Ten years ago, AlphaGo, an artificial intelligence system built by Google DeepMind, defeated the world champion at Go, the ancient strategy board game whose simple rules hide a ridiculous number of possible moves.
Prime Minister Mark Carney brought up the match at today’s launch of Canada’s national AI strategy, “AI for All,” because AlphaGo didn’t win by copying centuries of human play. It made moves experts had never seen.
“Instead of destroying the fun of Go, they opened up a whole new area on the board,” Carney told reporters. “When I see examples like that, it shows the ability to do things that we didn’t think were possible, combined with the human element.”
Today’s release has been a long time coming. Canada launched the world’s first national AI strategy in 2017, renewed it in 2021, and has been operating without an updated framework ever since.
The previous government’s proposed AI legislation, the Artificial Intelligence and Data Act, died when Parliament was prorogued in January 2025. Solomon launched a 30-day public consultation in October 2025 and promised a new strategy by year’s end. It took until today, and more than 11,300 Canadians weighed in along the way.
Canada helped build artificial intelligence. Geoffrey Hinton, Yoshua Bengio, and Richard Sutton all did foundational work here. National AI institutes in Montreal, Toronto, and Edmonton are globally recognized, and the country ranks among the top venture capital markets in the world.
But building AI and using it are two very different things.
Only 12% of Canadian businesses used AI to produce goods or services between mid-2024 and mid-2025, as opposed to 26% in Germany and 18% in France, found Statistics Canada. Among small and medium-sized enterprises, Canadian adoption sits at roughly 8%, well behind Nordic leaders at 29-42%
Canada also ranks 44th of 47 countries on AI training and literacy, and 42nd on trust in AI systems, according to a KPMG-University of Melbourne global study.
Canada might have helped invent the board, but now it has to prove more Canadian organizations can play.
The newly announced federal strategy commits more than $2.3 billion across six pillars, covering everything from AI adoption and sovereign compute to literacy, safety, and global partnerships.
It’s a plan with lofty goals, including a target of moving AI adoption in Canadian businesses from 12% today to 60% by 2034.
Now Canadian organizations have to make AI adoption survive budgets, workflows, governance, and people with actual jobs.
Canada has an adoption problem and this strategy knows it
According to Statistics Canada, 78% of Canadian businesses not planning to adopt AI say the technology isn’t relevant to the goods or services they provide.
The government is treating this as a translation problem. Businesses don’t need another abstract speech about AI, they need to know what it does on Monday morning.
“People need to see what AI does for them, not what AI is,” Evan Solomon, minister of artificial intelligence and digital innovation, said at the announcement. “How it’s supporting doctors and nurses and improving services.”
The key commitments for technology leaders include $700 million in additional funding for the Compute Access Fund to give SMEs access to affordable sovereign compute, $500 million through the Regional AI Initiative to accelerate adoption and commercialization across the country, and a $500 million Canadian Tech Growth Fund to provide growth capital and potential government equity stakes in high-potential AI firms.
The first real test will come through the strategy’s AI Missions Program. Its initial $200 million mission focuses on health care, with projects aimed at reducing ER wait times, expanding access to primary care, and reducing administrative work for physicians.
The strategy points to CHARTWatch at St. Michael’s Hospital in Toronto as the kind of outcome the missions model is designed to replicate. The AI early warning system monitors patients in real time and a study published in the Canadian Medical Association Journal found it reduced unexpected ward deaths by 26%.
More projects like CHARTWatch are the goal, where AI is applied to a specific problem with results you can point to.
Canada has the research credibility, but more organizations that know where AI fits, who is responsible for it, and how to move it into daily operations without turning governance into a scavenger hunt is the next step
Building sovereign infrastructure is a long game with near-term urgency
“Prosperity and sovereignty in the age of AI belong to those nations that can build, adopt, and govern AI on their own terms,” Carney said in his announcement.
The strategy’s focus on sovereignty brings it closest to the decisions most organizations are already making about vendors, data, cloud infrastructure, and risk.
“Most of our data that’s used in AI goes across our border or is governed by privacy regimes of other countries,” Carney said.
He added that Canada would expand sovereign compute and cloud infrastructure so more AI systems and data can be processed under Canadian control and Canadian law.
The strategy follows a build-partner-buy framework borrowed deliberately from the defence industrial strategy. Build sovereign capabilities domestically first, and where that isn’t possible, partner with like-minded allies. Buy from abroad only after exhausting those options.
On the build side, the strategy commits to a world-leading public AI supercomputer, expanded sovereign compute and cloud infrastructure, and nearly $350 million to expand Canada’s three national AI institutes: Mila in Montreal, Vector Institute in Toronto, and the Alberta Machine Intelligence Institute (Amii) in Edmonton.
