Artificial Intelligence: Both Contagion and Cure - Why the Equity Stakes for Digital Health Have Never Been Higher

AI represents both promise and peril in healthcare - requiring careful thought about implementation

The Technology That Heals and Harms Simultaneously

Artificial Intelligence (AI) has arrived in healthcare with a promise and a warning. Unlike previous medical innovations that simply did one thing well or poorly, AI represents something unprecedented: a technology that can simultaneously amplify both healing and harm, often through the exact same mechanisms. This duality isn't academic—it's playing out right now in hospitals, clinics, and community health centres worldwide. The same machine learning algorithms that can detect cancer earlier in well-resourced populations may miss it entirely in others. The clinical decision support systems that enhance care for some patients can systematically disadvantage those whose data wasn't included in training sets. The digital health assistants breaking down geographic disparities between urban and rural populations may be inaccessible to many communities lacking reliable internet. In all these examples, the promise of AI to heal quickly turns to harm implementation isn’t carefully and critically thought through.

The Equity Paradox

Here's what makes AI's health revolution different from previous technological disruptions: it's the first innovation where the mechanism of harm and its defence are inherently intertwined. AI is both the problem and the solution to its own equity challenges. Consider algorithmic bias in diagnostic imaging. Traditional approaches might address this through better training or policy guidelines. But with AI, we can actually build bias detection into the algorithms themselves—using AI to monitor AI for fairness in real-time. The same computational power that can perpetuate health disparities can also identify and correct them, if we design it thoughtfully. This creates an unprecedented opportunity: for the first time in medical history, we have technology that can actively self-monitor and self-correct for equity issues. But it also creates unprecedented risk—because the scale and speed at which AI can operationalize bias far exceeds anything we've seen before.

The Implementation Reality Check

The transformative promise is real. Machine learning algorithms are already identifying diseases earlier and more accurately than human clinicians in many contexts. Natural language processing is breaking down language barriers in clinical care. Predictive analytics are helping resource-constrained health systems allocate care more efficiently. But the emergent risks are equally real. Algorithms trained primarily on data from wealthy, urban populations are making decisions about rural and marginalized communities they've never "seen." Clinical decision support systems are being implemented without consideration of how they might perform differently across demographic groups. Digital health tools are being deployed without ensuring the infrastructure exists for equitable access.

The Historical Pattern We Can't Ignore

This isn't the first time the health sector has faced the equity implications of transformative technology. From the early computers of the 1960s that served only elite academic medical centres, to electronic health records that initially widened gaps between well-resourced and safety-net providers, we have decades of evidence about how health innovations can inadvertently amplify existing disparities. What's different now is our awareness of these patterns and our ability to actively counteract them using the technology itself. We don't have to wait for post-implementation studies to identify bias—we can build equity monitoring directly into AI systems from the ground up.

The Question That Defines Our Future

The fundamental question isn't "Will AI disrupt health systems?”—that disruption is already underway. The question is: "How do we deliberately shape AI's trajectory to maximize healing while vigilantly guarding against harm, especially for those most vulnerable to being left behind?" This requires frameworks that embed equity considerations at every stage: from dataset curation that includes diverse populations, to model development that accounts for different contexts, to implementation strategies that ensure access across different communities. It means engaging the voices of those most at risk of exclusion, not as an afterthought but as essential partners in design and deployment.

Both Vigilant and Optimistic

Understanding AI as both contagion and cure means holding two truths simultaneously: this technology has unprecedented potential to reduce health disparities AND unprecedented potential to amplify them. Our response must be equally dual—embracing innovation while remaining vigilant about its equity implications. The communities most likely to benefit from AI's healing potential are often the same ones most vulnerable to its harmful effects. Rural populations who could gain access to specialist expertise through AI-enabled telemedicine may also be excluded by digital infrastructure gaps. Low-income patients who could benefit from AI-enhanced diagnostic tools may be served by under-resourced clinics that can't afford this implementation. This is why equity can't be an add-on to AI development—it must be foundational, built directly into systems at every stage from formative research to implementation and evaluation. The decisions we make today about how to design, deploy, and govern AI in health will determine whether this technology becomes a force for justice or a sophisticated mechanism for perpetuating and further entrenching deeply-rooted health disparities. The choice is ours, but the window for thoughtful action is narrowing. As AI systems become more embedded in health service delivery, the patterns we establish now will become increasingly difficult to change. This is our moment to ensure that when we look back on the AI revolution in health, we can say it was a lot more cure than it was contagion.