI governance literature, legal frameworks and empirical case studies, this study identifies three interconnected ethical priorities under the umbrella of ethical accountability and transparency: mitigating bias, ensuring accessibility, and upholding informed consent. While we acknowledge that other concerns exist, these were selected based on (i) their consistent citation in regulatory debates as pressing ethical risks, (ii) their significant impact on equity, accessibility and transparency in AI-driven healthcare and (iii) their incomplete resolution in existing governance models.
2.3.1. Ethical Accountability: Mitigating Bias and Ensuring Accessibility
Algorithmic bias remains a critical accountability gap, as AI systems trained on unrepresentative datasets perpetuate disparities in healthcare delivery. For instance, racially biased referral algorithms have systematically underestimated the care needs of Black patients, reflecting historical inequities embedded in training data. 30 Ethical accountability requires governance frameworks to mandate rigorous bias audits, corrective measures (e.g., dataset diversification) and redress mechanisms for affected populations. Accessibility gaps further underscore accountability failures, as marginalized communities often face barriers to AI-enabled healthcare. For example, diagnostic tools in low-resource settings frequently lack infrastructure for deployment, exacerbating health inequities. 31 Accountability mechanisms must, therefore, enforce equitable resource allocation, requiring developers and policymakers to prioritize accessibility audits and infrastructure investments in underserved regions.
2.3.2. Transparency: Enabling Informed Consent and Trust
Informed consent is fundamentally compromised when AI systems operate as “black boxes,” obscuring how data is used or decisions are made. Patients cannot meaningfully consent to AI-driven interventions without a clear explanation of risks, benefits and algorithmic logic. 32 Transparency standards must mandate user-friendly explanations (e.g., visual dashboards for clinicians and plain-language summaries for patients) and integrate AI literacy into clinical workflows. Transparency also fosters public trust eroded by opaque AI decision-making in high-stakes applications like triage systems. XAI principles, like post-hoc interpretability methods, should be legally enforced for critical health applications. 33 Concurrently, participatory governance engaging patients, clinicians, and marginalized communities in AI design ensures transparency aligns with societal values. 34
2.3.3. Governance Implications
These ethical imperatives necessitate adaptive governance structures. For instance, Kenya’s codesign of a maternal health AI tool with local midwives improved rural adoption rates by 40%,Reference Tshimula 35 demonstrating how accountability and transparency protocols can address contextual inequities.Reference Mumo 36 Similarly, Norway’s XAI-driven consent form implementation reduced patient mistrust,Reference de Souza Filho 37 highlighting the practical impact of transparency mandates. Centering ethical accountability and transparency enables governance frameworks to bridge the gap between AI’s transformative potential and ethical risks, ensuring technologies serve as equitable tools for public health advancement.
2.4. Research Gaps
International law principles’ integration into AI governance remains insufficient, particularly in addressing the transnational complexities of public health. While ethical frameworks for AI proliferate, many lack legal enforceability or fail to reconcile with established international legal norms. For instance, proposals to recognize AI as a limited legal subject inadequately address harmonization with existing human rights law or cross-border liability mechanisms.Reference Xudaybergenov 38 Similarly, Greene and others critique the tendency of ethical AI discourses to sidestep legal accountability, underscoring the need for governance models that bridge ethical aspirations with binding legal standards.Reference Greene, Hoffmann and Stark 39
A critical gap persists in regulatory adaptability, particularly frameworks that balance global interoperability with localized socio-cultural and legal contexts. Current governance models often adopt rigid, “one-size-fits-all” approaches ill-suited to diverse health systems. For example, the GDPR’s stringent data localization requirements conflict with low-resource nations’ needs for agile health data sharing during emergencies. 40 Regulatory adaptability must extend beyond technical flexibility to incorporate cultural sensitivity, ensuring AI systems align with regional ethical norms, health practices and community trust dynamics. Jobin, Ienca and Vayena note that most AI ethics guidelines neglect cultural nuances, like collectivist societies’ prioritization of communal health outcomes over individual data privacy. 41 Adaptive frameworks must enable jurisdictions to tailor transparency protocols, consent mechanisms and bias mitigation strategies to local contexts while upholding universal human rights standards.Reference Zhang and Zhang 42 Finally, HRL’s underutilization in AI governance limits its potential to enforce equity and accountability. While Su argues that HRL provides a robust foundation for AI regulation,Reference Su 43 its application remains theoretical rather than operational. For instance, the right to health (United Nations Universal Declaration of Human Rights Article 25) could mandate equitable AI resource distribution, yet few governance models explicitly link AI deployment to HRL obligations. Closing these gaps requires interdisciplinary collaboration to develop frameworks harmonizing AI ethics, adaptable regulation and legally enforceable human rights principles.
3. Methodology
This study employs an interdisciplinary methodology, integrating legal doctrinal analysis, public health ethics and AI governance scholarship, to critically evaluate and operationalize the three pillars of ethical accountability, regulatory adaptability, and transparency. The research design combines qualitative case studies with comparative legal analysis, ensuring theoretical rigor while addressing real-world governance challenges.
3.1. Research Design and Data Sources
The methodology is anchored in three complementary approaches:
-
I. Legal doctrinal analysis: International treaties (e.g., International Health Regulations (2005)), regional regulations (e.g., GDPR, European Union (EU) AI Act) and national frameworks (e.g., Brazil’s AI Bill) were analyzed through a scoping review to map regulatory gaps and adaptability challenges like rigid data-sharing protocols hindering cross-border collaboration. This process followed established scoping review frameworks, such as Arksey and O’Malley’s methodology, to identify key themes and jurisdictional variations without exhaustive quality assessment. Legal principles (e.g., sovereignty, proportionality) were adapted to propose modular governance structures.
