Canada’s latest debate about digital sovereignty has moved from theory to urgency. The proximate trigger is not abstract regulation or a white paper from Ottawa. It is the abrupt reality that access to advanced AI systems can be shaped outside Canada’s control. Anthropic said on June 12 it had been ordered by the U.S. government to suspend access to its Fable 5 and Mythos 5 models for foreign nationals, a move that led to a wider shutdown for all customers while the company complied. That episode, whether temporary or not, exposed a strategic vulnerability for countries that depend on foreign AI infrastructure, foreign cloud providers, and foreign policy decisions for tools that are becoming central to economic life.
This is the backdrop for a new report from the Blockchain Research Institute, Rebuilding Canada for the New Technology Order, by Don Tapscott and Alex Tapscott. The report argues that Canada is facing two simultaneous upheavals: a fractured geopolitical environment and the rise of artificial intelligence as a general-purpose technology that is reshaping productivity, sovereignty, and state capacity. Its core claim is straightforward: Canada cannot assume that intelligence, compute, and digital infrastructure will remain abundant, neutral, and globally accessible. The country needs to design for resilience, not convenience.
That framing is sharper than the usual Canadian innovation rhetoric, and rightly so. For years, Canada has been comfortable celebrating strengths in AI research while underperforming on commercialization and scale. The BRI report identifies five priorities: sovereign digital infrastructure, stronger pathways to build globally competitive Canadian firms, AI-led productivity gains, defence modernization for algorithmic conflict, and a renewed social contract for work, identity, and democratic resilience. Taken together, these are less a shopping list than a warning that the old model—excellent universities, modest capital, and dependence on U.S. platforms no longer looks adequate.
Devising AI strategy
There is empirical support for the concern over productivity. Statistics Canada reported that 12.2 percent of Canadian firms used AI to produce goods or deliver services in 2025, double the previous year, with another 14.5 percent planning adoption over the next 12 months. Yet the same analysis places this against Canada’s longstanding productivity weakness, linked to weak business investment, sluggish R&D, and lagging intangible asset development. The Bank of Canada has also argued that AI could materially improve productivity, wages, and investment—but only if adoption spreads and complementary capabilities are built. Canada’s challenge is not inventing AI; it is embedding it across the economy.
This is where the report’s call for sovereign infrastructure deserves serious attention. Ottawa has already moved in this direction with the Canadian Sovereign AI Compute Strategy, which is built around public and commercial compute capacity, an AI Compute Challenge, and new supercomputing infrastructure. The federal government says the strategy is intended to keep Canadian data and intellectual property in Canadian hands while giving domestic researchers and firms access to the compute they need. That is not techno-nationalism for its own sake. It is recognition that compute has become a strategic resource, closer to energy or telecommunications than to ordinary software procurement.
Canadian firms leading the way
A real-world Canadian example already exists. In 2025, TELUS opened what it described as Canada’s first fully sovereign AI factory in Rimouski, Quebec: a facility controlled in Canada, designed to keep data within national borders, and aimed at supporting Canadian businesses, researchers, and public institutions with domestic AI compute. CBC later reported that sovereign data centres have become a focal point of Ottawa’s broader AI infrastructure push, with policy makers increasingly distinguishing between infrastructure that is merely located in Canada and infrastructure that is actually controlled here. That distinction matters. A server in Canada is not the same as sovereignty if the governance, commercial leverage, or legal exposure sits elsewhere.
Toronto-based Cohere offers one of the clearest signs of maturation. Focused on enterprise-grade large language models, the company has built a global customer base while pursuing a markedly different strategy from consumer-facing AI firms.
Infrastructure, however, is only half the story. Canada also needs firms that can convert research excellence into durable economic value. Here, Toronto-based Cohere is one of the clearest examples of what success can look like. CNBC reported in February that Cohere had reached roughly $240 million in annual recurring revenue, surpassed its growth target, and climbed to an estimated valuation of about $7 billion. That matters because it shows Canada can produce AI companies with global relevance in enterprise markets, not just academic prestige. The BRI report is correct to focus on commercialization pathways and capital allocation. Without them, Canada will continue to educate top talent only to see the largest gains captured elsewhere.
The defence pillar of the report may strike some readers as more speculative, but that would be a mistake. The same episode involving restricted access to frontier models shows that advanced AI is now entangled with national security, export controls, and cyber capability. Anthropic itself said the U.S. directive was linked to concerns about bypassing safeguards in a powerful model. Once intelligence becomes operational infrastructure, states do not treat it as a neutral consumer service. Canada therefore has to think beyond startup policy and ask harder questions about cyber resilience, autonomous systems, procurement, and the role of domestic technology in strategic sectors.
The social contract element is equally important, though it is often handled in vague terms. Canada does not merely need more AI; it needs institutions that widen the gains. Statistics Canada’s work shows that firms with cloud capability, data analytics, robotics, R&D, and ICT training are more likely to adopt AI effectively. That suggests the real dividing line will not simply be between firms that buy AI and those that do not. It will be between organizations with the managerial depth, worker training, and digital infrastructure to make AI productive, and those left behind as the technology diffuses. Canada’s policy response has to bridge that gap or risk deepening regional and sectoral inequality.
If a decision in Washington can disrupt access to advanced AI for Canadian users overnight, then digital dependence is no longer a theoretical concern. Canada has pieces of the answer already: sovereign compute planning, data-centre investment, and homegrown firms such as Cohere. What remains uncertain is whether these pieces will be assembled into a coherent national strategy before dependency hardens into permanent disadvantage.