Will the proposed investments in AI make us healthier — and at what cost?

A woman in her 30s arrived for a medical appointment a few months after delivering a healthy baby. She lay flat on an examination table as electrodes were placed across her chest. Within moments, an electrocardiogram, or ECG, traced the electrical activity of her heart into the familiar wavy line with spikes and valleys that indicate a heartbeat.

To most clinicians, the ECG probably would have appeared normal. But an AI-powered app reading the signals on her doctor’s smartphone flagged something concerning. Then, a second device — an AI-enabled digital stethoscope — also alerted.

Further tests confirmed that she had left ventricular systolic dysfunction, a potentially dangerous heart condition that can occur during or after pregnancy.

This wasn’t a demonstration or a hypothetical scenario. It really happened.  

Dr. Peter Noseworthy, a cardiologist at the Mayo Clinic and leader in AI-ECG research, points to cases like these as compelling examples of the technology’s potential.

“Those are the cases where AI may provide the greatest clinical value,” Noseworthy said in emails with Canada’s National Observer.

The word “may,” however, is important and can be difficult to parse. 

Two competing narratives are emerging about AI in health care: is it a technology poised to transform health care for all, or an industry expanding so rapidly that communities could be harmed? As Canada pours hundreds of millions of dollars into AI for health care, questions remain about how much of the promised AI health revolution is backed by evidence, what new issues it raises and how Canadians should weigh the potential benefits against the emerging costs.

You may not know it, but AI is already in use in health care

On June 4, 2026, Prime Minister Mark Carney stood in front of medical doctors in blue scrubs and white lab coats, many with stethoscopes slung around their necks, at Toronto General Hospital.

Carney was there to announce the launch of a national artificial intelligence strategy, AI for All, spotlighting $200 million for health care, with a goal to “take AI out of the lab and put it to work on real problems.”

Some of the strategy includes expanding the use of tools that make practical aspects of health care more efficient — from automating scheduling (sending appointment reminders) to improving doctor-patient interactions with AI-powered note-taking (called ‘AI scribes’) and review of diagnostic images and tests. 

Most people would likely see decreasing administrative burdens or increasing accuracy as good things. However, some doctors will point out that less paperwork or improved diagnostics does not automatically lead to improved health.

Does better health data lead to better health?

“Better health data can mean better health care,” said Evan Solomon, federal minister of Artificial Intelligence and Digital Innovation, as he announced an investment of $100 million in VITAL, a data health platform powered by AI.

“Every day, our hospitals generate information that could help researchers discover new treatments, improve services and build the next generation of Canadian health innovation,” he said. 

The promise of better health data is what originally drew Zahra Shakeri, an assistant professor at the University of Toronto, to the field of public health and medical AI. Shakeri wanted to use computer science to assist with triage and potentially help someone who might otherwise be overlooked.   

But as the plan to stitch together medical records from clinics and hospitals gets underway, Shakeri has concerns.

“Take a missing blood test, for example,” Shakeri said in an interview with Canada’s National Observer. A missing blood test could occur if the doctor didn’t think it was needed or the patient couldn’t get it done. Either way, the gap looks the same in a database. 

Health care data, Shakeri points out, is merely a reflection of a person’s access to care — the number of times the person interacts with medical officials and the quality (or lack thereof) of the human-to-human interaction. 

Issues of bias, privacy and inclusivity — problems that are already baked into the existing health care system — need careful consideration as AI models are trained for use in different demographics and regions.

According to Shakeri, a patient who gets an appointment quickly may leave a very different data trail than someone who waits months for care. It is unclear how a new AI-powered health data system will address those blind spots, which is a key consideration for those whose medical records are incomplete and for underserved demographics and communities.  

“What are we actually teaching AI to predict?” Shakeri asked.

Does the promise of accelerated drug development hold up?

After visiting ProteinQure — a Toronto-based company using AI to design cancer treatment drugs — on May 25, 2026, Solomon boasted about AI helping to “design new medicines” for cancer treatment.

“AI does not replace that human effort. It helps accelerate it,” Solomon wrote in his announcement. 

Traditionally, drug developers would consider a nearly unimaginable number of protein shapes and molecular combinations, like finding the precise key that fits a particular lock. AI tools can search that landscape of possibilities much faster than scientists testing candidate molecules one by one.

However, finding a promising molecule isn’t the same as developing a new medicine — it is only one small step in the process.  

For new drugs to emerge for treatments, researchers still must establish if a potential drug can reach the correct target cells, survive in the body, avoid causing toxicity to non-target cells and ultimately treat the problem and help patients’ health improve.

One bottleneck in the long, arduous process of drug development is clinical trials, which are stubbornly dependent on tests in animals and humans.

“Certain aspects of clinical trials are a rate-limiting step,” said Kelsey Hanson Satterly, founder of KH Satterly Consultancy, in an interview with Canada’s National Observer. 

“There are so many things you can’t really speed up,” she said.

Even with the correct combination of keys and locks, every new drug will still need toxicity studies, participants, ethics approval, patient recruitment, dosing, follow-up and enough time to discover whether the drug works.

Can a technology intended to improve health also impose health costs?

The AI tools currently in use or under evaluation for widespread adoption will ultimately depend on an expanding network of data centres.

Those data centres — the subject of many recent protests across Canada and the US — consume astonishingly large amounts of electricity and, when powered by fossil fuels, produce air pollution that can be problematic for surrounding communities. 

With AI infrastructure rapidly expanding — and since health risk assessments are not required for data centres — community groups and public health doctors are concerned about the potential environmental and public health costs.

The contrast between these two sides of the dilemma creates challenges for Canada’s AI ambitions. 

Should Canada embrace a technology that holds promise to improve human health but also requires infrastructure that can carry health costs of its own?

What kind of AI health system do Canadians want? 

For some doctors, the benefits of new AI tools — like reading ECGs of a new mom — outweigh the risks associated with AI development for health care.

Noseworthy appreciates the significant risks associated with AI but distinguishes the risks for AI technologies that are “purpose-built” for improving health.

“These are risks we can measure, study and mitigate,” said Noseworthy. “They should be considered in the context of the potential clinical benefit.”

For Dr. Douglas Manuel, a senior scientist at The Ottawa Hospital and professor at the University of Ottawa, the challenge of the moment is figuring out how to “maximize the pros and minimize the harms.”

“AI will affect every corner of the health care system and health,” he said. “The good parts can be where [AI] is not reducing our quality of care, but it’s improving how we deliver behind the scenes.”

According to Manuel, we can avoid introducing AI simply to save money and reduce costs, but instead dramatically improve access and care, “if we do this right.”

“We need to talk collectively about what we want,” he said. 

For clinicians, Manuel said, that can mean better quality of care and more time face-to-face with patients, with AI assisting rather than displacing health-care professionals.

Manuel describes himself as “more optimistic” about AI in health care. But, like Noseworthy, he acknowledges nervousness and fear about the well publicized risks of AI, including the potential for severe environmental, economic and societal consequences.

“Where I get scared is that we won’t have agency,” Manuel said. 

“AI is moving so fast,” he added. “Both the good and bad.”