Canada should lead AI health alliance

Health artificial intelligence will likely shape the next generation of medical advances, influencing everything from cancer identification to pandemic response. For Canada, this creates a difficult choice between competing directly with the United States or China or finding a different path forward.

Canada’s unique position in health data

Canada holds a valuable asset in its health data, a byproduct of a single-payer system and decades of immigration. The country’s population of 42 million is one of the most diverse on Earth, offering a dataset that existing AI systems trained on smaller or more homogeneous groups cannot match. This diversity means Canadian data sets can help reduce bias and make algorithms more globally relevant.

While the government’s “AI for All” strategy marks a step forward, acting alone will not be enough. The center of gravity in AI development is consolidating among a small number of American and Chinese technology giants with unmatched computing power and talent. Furthermore, much of the world’s digital infrastructure is controlled by U.S. companies subject to the CLOUD Act, which can compel them to provide access to data to Washington.

This creates a specific vulnerability. Access to cutting-edge AI algorithms can be summarily throttled; the U.S. government’s decision to suspend access to Fable 5/Mythos 5 models to all non-American entities illustrates the risk. As AI becomes embedded in routine health care, this represents a growing dependency that could compromise national security.

European countries have responded with aggressive data-sovereignty laws and steps such as banning American platforms. However, this fragmentation comes at a cost. Isolated national data sets limit the scale and diversity needed to develop AI systems that perform fairly and reliably across different populations. A fragmented approach to development hinders the ability of researchers to train models that work well for everyone.

Building a coalition of middle powers

Canada should lead the creation of an ecosystem among middle powers for health AI. This approach would allow nations to pool their strengths and share the enormous financial costs of development while preserving national control over data and algorithms. Emerging common data standards and “federated analyses” mean data do not need to leave a country to contribute to AI development.

Algorithms can be improved through the diversity and size of data sets spanning multiple countries, while the underlying health data remain secure within each country’s borders. Such a consortium could mitigate dependence on either the American or Chinese ecosystems. The global health AI market is projected to exceed US$1-trillion within the next decade, making this a matter of significant financial and strategic consequence.

International collaboration is also essential for better patient care. More diverse multinational data sets can reduce AI bias and improve fairness. Shared AI systems could also have far-reaching consequences for the health system, from improving medical triage to strengthening pandemic preparedness. In an era of worsening political division, health AI collaboration may prove politically valuable and sustainable because few issues command broader public support than improving health care.

Canada has historically exerted influence through coalition building and institution creation. This is precisely the kind of moment where that identity matters. Fahad Razak is Canada Research Chair in Healthcare Data and Analytics at the University of Toronto, and Amol Verma is Temerty Professor of AI in Medicine at the University of Toronto; they are internists at St. Michael’s Hospital, and co-lead the VITAL health-data initiative, which received funding from Canada’s AI strategy and provincial sources. Kumanan Wilson is professor of medicine at the University of Ottawa and an internist at Bruyère Hospital.