-
II. Case studies: Three case studies on public health AI applications further enriched the analysis, focusing on AI-driven disease surveillance and data sovereignty (Case Study 1: The EU and the Global South), bias in AI diagnostic tools (Case Study 2: The United States and the WHO) and AI diagnostic tools (Case Study 3: Kenya and Sweden). These cases were chosen to highlight failures in ethical accountability (e.g., biased US diagnostics), 44 transparency and adaptability (e.g., opaque EU-Nigeria data-sharing during COVID-19),Reference Akinloluwa 45 and inclusivity gaps linked to AI-enabled health interventions.
-
III. Policy briefs and research outputs from leading think tanks like WHO, UNESCO, the Organisation for Economic Co-operation and Development, and AI oversight bodies: This ensured empirical grounding and provided contemporary perspectives on governance challenges and a robust legal foundation to address the regulatory complexities associated with AI technologies. Cross-jurisdictional comparisons revealed how adaptable frameworks could resolve accountability gaps (e.g., Kenya’s codesigned HIV and tuberculosis (TB) tool). 46
Sources were identified through databases (Cambridge University Press, PubMed, Elsevier, Taylor and Francis) and gray literature using keywords like “AI governance” and “public health ethics.” Inclusion criteria prioritized jurisdictional diversity and thematic relevance to align with scoping review objectives.
3.2. Analytical Framework
A criteria-based framework guided the evaluation of governance systems:
-
• Ethical accountability: Assessed through bias audits, redress mechanisms, and equity impact metrics derived from public health ethics literature.
-
• Regulatory adaptability: Evaluated via modularity (e.g., Association of Southeast Asian Nations’ (ASEAN’s) regional sandboxes) and interoperability of cross-border protocols (e.g., Fast Healthcare Interoperability Resources (FHIR) integration).
-
• Transparency: Measured using XAI standards (e.g., post-hoc interpretability) and public trust indices (e.g., patient consent rates).
3.3. Validation and Limitations
Methodological rigor was ensured through triangulation (comparative legal analysis, thematic coding, and synthesis), consistent with scoping review objectives to map broad trends rather than answer narrow research questions. Legal findings (e.g., GDPR limitations) were cross-validated with case study outcomes (e.g., EU-Nigeria data conflicts) to strengthen adaptability recommendations. Ethical principles (e.g., WHO guidelines) were tested against empirical bias reports to refine accountability mechanisms. This approach ensures the framework is both theoretically robust and pragmatically actionable, addressing the complex interplay of ethics, law and technology in public health AI governance. The methodology’s strength lies in its synthesis of disciplines; (i) legal scholarship-informed scalable regulatory models for adaptability, (ii) public health ethics-grounded accountability mechanisms in equity principles, like justice and beneficence, and (iii) AI governance research-shaped transparency tools, balancing technical feasibility (e.g., XAI interfaces) with sociocultural needs (e.g., multilingual disclosures). Limitations include regional bias in case selection (e.g., underrepresentation of Asia-Pacific contexts) and reliance on self-reported bias incidents (e.g., U.S. diagnostics), 47 which may underestimate systemic accountability gaps. These limitations reflect the scoping review’s focus on breadth over depth, as well as pragmatic constraints in accessing global data.
4. Results
4.1. Key Challenges in AI Regulation
As AI technologies continuously transform public health systems globally, significant challenges emerge in their regulation, particularly when addressing data sovereignty, cultural and legal fragmentation and cross-border governance issues. These challenges must be addressed to ensure that AI is ethically, inclusively and effectively deployed in public health. This section outlines two key challenges that directly impact AI regulation.
4.1.1. Data Sovereignty Conflicts: Navigating Health Data Across Jurisdictions
Data sovereignty refers to the legal principle that data is subject to the laws and governance structures of the country in which it is collected, processed, or stored. This principle becomes particularly challenging when dealing with health data, which is often cross-border in nature. AI technologies in public health rely on vast amounts of data, including sensitive health information, which can be difficult to manage across multiple jurisdictions with differing legal standards. The key issues include:
Data protection regulations: Jurisdictions like the EU have implemented robust data protection laws like GDPR, which provide strict rules regarding consent, data access and the movement of personal data across borders. However, these laws can conflict with data-sharing requirements in other regions or countries that may not have equivalent protections in place. This creates barriers to AI-driven health interventions that rely on international data flows, like those for pandemic response or global disease surveillance. 48 The challenge lies in reconciling these regulatory differences to facilitate a human-centric approach to data governance that respects individual rights while promoting public health objectives.
Cross-border data transfers: The movement of health data across borders for AI processing and analysis is frequently obstructed by data sovereignty concerns.Reference Yun 49 This is particularly problematic in health emergencies where global cooperation and data sharing are essential for effective responses. Regulatory frameworks must reconcile these conflicts, ensuring that data sovereignty is respected while promoting international collaboration for public health benefits.Reference Cihon, Maas and Kemp 50 Establishing a common data-sharing framework that balances local laws with global health needs is essential for effective AI deployment.
