Artificial intelligence is already moving through healthcare, biomedical research, clinical decision-making, public health, and global health systems at a speed that makes hesitation less realistic than disciplined governance. At the People and Planet United Global Health and Purpose Summit, David Lazerson, Co-Founder and CEO of Briya, and Jonathan Samet, Professor in the Department of Environmental & Occupational Health and the Department of Epidemiology at the Colorado School of Public Health, join host Christina Raish of FINN Partners for a leadership conversation on Can AI Improve Global Health Without Trust, Equity, and Accountability?.

The session opens with a central premise. AI is no longer a future possibility waiting for permission. It is already shaping research workflows, regulatory expectations, clinical data extraction, real-world evidence, diagnostic support, population health analysis, and the future of healthcare delivery. Lazerson frames the shift directly, noting that the question is no longer whether AI should be used in healthcare and research. The more urgent question is how it will be used, who will be accountable for its impact, and whether the systems surrounding it can earn trust across very different healthcare environments.

The Shift Has Already Happened

Lazerson begins by describing an environment in which regulators, researchers, healthcare organizations, and technology companies are moving AI into health systems with accelerating momentum. He points to regulatory interest in extracting data from clinical records, using AI and real-world data in drug development, and the broader global vision for AI adoption in healthcare. The point is not that AI has solved healthcare. It is that the system has already crossed into a new operating reality.

That shift carries extraordinary promise. AI can help interpret diagnostic materials, structure handwritten medical records, support clinical decision-making, identify disease patterns, strengthen public health surveillance, accelerate research, and extend expertise into regions where specialist capacity is limited. It can help small teams analyze large bodies of data and allow smaller organizations to participate in research and development that once required much larger resources.

The same shift also brings significant risk. Lazerson points to irresponsible use, deliberate misuse, weak infrastructure, insufficient guardrails, and a lack of expertise as challenges already emerging. Samet brings the public health lens, asking what global health means in practice and which institutions, governments, professional bodies, and healthcare actors have the authority and responsibility to guide AI’s deployment across countries, populations, and care settings.

Accountability Across the Full Chain of Use

The strongest theme of the session is accountability. Lazerson raises the question of who should be responsible when something goes wrong, especially when large AI platforms are used in healthcare despite not being built specifically to assume clinical or research accountability. His comparison to automotive AI is instructive. Generic AI may be powerful, but health requires specialized tools, validated systems, and clear responsibility across the full chain of use.

Samet responds by widening the frame. Global health includes clinical care, public health, national health systems, international institutions, professional organizations, workers, children, and populations facing very different burdens of disease and access. Accountability cannot rest in one place because AI moves through a sequence of decisions, from development to procurement, implementation, adaptation, validation, clinical use, research use, and evaluation of impact.

That chain matters because AI tools can be developed in one environment and applied in another where clinical context, infrastructure, records, disease patterns, staffing, and patient needs may be entirely different. Samet gives the example of a clinical tool developed and tested in New York City that may not necessarily be applicable in Uganda. The issue is not whether AI can be useful in low- and middle-income settings. It may be especially useful there. The issue is whether tools are adapted, validated, monitored, and governed in ways that fit the local context.

Global Health Requires Local Fit

Samet’s Uganda example gives the discussion practical force. He describes a student who developed appendicitis during a field experience near a regional hospital. In an AI-enabled world, ultrasound imaging could potentially be read on site or transmitted elsewhere for rapid interpretation. Such a capability could be transformative in settings with limited specialist access. Yet the usefulness of the tool still depends on accuracy, context, quality control, and accountability.

AI can bring high-level expertise closer to the point of need, but global deployment cannot be treated as simple technology transfer. Tools built in high-resource settings may not automatically perform well in low-resource environments. Records may be handwritten, infrastructure may be inconsistent, training may be limited, and disease burden may differ substantially. The promise of AI in global health depends on the discipline with which tools are matched to the communities and systems they are meant to serve.

Samet’s framing moves the conversation away from abstract optimism. AI’s value must be measured by impact, including what good has been done, what risks have been introduced, who benefits, and who could be harmed when tools do not work as expected.

Researchers, Clinicians, and Data Quality

The conversation also turns to research integrity. Samet draws a clear line between using AI responsibly to improve clarity and using AI irresponsibly to generate analysis and scholarship without appropriate human accountability. His comparison to field research is useful. Decades ago, an investigator who sent a team to collect data remained responsible for quality control. Today, if an AI tool is used to crawl millions of clinical records and create an analytical dataset, the investigator still remains responsible for the validity of the information.

That principle matters for universities, research organizations, journals, peer reviewers, clinicians, regulators, and companies building AI-enabled research infrastructure. AI can accelerate discovery, but it cannot remove responsibility from those who use it. If a tool is faulty, the accountability chain moves back toward the developer and the implementation context. If the tool is used carelessly, responsibility rests with the user. Strong AI governance needs to recognize both.

Samet also points to the publication challenge. AI may appropriately improve language, clarity, and accessibility, especially for researchers who need support in preparing manuscripts. Using AI to analyze data, generate tables, write a paper, and then place a human name on the result is a different matter. The future of scientific publication will require stronger standards for disclosure, review, data provenance, and investigator responsibility.

