Canada has released its national artificial intelligence strategy, titled AI for All, launched on June 4 by Prime Minister Mark Carney alongside Evan Solomon, the country's first Minister of Artificial Intelligence. The strategy rests on three pillars, Trust, Opportunity, and Sovereignty, tied together by a single claim: trust is what makes adoption possible, and opportunity and sovereignty are what make adoption pay off for Canadians. Its stated targets are ambitious, an additional 200 billion dollars in economic growth, 250,000 new jobs over five years, and a rise in business AI adoption from just over 12 percent today to 60 percent by 2034.
On trust, Canada calls the pillar its north star but reaches for soft instruments. The country's one attempt at a binding law, the Artificial Intelligence and Data Act, died in Parliament at the start of 2025 and is not being revived. In its place the strategy leans on modernized privacy legislation, a voluntary Canada Trusted AI Certification, and a safety institute funded at 50 million dollars. Critics have been pointed: one analysis called the approach a framework without a rulebook, and another noted that the phrase human rights does not appear in the strategy even once. The instinct, that trust enables adoption rather than slowing it, is sound. The instrument, voluntary rather than enforceable, is where the doubts gather.
On opportunity, the plan hangs on one number. Lifting business adoption from 12 percent to 60 percent in under a decade is a roughly five-fold jump, and it is the assumption every other figure depends on, since the jobs and the 200 billion dollars only arrive if that curve bends. A 200 million dollar program of AI Missions, beginning in health, is meant to pull adoption forward. The sharper worry is about consent: critics argue the strategy treats refusal as not an option by pushing AI into schools, services, and health care, and several noted the awkward optics of the government using AI to help review the 11,000 public comments submitted during consultation. Literacy, they argue, is not the same as agreement.
Sovereignty is the pillar with the most weight behind it. A Compute Access Fund now totaling around 1 billion dollars, a public supercomputer promised for 2031, and a plan to scale domestic sovereign compute to 850 megawatts by 2030 with a path beyond 2 gigawatts give this section real dollars and real deadlines. The posture is build, partner, buy, and it treats data as a strategic national asset. The contradiction is open, though: sovereign data centers that run mostly American hardware and software, copying the energy-hungry hyperscale model, with one flagship facility exempted from environmental assessment and partly powered by natural gas in a country that spent the summer fighting wildfires. The unresolved question is whether sovereignty means a count of megawatts or the capacity to run smaller, purpose-built models on infrastructure a country actually controls.
Read whole, AI for All is a genuinely considered document that knows its own size. Canada looked at the race between the United States and China and declined to pretend it could win the frontier-model arms race, choosing instead to compete on trust, sovereign and applied strength, and coalitions with like-minded nations. It names the energy problem out loud, which most strategies dodge, and writes the French language, Indigenous leadership, and an explicit equity lens into the foundation rather than the footnotes. The risk was never the vision. It is the distance between a framework and a rulebook, and the experts who study this are unusually united that the rulebook is the missing half. Sarah Chen, who covers AI for Zubnet, published a longer analysis of the three pillars on her blog at sarahchen.ai, landing on the view that the more durable sovereignty is the small, owned, well-understood model rather than the largest possible machine.