September 28, 2025

Over the past year, I've had the privilege of addressing healthcare leaders, technologists, and policymakers at major conferences in Toronto, New York, Austin, Atlanta, Chicago, Doha, etc.  on one of the most pressing issues of our time: the ethical deployment of artificial intelligence in healthcare. These conversations have reinforced my conviction that while AI holds unprecedented potential to revolutionize healthcare delivery, we must proceed with deliberate care to ensure these technologies serve all communities equitably.

The Promise and the Peril

Artificial intelligence is already transforming healthcare in remarkable ways. From diagnostic imaging that can detect cancers earlier than human radiologists to predictive algorithms that identify sepsis before clinical symptoms appear, AI's capabilities continue to expand. The FDA has approved over 880 AI-enabled medical devices as of 2024, with the majority concentrated in radiology, cardiology, and neurology.However, my discussions with international healthcare leaders have made it clear that AI's greatest promise in improving health outcomes for all still remains at risk due to systemic biases and inequitable development practices. We're at a critical juncture where the decisions we make today about AI governance will determine whether these technologies amplify existing healthcare disparities or help eliminate them.

The Challenge of Algorithmic Bias

Through my work with Ask Me Your MD and our consulting practice, I've witnessed firsthand how AI systems can perpetuate or even exacerbate existing healthcare inequities. Consider the sobering reality that many AI algorithms require patients of color to present with more severe symptoms than white patients to receive equivalent diagnoses or treatments. This isn't a technical glitch, it's a reflection of historical biases embedded in the data we use to train these systems.

The sources of bias in healthcare AI are multifaceted and occur throughout the development lifecycle:

• Data Bias: Historical medical data reflects decades of healthcare disparities, underrepresentation of minority populations, and socioeconomic inequalities. When we train AI models on this data without correction, we digitize and scale these biases.
• Development Bias: AI development teams often lack diversity, leading to algorithms that don't account for the varied experiences and needs of different patient populations.
• Implementation Bias: Even well-designed algorithms can produce biased outcomes if deployed in healthcare systems that lack appropriate governance structures or user training.

My Recommendations: A Framework for Ethical AI Implementation

Based on my research and experience speaking at healthcare conferences internationally, I've recommend a comprehensive approach to ethical AI in healthcare that addresses five critical domains:

Community-Centered Design and Development

Healthcare AI must be developed with communities, not just for them. This means:

• Authentic Community Engagement: Involve patient communities, especially those from underrepresented groups, throughout the AI lifecycle, from problem definition to post-deployment monitoring.
• Compensated Participation: Provide fair compensation to community members who contribute their expertise and lived experiences to AI development.
• Cultural Competency: Ensure AI systems are designed with cultural sensitivity and can accommodate diverse health beliefs and practices.
• Language Accessibility: Develop multilingual capabilities that go beyond translation to include cultural nuance and health literacy considerations.

Inclusive Data Governance and Management

The foundation of fair AI lies in representative, high-quality data:

• Diversity Requirements: Mandate that training datasets include adequate representation across race, ethnicity, gender, age, socioeconomic status, and geographic location.
• Prospective Data Collection: Where historical data is biased, invest in prospective data collection that actively addresses representation gaps.
• Open Science Practices: Engage with initiatives like STANDING Together and other accessible data sharing platforms to enhance dataset diversity.
• Bias Assessment Tools: Implement systematic bias detection at every stage of data collection and preprocessing.

Transparent and Accountable Algorithm Development

AI systems in healthcare must be explainable and auditable:

• Explainable AI: Develop algorithms that can provide clear, clinically relevant explanations for their decisions.
• Multidisciplinary Teams: Include statisticians, ethicists, clinicians, and community representatives in development teams.
• Bias Mitigation Techniques: Implement technical approaches such as stratified batch sampling, fair meta-learning, and protected group models where appropriate.
• Continuous Monitoring: Establish "AI-vigilance" systems similar to pharmacovigilance to monitor algorithmic performance across different populations post-deployment.

Robust Governance and Regulatory Frameworks

We need governance structures that promote innovation while protecting patients:

• Ethics Review Boards: Establish AI ethics committees with diverse membership, including patient advocates and community representatives.
• Mandatory Impact Assessments: Require health equity impact assessments before AI deployment, with disaggregated outcomes reporting by demographic groups.
• Regulatory Harmonization: Work toward international standards that ensure AI systems meet ethical obligations regardless of their country of origin.
• Post-Market Surveillance: Implement mandatory post-release auditing by independent third parties, with public reporting of outcomes.

Education and Trust Building

The success of ethical AI depends on informed stakeholders:

• Provider Training: Educate healthcare workers on AI capabilities, limitations, and potential biases.
• Patient Education: Develop clear, accessible information about AI use in healthcare settings.
• Public Engagement: Foster transparent dialogue about AI's role in healthcare through community forums and educational initiatives.
• Trust Metrics: Develop and track indicators of public trust in healthcare AI systems.

The Global Imperative

My conversations with healthcare leaders in across the global have highlighted that ethical AI in healthcare isn't just an American challenge, it's a global imperative. Countries with emerging healthcare infrastructures have the opportunity to build ethical AI principles from the ground up, while established healthcare systems must work to retroffit existing systems with fairness considerations.

The World Health Organization's recent guidance on AI ethics provides a foundation, but implementation requires local adaptation and international cooperation. We need to share best practices across borders while respecting cultural differences in healthcare delivery and patient values.

The Path Forward

As we stand at this crossroads, the healthcare community has a choice. We can allow AI to perpetuate the inequities that have long plagued our healthcare systems, or we can use this moment to build something better. AI systems that actively promote health equity and serve all communities fairly.

This will require sustained commitment from multiple stakeholders:

• Technology Companies must prioritize fairness over speed to market
• Healthcare Systems must invest in the infrastructure needed for responsible AI deployment
• Policymakers must create regulatory frameworks that incentivize ethical practices
• Communities must remain engaged as active participants rather than passive recipients

The potential benefits are too significant to ignore:

AI that truly serves all communities could help address physician shortages in underserved areas, make high-quality diagnostics available globally, and personalize treatments in ways that account for genetic, cultural, and social diversity.

A Call For Action

The future of healthcare AI is not predetermined. The choices we make today about how we develop, deploy, and govern these technologies will determine whether AI becomes a tool for equity or inequality. Based on the conversations I've had with global healthcare leaders over the several months, I'm cautiously optimistic that we can choose the path toward equity, but only if we act with intentionality, transparency, and unwavering commitment to the communities we serve.

The technology exists to create fair, transparent, and beneficial AI systems. What we need now is the collective will to demand nothing less than ethical AI that serves everyone equally. Our patients, and our future depend on getting this right.

Christopher Kunney is Managing Partner at IOTECH CONSULTING and Chief Technology Officer at Ask Me Your MD. He frequently speaks on healthcare technology and digital equity at international conferences and consults with healthcare organizations on ethical AI implementation.