Together, the institutes anchor the Canada CIFAR AI Chairs program, which funds long-term research positions across nine universities, and is designed to bridge the distance between lab work and industry application.
Today’s strategy commits to expanding the program from its current 143 chairs to nearly 200.
“The next couple of years will shape the coming decades of AI and we can’t afford to wait,” Amii CEO Cam Linke said in a statement. “The investment in this strategy secures our leadership in the global narrative, ensuring Canada isn’t just reacting to the future of AI but actively shaping it.”
Canada is also one of only four countries in the world with a large language model (LLM) capability, Carney said, which the strategy treats as a strategic anchor for data governance and national security.
This is where the strategy gets closest to decisions already being made inside Canadian organizations. Vendor selection, cloud infrastructure, and data residency choices are now part of the sovereignty conversation the strategy is trying to shape.
There is a complication, though.
The U.S. Trade Representative’s 2026 National Trade Estimate Report flagged Canada’s Sovereign Cloud Initiative as a trade concern for the first time this year, treating it as a procurement barrier.
That gives Canada’s AI sovereignty push a trade dimension at the same time the country is heading into the CUSMA review without AI legislation in force.
The new strategy begins to sketch the necessary policy framework.
Companies are already making decisions on vendors, cloud infrastructure, data residency, and internal AI governance without a complete federal rulebook.
Three questions Canada’s strategy doesn’t answer yet
The strategy is clear on what Canada needs to build. Who’s responsible for making it real inside a Canadian organization is a little murkier.
The first question is the one most immediately in front of technology leaders. How do I govern AI inside my own organization right now?
The strategy covers consumer privacy, child safety, and government AI procurement in reasonable detail, but says almost nothing about the governance challenges those same leaders are managing internally.
A 2026 Deloitte study found only one in five companies has a mature governance model for overseeing how agentic AI is being used, even as worker access to AI rose 50% in 2025 alone.
Employees are using tools outside approved channels, sharing sensitive data with systems their organizations have no agreements with, and making decisions the audit trail can’t follow. Model risk is landing on audit agendas, and boards are asking accountability questions that don’t have standard frameworks yet.
The strategy’s answer to all of this is a single phrase, used once, in the context of public service delivery.
“Human in the loop.”
When organizations are navigating model risk, shadow AI, and board-level accountability questions without frameworks to point to, that’s not enough to work with.
The second is a timing question every compliance team is already asking. When does the regulatory framework arrive?
Privacy law modernization, online safety legislation, and the Canadian Trusted AI Certification program are all signalled, but none are dated.
Organizations building compliance roadmaps need more than a directional commitment. Without dates, they can’t decide what to build now versus what to wait on. The CUSMA review adds urgency, too. If the U.S. succeeds in framing sovereign cloud as a trade barrier, any AI legislation Canada introduces could become a negotiating chip before it even passes.
The third question is how do I keep the people I need?
The Canada CIFAR AI Chairs expansion strengthens the research pipeline, but doesn’t touch the 25-35% wage premium drawing Canadian AI talent to U.S. roles, and that premium is the reason most technology leaders have lost people they couldn’t afford to lose.
It’s a concern shared by the Council of Canadian Innovators (CCI).
“While ‘AI for All’ contains a number of promising ideas, it spreads its priorities broadly and does not yet provide a sufficiently clear roadmap for helping Canadian AI companies grow into globally competitive firms that create and retain economic value in Canada,” Laurent Carbonneau, CCI’s vice president of policy and advocacy, said in a statement today.
Carney closed his remarks with a line that applies as much to this strategy as it does to the technology it’s describing.
“Too often AI is talked about, including by those who are developing it and building the infrastructure to support it, as if AI were the end in itself,” he said. “Too seldom is it connected to people and for people.”
The “AI for All” strategy is at its strongest when it’s specific about people, like the farmer using soil mapping in Saskatchewan or the surgeon using imaging AI in Toronto.
It becomes harder to use when it returns to pillars, targets, and frameworks without naming who has to make them real inside an organization.
Getting from 12-60% requires that translation problem to be solved inside organizations, by the people deciding where AI belongs, who owns it, and how much risk they are willing to carry.
The strategy does give them more to work with than they had yesterday, but making the moves is a different question, and one the strategy can’t answer for them.
Final shots
- The national strategy puts money behind adoption, sovereign compute, AI literacy, safety, research institutes, growth capital, and global partnerships.
- The adoption target only matters if organizations can see where AI fits into daily operations.
- Sovereign compute matters because vendor, cloud, and data residency choices are becoming strategic decisions.
- Leaders still need clearer answers on accountability, approved use, risk, and oversight.