Privacy and ethical considerations: The lack of uniform standards for health data privacy leads to ethical dilemmas. 51 In some jurisdictions, AI-driven health interventions may be hindered by privacy laws limiting the use of health data, while in others, data is shared with few restrictions. Finding common ground on data sharing, security and privacy standards in a way that maintains ethical integrity across jurisdictions is a fundamental challenge. 52 This necessitates a human-centric approach that prioritizes transparency and accountability in AI applications, ensuring that ethical considerations are integrated into the regulatory framework.
4.1.2. Cultural and Legal Fragmentation in AI Governance
AI technologies, particularly those applied in public health, are influenced by cultural and legal contexts that vary significantly across jurisdictions. This fragmentation poses challenges for creating a cohesive international regulatory framework that balances ethical AI governance with respect for cultural diversity and legal pluralism. The key issues are given as follows:
Cultural differences in ethics and AI: Different societies have varying conceptions of ethics, justice and privacy, which affect how AI systems are perceived and adopted in public health. 53 For example, individualistic cultures may emphasize personal privacy and consent, while collectivist cultures may prioritize the broader community’s health and well-being. These differences can lead to conflicts in the implementation of AI systems in global public health initiatives, as what is considered ethically acceptable in one jurisdiction may not align with the values of another. A human-centric AI governance framework must account for these cultural differences to foster inclusivity and respect for diverse ethical perspectives.
Fragmentation of legal frameworks: AI regulation remains fragmented due to different countries’ adopted diverse legal, cultural and policy approaches. 54 While this regulatory diversity reflects national priorities and governance structures, it also creates inconsistencies in AI oversight, particularly in public health applications that require cross-border collaboration. Recognizing that comprehensive global AI regulation may not be feasible, this study argues for an international framework that focuses on harmonizing core principles like transparency, accountability, and equity while allowing flexibility for national adaptation. This approach aligns with established international regulatory models in other sectors, where global agreements set foundational standards while granting nations autonomy in implementation. For example, the IHR provide a framework for cross-border health security, 55 yet countries retain discretion over national public health policies. Similarly, the GDPR has influenced global data privacy standards, even though each jurisdiction tailors its regulations to local contexts. AI governance can follow a similar model by establishing guiding principles at the international level while enabling nations to customize policies based on their legal and socio-economic conditions. The impact of globalization on AI regulation mirrors challenges seen in other domains, like trade and environmental law, where international cooperation is necessary but does not eliminate national sovereignty. Instead of pursuing a rigid, “one-size-fits-all” global AI regulation, this study advocates for a modular and scalable governance approach that integrates core ethical and legal principles into national frameworks, ensuring both international alignment and local adaptability.
Lack of global consensus: Despite efforts by international organizations like the United Nations (UN) and the WHO to create global standards, 56 there is no universal agreement on key aspects of AI governance. For example, while the EU has proposed strong ethical guidelines for AI development, other nations, particularly in the Global South, may prioritize innovation and economic growth over strict regulation. This lack of consensus hinders the creation of a cohesive international framework for AI in public health, limiting the ability to regulate and deploy AI technologies effectively and ethically on a global scale. A human-centric approach to AI governance must strive for inclusivity and collaboration among diverse stakeholders to foster a shared understanding of ethical AI practices.
The key challenges of data sovereignty conflicts and cultural and legal fragmentation in AI regulation underscore the complexities of governing AI in public health. These challenges require an adaptive, inclusive and flexible international legal framework that not only respects national sovereignty but also fosters global cooperation. Addressing these issues will ensure the ethical and effective use of AI technologies in public health, especially as AI systems become more integrated into global health responses and public health initiatives. The development of such a framework will need to consider the diversity of cultural, legal, and ethical perspectives across jurisdictions while balancing the need for innovation with protecting human rights and privacy. Prioritizing human-centric principles helps stakeholders to work towards a more cohesive and effective regulatory environment that enhances public health outcomes globally.
4.2. Proposed Adaptive Governance Framework
In response to the key challenges outlined in Section 4.1, a robust and adaptive AI governance framework’s development is critical to ensuring AI technologies’ ethical and inclusive deployment in public health. The framework addresses two overarching challenges in AI governance for public health data sovereignty conflicts and cultural-legal fragmentation through three core pillars: ethical accountability, regulatory adaptability and transparency. Each pillar is contextualized below to demonstrate how it resolves these systemic barriers.
4.2.1. Human-Centric Design
This framework prioritizes human-centric design, emphasizing inclusivity, accessibility and fairness as foundational principles for ethical and equitable AI in public health. These principles were selected due to their direct role in addressing ethical, legal and societal challenges in AI deployment. While other principles like privacy, security and autonomy remain critical, inclusivity, accessibility, and fairness are essential, particularly for mitigating bias, reducing healthcare disparities, and ensuring equitable AI adoption across diverse populations. Furthermore, they align with international legal standards, including the WHO AI ethics guidelines.
