This week, Canada is supposed to reveal its new national AI strategy. The strategy's six pillars are: Protecting Canadians and democracy; Empowering Canadians; Powering AI adoption; Building sovereign AI foundation; Scaling Canadian champions; and Building Global Partnerships. These are important goals. However, most of the current discourse focuses only on three questions: Can we build it? Can we scale it? and can we deploy it?
Those are necessary questions, but they are not sufficient to make sure that Canada doesn't just build, scale, and deploy. Canada needs to build, scale, and deploy AI right. That requires us to answer harder questions, especially as AI becomes embedded in government services, banking, healthcare, education, and critical infrastructure. As Canada builds, scales, and deploys AI, it needs to ask: Can we trust it? Can we afford it? and can we control it?
Canada recognizes the importance of food security, energy security, and telecommunications independence. Artificial intelligence should be viewed through this lens. Real AI sovereignty is more nuanced and complex than what is covered by the average discussion. Real AI sovereignty requires:
- Hardware sovereignty: chips, compute infrastructure, and data centres
- Model sovereignty: trusted AI systems and foundational models
- Decision sovereignty: the ability of government and critical institutions to continue operating regardless of decisions made by foreign powers or foreign tech
Current policy discussions focus only on the first level, but any effective AI strategy will include all three in a meaningful way.
Canada has legitimate reasons to invest in AI infrastructure. Domestic compute capacity, data centres, research institutions, and to a lesser extent commercialization initiatives can strengthen economic competitiveness and reduce our dependence on foreign infrastructure. However, infrastructure is just a foundation and should not be framed as an end goal. AI policy should not be measured simply by the number of data centres built, GPUs deployed, or money doled out to private sector companies. The most important question Canada needs to answer, is whether these investments contribute to long-term national resilience, economic competitiveness, and public benefit for all.
Canada has excelled at AI research but struggled to retain our intellectual property and talent. Supporting Canadian firms, encouraging targeted adoption, and improving productivity are worthwhile goals. However, scaling AI must be connected to broader public good outcomes. No one should be trying to transform our country into a tech powered Canada Inc. Every decision regarding AI needs to consider who will benefit from AI-driven productivity gains, will economic benefits remain in Canada, and whether or not Canadian firms can compete against larger foreign firms. An AI strategy that only focuses on growth misses the mark by overlooking ownership, value capture, and long-term economic resilience.
Government, businesses, non-profits, workers, and citizens are being pushed into integrating AI into everyday operations. While AI adoption can improve efficiency, without being done strategically, this can increase dependency and risk. As AI becomes embedded in government, healthcare, banking, and critical infrastructure, policymakers must consider not just how quickly AI can be adopted, but also what happens when those systems fail, become unavailable, or become prohibitively expensive.
Canada's AI strategy needs a strong focus on AI safety and public trust. Issues such as bias, accuracy, transparency, accountability, privacy, and security are far from resolved. This is particularly important when AI is used in government services, military deployment, healthcare, financial systems, and regulatory processes. An AI system that produces inaccurate, discriminatory, opaque, unfair, or unaccountable outcomes is a failure that will undermine public confidence. Clear standards for transparency, human oversight, accountability, and redress mechanisms are essential.
Almost all the public discussions around AI assume that AI will increase productivity and reduce costs. The thing is, AI is not just software. AI is resource intensive infrastructure technology that needs massive investments in compute, electricity, cooling systems, cloud services, ongoing model access, and tokens per use. Every single AI query consumes these resources. Who has to pay for all this? Are Canadians supposed to foot the bill through our tax dollars on data centres and expanding the grid for private sector demands? Will scarce grid capacity prioritize households, hospitals, schools, and other essentials citizens depend on or prioritize private data centres? Will Canadians face higher electric bills as the demand for power grows?
The labour implications of AI adoption have not been honestly addressed. AI might improve productivity, but it is already displacing workers, changing jobs, and shifting economic gains to the small few who own tech companies and other capital rather than equally benefiting all Canadians. Any credible AI strategy must address the costs of AI rather than just the perceived benefits.
Current AI sovereignty discussions limit their scope to domestic compute capacity and data centres. However, even with domestic compute capacity and data centres, Canada would still be dependent on foreign controlled AI systems and supply chains. For example, advanced semiconductors and chips manufactured abroad, foreign cloud infrastructure, foreign-owned foundation models, foreign technology platforms. This is a brand-new form of strategic dependence and extreme vulnerability.
Unlike traditional supply chain disruptions like a tariff dispute affecting manufactured goods where challenges take effect over the course of weeks or months, AI systems that are dependent on foreign powers and private companies create crises immediately. Can we afford to inject AI into government, hospitals, banks, businesses, non-profits, and critical infrastructure without actual independence and sovereignty? There is no complete AI sovereignty for Canada unless we have domestic capability across the entire AI tech stack from chip design to compute infrastructure to sovereign AI models for public interest and critical sector applications. The objective here is not isolation, it is resilience.
So far, Canada's AI policy discussions have ignored whether or not we need a domestic model. This does not require building a frontier model that can compete directly with the world's largest technology companies. However, government and critical sectors like healthcare need access to a trusted domestic model that can continue operating regardless of geopolitical disputes, commercial disagreements, export restrictions, or changes in a foreign provider's priorities. A Canadian public-interest model can support the systems and services Canadians depend on the most. This is no different than Canada's strategic capacity in telecommunications, energy, and other essential sectors. Moreover, having a domestic public-interest models is not unprecedented, Switzerland built one that is reliable and effective.
Canada's AI strategy has 6 solid pillars. However, it needs more clarity and a broader scope to tackle national resilience; clear standards for government use including transparency and human oversight; a meaningful concept of AI sovereignty; a Canadian public-interest AI model; a full review of Canada's domestic AI supply chain; workforce transition measures; in-depth analysis on energy considerations; public procurement that supports Canadian companies, institutions, and talent; and contingency plans for disruptions due to foreign powers and foreign companies.
Canada can be a global leader in artificial intelligence, but this leadership needs more than just building partial infrastructure and adopting AI quickly. Canada needs to build AI systems that can be trusted, affordable, resilient, sovereign, and that benefit all Canadians. Any meaningful AI strategy has to be to honest, rational, and visionary about how artificial intelligence can shape the economy, institutions, and the daily lives of Canadians.