Making AI Ethics a Shared Language for Trusted and Equitable AI for Health
As AI capabilities advance rapidly, safety evaluation, regulation, and institutional governance are struggling to keep pace. Slowing the pace of AI development may buy time, but it does not answer the more fundamental question: how should that time be used to ensure that AI is developed and deployed responsibly? Recent global discussions highlight a shift from focusing solely on AI safety toward broader human-development concerns, including preserving human capabilities, reflecting diverse knowledge and cultural contexts, maintaining human accountability, enabling responsible data use, and ensuring that AI systems are verifiable and inclusive.
These challenges are particularly significant for AI for Health in low- and middle-income countries (LMICs). Models trained predominantly on data from higher-income settings may not perform reliably across different epidemiological and population contexts. At the same time, digital determinants of health, including connectivity, interoperable health information systems and digital literacy, influence who can benefit from AI. Limited capacity for local validation and post-deployment monitoring may further widen the gap between ethical principles and their practical implementation.
To bridge this principles-to-practice gap, the report proposes a shared risk-check framework aligned with the AI for Health lifecycle. Across ideation and R&D, data procurement and model training, clinical validation and regulatory clearance, and deployment and post-market surveillance would apply a common set of questions covering issues such as data representativeness, local validity, human oversight, accountability, model drift, accessibility, and health equity.
The objective is not to create another compliance checklist, but to establish a common operational language that enables developers, policymakers, regulators, procurers, funders, and other stakeholders to identify and address risks consistently. In this way, AI ethics can become an enabler of trusted, scalable and equitable AI for Health, rather than a brake on innovation, while keeping the needs of the most vulnerable populations at the center of technological progress.
Author
KPMG AZSA LLC
Director - Healthcare & Well-being (HC&WB)
Michikazu Koshiba
Connect with us