Cultural and legal fragmentation, as evidenced by conflicting privacy norms (e.g., the EU’s GDPR 57 vs. Nigeria’s Nigeria Data Protection Regulation)Reference Tripathi, Rathore and Singh 58 and cultural biases in AI tools (e.g., mental health chatbots in China), 59 necessitates an inclusive approach to AI development. Participatory design, involving marginalized communities, is crucial to align AI systems with local values and ensure user-friendliness across diverse populations, including individuals with disabilities and economically disadvantaged groups. For example, a maternal health AI tool codesigned with Kenyan midwives achieved a 40% higher adoption rate in rural clinics, demonstrating the efficacy of this approach. 60 Accessibility is vital to prevent AI-driven healthcare solutions from deepening digital divides. Ensuring compliance with assistive technologies (e.g., Web Content Accessibility Guidelines (WCAG) 2.1) and addressing infrastructural gaps, as evidenced by the challenges faced in Sweden’s rural AI diagnostic deployment, 61 is crucial for promoting equitable access to AI-enabled healthcare. Fairness is operationalized through bias audits and equity standards to mitigate algorithmic discrimination and prevent inequitable treatment recommendations, particularly in contexts where AI systems trained on biased datasets may perpetuate healthcare disparities. For instance, racial biases embedded in US diagnostic algorithms underscore the urgency of embedding fairness mechanisms. 62
To translate these principles into actionable practices, the framework mandates legal mechanisms enforcing transparency across AI design, implementation and operation. These include compulsory disclosure of performance metrics, accessibility audits aligned with standards like WCAG 2.1,Reference Surjit 63 and user satisfaction evaluations, with noncompliance triggering penalties. Concurrently, cultural sensitivity is maintained through UNESCO or WHO-certified training for developers and policymakers and the adaptation of AI systems to regional health practices, including traditional medicine databases. Stakeholder collaboration is essential, with governments funding community engagement, tech companies establishing ethics boards, and NGOs conducting independent audits to ensure inclusivity and accountability. This human-centric approach promotes equitable, accessible and culturally respectful AI deployment in healthcare.
4.2.2. Modular and Scalable Regulation
This framework recognizes the challenges posed by diverse regulatory environments and data sovereignty conflicts, as exemplified by the delays in EU-Nigeria data sharing during the COVID-19 pandemic. 64 To address these challenges, it proposes a modular and scalable approach to AI governance, offering the flexibility needed to accommodate varying legal frameworks and public health contexts while ensuring the seamless progression of global health initiatives. This approach emphasizes the development and dissemination of Regulatory Self-Assessment Toolkits, hosted on a digital platform by the WHO or a designated international body. These toolkits would provide a step-by-step approach to AI governance, including a diagnostic tool for countries to evaluate gaps in their existing frameworks and customizable regulatory templates (e.g., GDPR-based data-sharing agreements with public health exceptions).Reference Zeb 65 This modular approach allows for phased implementation, enabling low-resource countries to begin with foundational modules, like data anonymization, while advanced economies adopt stricter transparency protocols. Furthermore, the framework emphasizes the importance of cross-border data protocols to address challenges related to data sovereignty and facilitate international collaboration. Building upon the EU’s GDPR (with the inclusion of a “public health exception” for emergencies),Reference Quinn, Ellyne and Yao 66 these protocols would harmonize data-sharing practices, ensuring that health data can be shared internationally while complying with both local laws and global standards. The adoption of technical standards like FHIR for health data exchange is crucial for seamless and secure data flow across borders.
Stakeholders play crucial roles in this process. International organizations, like the WHO and UN agencies, are responsible for hosting regulatory repositories, mediating disputes and promoting the adoption of these frameworks. National legislators play a pivotal role in adapting these modules into national law, as exemplified by Brazil’s AI Bill, which mirrors the EU AI Act while prioritizing Amazonian health data sovereignty.Reference Mendes, Viola, Søndergaard, de Sá and Barros-Platiau 67 This modular approach, as demonstrated by ASEAN nations, can effectively harmonize AI regulations across regions, as evidenced by their successful reduction of data-sharing and reduced information loss, which provided the best COVID-19 detection performance of 98.06% in federated X-ray image learning settings. 68 Through collaboration and adaptation, this framework offers a practical solution to the complex challenges of AI governance in public health.
4.2.3. Accountability Mechanisms
This framework prioritizes robust accountability mechanisms to ensure the responsible use of AI in public health and foster public trust. Fragmented accountability, exemplified by opaque AI triage systems 69 and the limited enforceability of GDPR “right to explanation” clauses, can erode public trust and hinder the effective governance of AI in healthcare. To address this, the framework recommends the establishment of an International AI Governance Board, modelled after existing international organizations like the WHO or the UN, to oversee the ethical and responsible use of AI in public health applications. Drawing examples from the COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) algorithm, in which a risk assessment tool used in the US criminal justice system was found to be racially biased, disproportionately predicting recidivism for Black defendants,Reference Belenguer 70 the demand for such a board remains high. This board would have the authority to audit high-risk AI systems, like diagnostic algorithms, revoke certifications for noncompliant systems and maintain a public registry of noncompliant entities. Furthermore, the framework mandates XAI for high-stakes decisions (e.g., diagnostics) and multilingual interfaces for vulnerable populations to enhance transparency and ensure inclusivity. Embedding ethical AI principles into global health governance frameworks is crucial for fostering accountability. This would involve establishing global ethical guidelines for AI in public health, encompassing data privacy, human rights protection, non-discrimination and the right to transparency. These guidelines would serve as a foundational framework for all participating nations, ensuring that AI applications in public health serve the common good and uphold basic human values.
Stakeholders play crucial roles in this framework. Healthcare providers are responsible for reporting AI errors to oversight bodies via standardized incident forms. Tech companies are required to submit transparency reports detailing training data sources and bias audits. Developers must obtain certifications, like IEEE CertifAIEd, before deploying AI systems in healthcare and are subject to penalties including fines of up to 2% of global revenue for violations. This multifaceted approach aims to establish a robust and accountable framework for AI in public health, ensuring that these technologies are developed and deployed responsibly, ethically and in a manner that benefits all of humanity.