Equity as a Design Requirement

The session’s equity discussion is among its most important leadership contributions. Lazerson raises the possibility of AI tumor boards that can support local physicians in places without access to top-level oncologist panels. That kind of application could help close expertise gaps and improve care in smaller or lower-resource settings. At the same time, AI can reinforce inequality when access, infrastructure, training, and representative data are uneven.

Samet is realistic about the risk. New technologies often reach those with more resources first. AI is no exception. The cost of tools, implementation, training, and correct use can widen gaps if global health leaders do not deliberately design against that outcome. A rural clinic or hospital in Sub-Saharan Africa may have very different needs from a large hospital in Manhattan, and the value of AI may be highest precisely where infrastructure is weakest.

Public health thinking becomes essential here. Lazerson suggests that AI equity begins with research and data. If training data reflects only major academic centers or high-resource populations, AI will reproduce those limitations. If data comes from many regions and population groups, and if the tools are designed with public health in mind, AI can move closer to reducing inequality rather than deepening it.

A Disease Burden Framework for AI Investment

Samet offers a thought experiment that becomes one of the session’s clearest frameworks. If given a billion dollars to use AI to improve health globally, he would focus on the places experiencing the greatest disease burden, examine the causes of that burden, and identify where AI could make a strategic difference. That approach would treat AI not as a technology looking for a market, but as a tool directed toward the health problems where it can produce the greatest benefit.

That framework is powerful because it links AI strategy to public health need. Instead of starting with the tool, leaders would start with the burden of disease, the affected populations, the available infrastructure, the likely benefit, and the path to implementation. Such an approach would help governments, funders, health systems, public health agencies, researchers, and companies decide where AI can create meaningful global health value.

AI can support interpretation of imaging, management of clinics, analysis of health records, surveillance, forecasting, and disease-specific interventions. Samet mentions the possibility of AI approaches for childhood pneumonia in Sub-Saharan Africa, including chest X-ray interpretation, physical sign assessment, and even video evaluation of respiratory distress. The examples show how AI could become more equitable when its development is shaped by the needs of populations that carry high disease burdens.

AI Should Not Replace Hard Public Health Work

Lazerson also raises a concern that AI could become another technological fix, drawing attention toward tools while distracting from deeper social determinants of health. Samet resists anthropomorphizing AI. The technology itself does not decide whether global health systems solve the right problems. Leaders, institutions, governments, researchers, and healthcare organizations decide how it is used.

Used well, AI can support public health, clinical care, population health, and global preparedness. Samet points to climate change and wildfires as examples of broader global health challenges where AI could potentially help with warning systems, forecasting, real-time predictions, and region-specific solutions. That perspective keeps the conversation grounded. AI is not the answer to every health challenge, but it can become an important tool when directed by public health priorities and responsible governance.

Five years ago, this conversation would have sounded very different. Five years from now, it will likely be different again. The challenge for leaders is to build enough governance, accountability, validation, and equity into the current moment so the next phase of AI in global health can produce trusted examples of success.

Democratizing Research and Development

Lazerson closes with one of the session’s most constructive points. AI can democratize access to research and development. Studies that once required large teams and the resources of major pharmaceutical companies may now become possible for small biotech companies, academic teams, and more focused research groups. With smaller teams able to do more, research may expand into niche areas and underserved topics that would otherwise remain neglected.

Samet agrees with the broader possibility. A future in which massive amounts of money are not always required, and in which accessibility and accountability are built together, would change who can participate in research and innovation. AI gives individuals and small teams the ability to work with large bodies of data in ways that were previously impossible.

That democratization will be valuable only if quality, accountability, and trust grow alongside it. More research is not automatically better research. More tools are not automatically better care. The value of AI in global health will depend on whether new capabilities are matched with rigorous standards, representative data, public health priorities, responsible implementation, and respect for the people and populations affected by the decisions.

Trust as the Operating Condition

The session’s title asks whether AI can improve global health without trust, equity, and accountability. AI can accelerate health progress, but only if the surrounding systems are strong enough to guide it.

Trust requires transparency, validation, explainability where possible, and confidence that tools are fit for purpose. Equity requires representative data, infrastructure support, training, access, and development priorities shaped by real disease burden. Accountability requires a chain of responsibility across developers, implementers, clinicians, researchers, institutions, regulators, and public health authorities.

The future of AI in global health will not be defined by technology alone. It will be shaped by the choices leaders make now about governance, evidence, inclusion, implementation, and responsibility. AI can improve global health when it is designed not only to scale intelligence, but to strengthen trust, reduce gaps, protect patients and populations, and support the public health mission of improving lives across every setting where care is delivered.

Session Intelligence

AI in global health is moving from possibility to implementation. The central leadership challenge is to ensure that adoption advances with accountability, equity, trust, and public health purpose.

Core Leadership Insight

AI’s value in global health depends on accountability built into every step, from model development to clinical and public-health use.

Strategic Relevance

The strongest use cases begin with disease burden and local need, then apply AI to diagnostics, records, research, surveillance, and care delivery.

Execution Model

Responsible adoption requires representative data, context-specific validation, user training, quality control, and clear ownership of outcomes.

Global Health Impact

AI can reduce gaps only when access, infrastructure, and governance reach the communities most at risk of being left behind.

AI in Global Health Trust Equity Accountability Public Health Clinical AI Research Integrity Representative Data Disease Burden Health Systems Global Health Equity Real-World Data