Empowering teams with clarity, consistency, and lasting impact across sectors.

Healthcare AI has a governance Problem and DiMe is trying to solve it. Canada should pay attention.

Artificial intelligence in healthcare is rapidly becoming part of clinical decision making, operational planning, diagnostics, scheduling, documentation, and resource allocation. The question is no longer whether healthcare organizations will adopt AI, but whether they have the governance systems necessary to ensure those tools remain safe, transparent, accountable, and worthy of patient trust.

Last week, the Digital Medicine Society (DiMe), and American organization, announced a new initiative titled “Operationalizing AI Governance in Healthcare” bringing together health systems, technology developers, regulators, and industry experts to create the first open source toolkit designed to help organizations implement practical AI governance, not simply talk about it. Rather than creating another high level ethical framework, the initiative focuses on translating existing guidance into operational processes that healthcare organizations can actually deploy.

Despite the growing number of ethical AI frameworks, healthcare is not lacking in principles, it is lacking in implementation. Organizations worldwide have established guidance centred on fairness, transparency, explainability, privacy, bias mitigation, and accountability, yet many health systems still struggle to answer fundamental operational questions:

  • Who approves an AI tool before deployment?
  • How is model performance monitored over time?
  • What happens when an AI system begins producing unexpected results?
  • Who is ultimately accountable for AI-assisted clinical decisions?

According to research cited by DiMe, more than 80% of surveyed health systems in the US have little or no formal governance process for the AI tools they are already in use. That statistic should concern every healthcare executive south and north of the border.

The strength of DiMe's initiative is not that it introduces another governance framework. Instead, it recognizes that healthcare leaders are experiencing "framework fatigue". Most organizations already know what responsible AI looks like. What they lack are practical operating systems that answer:

  • What committees should review AI?
  • What documentation should vendors provide?
  • How should risks be assessed?
  • How should AI systems be monitored after implementation?
  • When should humans intervene?

These operational questions determine whether AI governance becomes embedded in daily practice, or whether it remains a policy sitting in a cloud file. The question is, does Canada have an equivalent? The answer is, not exactly.

Canada has several organizations advancing trustworthy AI. For example, the Vector Institute has become a national leader in responsible AI research and has published guidance for organizations implementing AI responsibly. Likewise, the Mila – Quebec Artificial Intelligence Institute continues to produce internationally recognized research in AI ethics and responsible machine learning. National organizations such as the Canadian Institute for Advanced Research (CIFAR)have also contributed significantly to Canada's leadership in AI policy and research.

However, these organizations primarily focus on research, education, policy development, and scientific advancement. What Canada currently lacks is a national collaborative specifically dedicated to operationalizing AI governance within healthcare delivery, bringing together hospitals, clinicians, vendors, regulators, and implementation experts to develop standardized governance processes that health systems can adopt consistently.

In this respect, DiMe represents something different. It focuses less on developing new AI principles and more on building the operational infrastructure within systems needed to govern AI safely across healthcare organizations. For example, Canadian healthcare systems are adopting generative AI for clinical documentation, diagnostic support, imaging, scheduling optimization, patient communication, and administrative automation, yet the level of AI maturity varies significantly.

Some systems have sophisticated governance committees, while others rely primarily on existing privacy or IT review processes that were never designed for continuously learning AI systems. Without standardized governance, every health authority risks creating its own approach and duplicating effort, introducing inconsistency, and increasing organizational risk along the way.

This is where Canada has an opportunity. Rather than waiting for regulation alone, healthcare organizations could collaborate to develop shared governance models, implementation checklists, monitoring frameworks, procurement standards, and human oversight mechanisms tailored to the Canadian healthcare environment. The emergence of initiatives like DiMe signals a broader shift as AI governance is evolving beyond compliance and becoming an organizational capability. Successful healthcare organizations will not distinguish themselves solely by adopting AI faster than others. They will differentiate themselves by governing AI more effectively by ensuring innovation remains transparent, accountable, measurable, and firmly under meaningful human oversight.

While technology improves healthcare, governance is what ensures it improves safely. As healthcare continues its digital transformation, perhaps the most valuable AI investment organizations can make is not another algorithm, but the governance system that ensures every algorithm deserves the trust placed in it.

CT