4.2.4. Transparency and Trust in AI
This framework prioritizes transparency as a cornerstone for building trust in AI within the public health domain. Recognizing that opaque AI systems can erode public trust and hinder the effective governance of AI in healthcare, 71 the framework emphasizes the importance of XAI for high-risk applications like diagnostics and treatment recommendations. While acknowledging the limitations of explainability in certain complex models, 72 like Large Language Models (LLMs) or deep learning models operating as “black boxes,” the framework advocates for a risk-tiered approach. This approach mandates that high-risk AI systems, when used in clinical decision-making, provide interpretable decision pathways through surrogate models or post-hoc explanations to ensure that both clinicians and patients can comprehend AI-generated outputs. This is further strengthened by rigorous clinical trials, third-party audits and the implementation of redress mechanisms to address potential discrepancies in AI-driven decisions. A notable example is Norway’s XAI-driven consent forms, which reduced patient mistrust following their implementation. 73
Building on this, in lower-risk applications (e.g., administrative workflows), AI models that are not inherently explainable, like LLMs, may be permitted under strict safeguards, including real-time human oversight, 74 adversarial testing and bias documentation. 75 As a result, redress mechanisms establishing clear appeal processes allow individuals to challenge AI-generated decisions when discrepancies arise. The framework recognizes the importance of international cooperation and the need to adapt to diverse legal and cultural contexts. This includes the development of standardized data-sharing agreements, and the adoption of technical standards like FHIR for seamless and secure data flow across borders is also essential.
Drawing further, the framework emphasizes the importance of addressing the ethical and societal implications of AI in public health. This includes mitigating bias, ensuring inclusivity, and promoting equitable access to AI-powered healthcare solutions. Recognizing that AI systems trained on biased datasets can perpetuate existing health disparities, the framework mandates predeployment audits to identify and mitigate biases. For instance, racial biases embedded in US diagnostic algorithms 76 and other biased health screening algorithms 77 underscore the urgency of embedding fairness mechanisms to prevent inequitable treatment recommendations.
To ensure inclusivity and address the diverse needs of different populations, the framework emphasizes the importance of human-centric design principles, including accessibility and cultural sensitivity. This includes mandating compliance with accessibility guidelines (e.g., WCAG 2.1) and ensuring that AI interfaces are multilingual and culturally appropriate. For example, the successful implementation of a maternal health AI tool in Kenya, which achieved a 40% higher adoption rate in rural clinics following codesign with local midwives, 78 exemplifies the importance of involving local communities in AI solutions development. Meanwhile, public input is integrated into AI governance through region-specific engagement strategies, like South Korea’s citizen juriesReference Jain and Dienel 79 and Kenya’s codesign workshops with rural healthcare workers. 80 Additionally, AI interfaces must be multilingual and locally adapted to ensure accessibility. A successful example is China’s Xiaoice chatbot, which incorporated regional linguistic norms to enhance acceptance in mental health support applications. 81
Furthermore, the framework underscores building trust through transparent communication and stakeholder engagement. This includes public awareness campaigns, like the WHO-sponsored “AI in Health: What You Need to Know” campaign and the development of educational resources to improve AI literacy among healthcare professionals. Legislating “right to explanation” policies compel AI developers to disclose the logic behind AI decision-making, while media collaborations help demystify AI risks and benefits for the general public. For instance, the framework encourages the adoption of the EU’s GDPR while acknowledging the need for regional adaptations, like the inclusion of “public health exceptions” 82 as demonstrated by the Nigerian Data Protection Regulation. By fostering collaboration among stakeholders, including governments, researchers and the public, this framework aims to ensure that AI technologies in public health are developed and deployed ethically, responsibly and in a manner that benefits all of humanity.
5. Case Studies: Applying the Framework to Real-World AI Governance Challenges
To illustrate the practical application of the proposed adaptive AI governance framework, this section presents three case studies demonstrating how it can address specific ethical challenges in AI-driven public health initiatives across different jurisdictions and international bodies. These examples highlight the framework’s ability to navigate legal fragmentation, data sovereignty conflicts and inclusivity concerns.
5.1. Case Study 1: AI-Driven Disease Surveillance and Data Sovereignty — The EU and the Global South
AI-powered epidemiological tools, like the BlueDot system, have proven instrumental in tracking disease outbreaks. 83 However, their deployment raises significant ethical concerns, particularly regarding cross-border data sharing. A notable issue is the exploitation of health data from low-resource countries by high-income nations, 84 often without reciprocal benefits. For instance, the EU’s GDPR enforces stringent data localization policies, which complicate the global exchange of health data necessary for effective pandemic responses. Simultaneously, many nations in the Global South lack comprehensive AI regulatory frameworks, heightening the risk of data misuse and inequitable exchanges.
To address these challenges, the proposed governance framework enables low- and middle-income countries income (LMICs) to adopt tailored AI governance modules while adhering to core global data ethics principles. This approach facilitates the establishment of fair data-sharing agreements, ensuring that nations contributing health data also benefit from AI-driven health innovations. Furthermore, international oversight bodies, like the WHO, would play a pivotal role in auditing AI deployments to ensure compliance with ethical standards. For example, the framework mandates transparency and accountability in data usage, mitigating the risks of exploitation and fostering trust among participating nations. If implemented, this framework would promote equitable collaboration between the EU and the Global South, balancing the need for robust data privacy protections with the imperative of global public health intelligence. By addressing disparities in regulatory capacity and data governance, the framework ensures that AI technologies serve as tools for collective benefit rather than sources of inequity. 85
5.2. Case Study 2: Bias in AI Diagnostic Tools – The United States and the WHO
AI-driven diagnostic tools, like IBM Watson for Oncology, have shown racial and socioeconomic biases due to non-representative training datasets. 86 Such biases have led to disparities in cancer treatment recommendations. 87 The US currently lacks a comprehensive federal AI regulatory framework, leading to inconsistent oversight. 88 Moreover, international bodies like the WHO have yet to establish enforceable global AI equity standards. 89 Amid such observations, concerning the framework’s application, the framework’s human-centric design component mandates participatory governance, requiring healthcare workers, patient advocates and marginalized communities to be involved in AI training. Moreover, the framework’s XAI guidelines would ensure that diagnostic recommendations are interpretable by physicians, mitigating blind reliance on flawed AI outputs. Furthermore, through this framework, global standardization can be achieved. For instance, WHO-backed global equity benchmarks would be integrated into AI regulations, ensuring international consistency in bias detection and mitigation. 90 Implementing these measures would create more reliable and equitable AI diagnostic tools, fostering trust and compliance across jurisdictions.
5.3. Case Study 3: AI Diagnostic Tools in Kenya and Sweden
In Kenya, a low-resource setting, AI tools exhibited racial bias, 91 largely due to the underrepresentation of African datasets during training. This data imbalance can lead to skewed diagnostic outcomes, as AI models often underperform when applied to populations they were not adequately trained on.Reference Alshehri, Alahmari and Alasiry 92 Conversely, in Sweden, a high-resource environment, elderly rural populations faced barriers to accessing AI-driven healthcare, highlighting that disparities in healthcare access persist even in developed nations.Reference Halabhavi 93 To mitigate these challenges, our framework incorporates participatory design workshops with local healthcare workers, ensuring AI retraining on diverse datasets to minimize diagnostic bias.Reference Bhatt 94 Additionally, equity standards mandate monthly accessibility audits to facilitate hardware upgrades in marginalized regions, thereby promoting equitable healthcare access.Reference Igwama 95 In Sweden, the framework also prioritizes transparency protocols,Reference Ek and Öman 96 requiring developers to implement multilingual explainable AI interfaces tailored to elderly users. This initiative fosters trust and enhances comprehension among populations less familiar with advanced technologies. Public awareness campaigns further support this effort by educating rural communities on the benefits of AI, thereby increasing acceptance and adoption. A WHO-led oversight body could monitor performance metrics across both contexts, enforcing penalties for noncompliance with inclusivity benchmarks.
Empirical evidence underscores the framework’s efficacy: in Kenya, a computer vision AI tool achieved 100% sensitivity and 98.8% specificity,Reference Roche 97 while in Sweden’s rural areas, it attained a 90% adoption rate. 98 These outcomes demonstrate that with appropriate frameworks and community engagement, equitable AI deployment is feasible. These case studies illustrate the framework’s adaptability in addressing AI governance challenges across diverse regulatory environments. By integrating modular regulations, transparency mandates and international cooperation mechanisms, the framework advances ethical AI deployment while respecting jurisdictional diversity.
6. Discussion
This study proposed a human-centric AI governance framework grounded in three pillars: ethical accountability, regulatory adaptability, and transparency to address systemic challenges in public health AI deployment. The discussion below evaluates how the findings align with the study’s objectives, emphasizing the framework’s capacity to harmonize ethical principles, legal interoperability and sociocultural inclusivity.
6.1. Contributions to Public Health AI Governance
6.1.1. Ethical Accountability: Strengthening Governance Mechanisms
The framework institutionalizes accountability through bias audits, redress mechanisms, and equity impact assessments. The EU AI Board’s revocation of a racially biased cancer screening tool (The United States and the WHO Case Study (see also Zeb and others)) 99 illustrates how enforceable audits rectify disparities, reinforcing the need for mandatory ethical compliance in AI governance. Beyond the EU, models like Brazil’s AI Ethics Guidelines and India’s National Digital Health Blueprint provide further examples of accountability mechanisms tailored to diverse healthcare systems. Participatory design is another key accountability measure. The codevelopment of Kenya’s maternal health AI, codeveloped with midwives, highlights how inclusive AI governance ensures marginalized voices shape AI systems, addressing a gap in the WHO’s aspirational guidelines. Expanding such participatory models across low-resource regions would enhance AI’s equity impact while ensuring greater accountability in public health deployment.
6.1.2. Regulatory Adaptability: Overcoming Fragmentation Through Modular Regulations
One of the biggest challenges in AI governance is regulatory fragmentation, where differing national laws impede international collaboration. The proposed framework’s modular regulations offer a scalable solution, as evidenced by ASEAN’s success in reducing data-sharing delays through regionally tailored protocols.Reference Cocq 100 Unlike the rigid structure of GDPR,Reference Elfenbein and Hoffman 101 this framework allows jurisdictions like Nigeria to adopt GDPR-inspired data-sharing agreements while maintaining “public health exceptions” during emergencies 102 (see the EU and Global South Case Study). Thus, our framework ensures that national AI governance structures remain adaptable to evolving health challenges by balancing sovereignty with global collaboration. Further, regulatory adaptability is reinforced through policy layering, where AI governance aligns with existing national health strategies rather than imposing an entirely new regulatory system. Countries like Rwanda, which integrates AI into universal health coverage, offer promising models for embedding AI policies within existing health infrastructures.
6.1.3. Transparency: Enhancing Trust Through Risk-Tiered Standards
Transparency is critical in AI governance, particularly in high-stakes applications like diagnostics. The framework’s risk-tiered transparency model mandates XAI for clinical decisions (e.g., diagnostics), ensuring that patients and practitioners can interpret AI-generated outcomes. This aligns with emerging global regulatory paradigms, including the EU AI Act’s risk-based tiers,Reference Ebers 103 which distinguishes between low-risk and high-risk AI applications. However, while the EU AI Act emphasizes compliance, it lacks mechanisms for fostering public trust. The proposed framework addresses this gap by integrating transparency with public participation strategies. Norway’s implementation of XAI-driven consent forms, which reduced patient mistrust, 104 exemplifies how transparency fosters trust, a crucial element missing in many compliance-centric AI models. Additionally, the framework promotes governmental transparency in AI governance. Establishing public AI registries and independent oversight bodies ensures that AI decision-making processes remain auditable and publicly accessible, further reinforcing institutional trust in AI applications.
6.1.4. Advancing Human-Centric AI Principles
This framework distinguishes itself through its commitment to human-centric design, prioritizing inclusivity, accessibility, and fairness. By integrating participatory governance, modular regulation, accountability mechanisms, and transparency, the framework ensures AI governance is adaptable to diverse cultural, legal, and health system needs. To mitigate digital inequalities, the framework mandates accessibility audits and equity standards to monitor biases in AI health applications. 105 Legal mechanisms, like mandatory AI performance reporting and inclusive policymaking, ensure that AI is deployed responsibly without exacerbating disparities in healthcare access. Additionally, the framework emphasizes public participation and stakeholder engagement in AI governance. Participatory decision-making models ensure that AI systems align with societal values, fostering trust and amplifying the voices of marginalized communities. This approach promotes a governance model that genuinely serves all people.
6.1.5. Addressing Regulatory Fragmentation Through International Public Law
Regulatory fragmentation remains a significant challenge in AI governance due to divergent legal, cultural, and policy approaches. The proposed framework bridges this gap by offering a modular and scalable regulatory structure that aligns with global principles while allowing national customization. This flexibility enables countries to tailor AI regulations to their public health contexts, ensuring alignment with international standards on data privacy, human rights, and non-discrimination. The framework also proposes cross-border data protocols to address challenges like data sovereignty and international data-sharing conflicts, like those experienced in Nigeria. 106 These protocols would harmonize data-sharing practices across jurisdictions, ensuring that health data can be exchanged securely and ethically. This framework provides a roadmap for balancing national sovereignty with global health collaboration by establishing clear guidelines on data privacy, security and access.
At the international level, the framework advocates for global oversight bodies to monitor AI use in public health. Establishing WHO-backed AI compliance bodies would ensure accountability, facilitate dialogue between nations, and promote regulatory consistency. These mechanisms support the development of a unified AI governance model, fostering global cooperation while respecting national differences. Furthermore, the framework contributes to building an international consensus on ethical AI standards. Despite regulatory variations across jurisdictions, core principles like transparency, accountability, and human rights protections provide a foundation for regulatory alignment. Promoting ethical AI guidelines in public health through international organizations like the WHO and UN would encourage widespread adoption of universal standards, reducing fragmentation and enhancing interoperability.
6.2. Comparative Analysis
When comparing the proposed adaptive governance framework with existing models like the WHO’s AI initiatives and regional frameworks like the European Union’s GDPR, several key distinctions and complementary aspects emerge.
6.2.1. WHO’s AI Initiatives
The WHO has played a key role in shaping AI governance in public health through its “Ethics and Governance of Artificial Intelligence for Health” framework. 107 While comprehensive in its ethical approach, the WHO model lacks flexibility and mechanisms for cross-border AI governance. Without cross-border cooperation, AI-driven health solutions risk regulatory bottlenecks, data privacy conflicts and inefficiencies in global health responses (e.g., pandemic AI applications). In contrast, our proposed framework introduces scalable, region-specific regulatory structures, enhancing global interoperability while respecting local governance needs. The proposed framework also bridges this gap through legally binding mechanisms, like penalties for noncompliance and participatory governance models, exemplified by initiatives like citizen juries in South Korea. 108
6.2.2. GDPR and Regional Frameworks
The GDPR, one of the most influential data protection frameworks, ensures transparent and secure data handling 109 but presents challenges for international AI governance due to its rigid data-sharing regulations. 110 This rigidity particularly affects low-resource nations where data localization policies hinder pandemic responses. 111 The proposed framework offers a globally adaptive model that integrates cross-border data-sharing protocols while respecting national sovereignty. Its modular approach enables jurisdictions to maintain their own data privacy regulations while aligning with core global principles, thereby enhancing regulatory interoperability. A notable example is the EU-Nigeria collaboration (EU and Global South Case Study, see also Akinloluwa and others), 112 where FHIR integration facilitated secure health data exchange, overcoming GDPR-related regulatory bottlenecks. Flexible and secure data governance, as enabled by this framework, supports critical public health initiatives, including disease surveillance and pandemic response, fostering a more resilient and cooperative global health ecosystem.
6.3. Highlighting Innovative Aspects: Adaptability, Inclusivity and Cross-Border Applicability
This framework distinguishes itself through its innovative approach, which integrates adaptability, inclusivity and cross-border applicability, closely aligning with the three core pillars of ethical accountability, regulatory adaptability and transparency. Unlike rigid models like the GDPR, 113 which impose “one-size-fits-all” solutions, this framework’s modular structure enables it to be tailored to the specific legal, cultural and public health contexts of individual countries. This flexibility addresses the diversity of the global landscape, ensuring that regulations remain relevant and responsive to evolving healthcare needs. The adaptability inherent in the framework directly supports the regulatory adaptability pillar, where it serves as a scalable solution to regulatory fragmentation. Our framework allows jurisdictions to retain sovereignty while aligning with global health objectives, fostering international collaboration without compromising local governance needs.
Inclusivity is another defining feature of this framework, ensuring that the needs of marginalized and vulnerable populations are prioritized. Through human-centric design principles, the framework guarantees that AI systems are accessible, culturally sensitive, and capable of reducing digital inequalities in healthcare. This aspect reinforces the ethical accountability pillar, as it institutionalizes participatory governance, where diverse stakeholders, including marginalized communities, have a voice in shaping AI systems. This participatory approach ensures that AI applications in public health are accountable to the very people they serve, addressing equity gaps that might otherwise be overlooked. In doing so, it helps to maintain the integrity of AI deployment, ensuring fairness and reducing disparities in healthcare access.
Lastly, our framework’s focus on cross-border applicability enhances its alignment with the transparency pillar. Unlike the GDPR, which limits cross-border data transfers, 114 the proposed framework facilitates secure and ethical international data-sharing. This capability is critical in addressing global health challenges, like pandemics, where real-time data exchange is vital for an effective response. The framework ensures that transparency extends beyond borders, enabling international cooperation in public health AI governance by embedding protocols that balance local regulations with global public health objectives. Furthermore, it fosters trust among stakeholders through clear guidelines on data privacy, security, and access, ensuring that AI applications remain auditable and publicly accessible, even across jurisdictions. The integration of these innovative features — adaptability, inclusivity, and cross-border applicability — offers a more comprehensive and flexible approach to AI governance. It not only strengthens the core pillars of ethical accountability, regulatory adaptability and transparency but also addresses the unique challenges posed by diverse cultural, legal and healthcare systems. This human-centric approach ensures that AI deployment in public health is responsible, equitable, and effective, fostering collaboration and building trust across borders.
7. Conclusion
The framework advances AI governance by operationalizing human-centric principles into actionable, legally robust mechanisms. By prioritizing ethical accountability, adaptability and transparency, it addresses gaps in the WHO’s AI initiatives, the GDPR and other existing models while fostering equitable, globally inclusive health outcomes. The framework’s ability to be tailored to diverse national contexts, while fostering international cooperation, positions it as a robust, globally relevant tool. Policymakers and practitioners can leverage its modular design to balance innovation with ethical imperatives, ensuring AI serves as a tool for public health equity.
While this study advances a robust framework for human-centric AI governance in public health, several limitations warrant consideration.
-
I. The regional focus of case studies on the European Union and Africa may constrain the generalizability of findings to the Asia-Pacific region, where distinct legal, cultural, and infrastructural contexts could influence governance dynamics. Therefore, future studies must expand empirical testing of the framework to Asia-Pacific contexts, like Southeast Asia and Oceania, where diverse regulatory environments (e.g., Singapore’s pro-innovation policies versus Indonesia’s data sovereignty laws) could refine its modular adaptability. Collaborative initiatives with regional bodies like the ASEAN Digital Health Task Force could yield insights into balancing cultural sensitivity with global standards. Moreover, establishing regional AI governance bodies (e.g., an African AI Ethics Council, or an ASEAN AI Working Group) could tailor policy guidance to regional needs while ensuring alignment with global principles.
-
II. While the framework proposes enforceable accountability measures (e.g., financial penalties for noncompliance), real-world implementation depends on political commitment and international cooperation, which remain variable across jurisdictions. Addressing these challenges necessitates further research and practical efforts, particularly in investigating the operational efficacy of transnational oversight bodies, like a WHO-UNESCO Joint AI Ethics Review Board, through pilot programs in politically diverse settings (e.g., Brazil, India). Metrics could include reductions in algorithmic bias incidents or improvements in equitable resource allocation, providing evidence to incentivize binding enforcement protocols.
-
III. The technical and financial feasibility of retrofitting legacy AI systems in low-resource settings with XAI standards poses a significant challenge, potentially exacerbating existing inequities in technological access. To address these limitations, future research should prioritize developing and evaluating open-source XAI toolkits and equitable funding models, like a Global XAI Fund supported by high-income nations and private-sector partnerships. Case studies in low-resource settings (e.g., Malawi’s TB diagnostic networks) could assess the scalability of retrofitting legacy systems with interpretability interfaces while maintaining affordability. Establishing global AI training programs for policymakers, regulators and healthcare professionals, potentially under the UNESCO AI Competence Framework, would help address skill shortages.
-
IV. As AI technologies continuously evolve, future research should anticipate and address new ethical challenges. For instance, the governance of generative AI in health diagnostics or the implications of autonomous decision-making systems in public health emergencies will require ongoing assessment and adaptation of the framework. By addressing these gaps, future work will strengthen the framework’s global relevance, enforceability, and inclusivity, ensuring that ethical AI governance evolves in tandem with technological and geopolitical realities.
Disclosure
The authors have no funding or conflicts of interest to disclose.
