Canada’s AI Council Faces Its Sovereignty Test

Canada has assembled an unusually strong group of AI researchers, business leaders, investors and public officials. The harder task begins now: turning that expertise into measurable control over how AI is built, deployed and governed in Canada.

Prime Minister Mark Carney has launched the Prime Minister’s National Council on Artificial Intelligence to advise the federal government on Canada’s AI transformation and the implementation of its AI for All strategy. The council will examine adoption, investment, infrastructure, safety, democratic protections and the development of Canadian AI companies.

That mandate places the council at the center of a major national policy experiment. The government has set ambitious economic targets, including an additional $200 billion in economic growth and 250,000 new AI-related jobs over five years. It also wants AI adoption to rise from just over 12 percent to 60 percent by 2034, while creating up to 90,000 AI-related jobs and work placements for young Canadians.

Those figures give the council a clear test. Advice will matter only when it helps departments make decisions, directs investment toward lasting capabilities and gives Canadians a way to judge progress. Expertise creates a strong starting point. Implementation will determine the result.

The council’s responsibilities cover several connected parts of Canada’s AI agenda:

The combination matters. A country can have excellent researchers and still depend heavily on foreign computing capacity, private platforms and imported technical standards. It can also deploy systems quickly while leaving public agencies unsure about procurement, accountability and risk management. The council has been asked to address those issues together.

The phrase “sovereign AI infrastructure” will carry particular weight. In practical terms, it points toward Canada’s ability to access the computing power, talent, data systems and technical expertise required to develop and operate important AI applications. It also raises questions about who controls critical systems during a commercial dispute, a supply disruption or a national emergency.

A useful council agenda will connect these high-level concerns to specific decisions. It could examine which capabilities require public investment, which projects should be developed through partnerships and which systems demand stronger Canadian oversight. It could also identify areas where domestic control carries a clear public benefit and areas where international cooperation offers better value.

Canada’s new National Council on Artificial Intelligence brings researchers, business leaders, investors and public officials together to advise on adoption, investment, infrastructure, safety and democratic protection. Its central challenge is converting ambitious AI targets into durable national capabilities, measurable outcomes and accountable government decisions. The council’s credibility will depend on whether its recommendations lead to practical procurement rules, workforce programs, infrastructure choices and public reporting.

Canada’s AI sovereignty will take shape through ordinary institutional choices. Federal departments will decide which systems to purchase. Businesses will choose whether to adopt AI tools and where to locate research and production. Universities will train workers and conduct foundational research. Infrastructure providers will determine how much computing capacity is available to Canadian users.

The council’s advice should therefore reach beyond broad calls for innovation. It should help answer practical questions such as:

A Canadian company can build valuable products on foreign infrastructure, and that may be efficient for many uses. The strategic question concerns the systems and capabilities on which Canada cannot afford to lose access or influence. The council’s credibility will depend on distinguishing those priorities from every project that happens to carry an AI label.

The same discipline applies to national champions. Supporting Canadian companies can strengthen talent, investment and export capacity, yet public support needs clear conditions and measurable outcomes. The government will need to know whether an investment builds durable research capacity, expands access to infrastructure, creates quality jobs or simply shifts spending toward firms that would have grown anyway.

The AI for All strategy sets a scale for the council’s work. The government wants $200 billion in additional economic growth, 250,000 new AI-related jobs and a fivefold increase in adoption by 2034. These goals provide political momentum, yet they also create a measurement burden.

The adoption target needs careful definition. A business using an automated customer service tool, a hospital using an AI diagnostic system and a department using a language model represent very different levels of capability and risk. A simple percentage could count all of them as adoption while hiding differences in productivity, reliability and public value.

The jobs target raises similar questions. It should distinguish between temporary placements, specialized research positions, technical trades connected to infrastructure and long-term roles created through higher productivity. It should also track where the jobs appear, which workers can access them and whether training programs lead to employment.

A public scorecard could track several categories:

The council does not need to manage every indicator. It does need to recommend definitions that allow Parliament, businesses and the public to see whether policy is producing durable capacity. Clear measurement also gives the government a way to revise targets when conditions change.

Canada’s AI sovereignty will be decided in procurement offices and infrastructure contracts, not in the council’s meeting room.

The council can clarify priorities, but departments and ministers will determine whether Canada controls critical capabilities, builds durable firms and protects people affected by AI systems.

The council includes a wide range of experience. Its members include AI researchers Yoshua Bengio and Sanja Fidler, economist and investor Ajay Agrawal, entrepreneur Jordan Jacobs, technology and consulting executive Eric Lamarre, former public officials Jody Thomas and Senator Thomas Pitfield, former banker Louis Vachon, former Vector Institute leader Valérie Pisano and technology executives Michelle Zatlyn and Patrick Pichette, who will serve as an ex officio member.

Gillian K. Hadfield brings a particularly relevant perspective on the relationship between technical systems, law and public values. Her work focuses on AI governance, legal and regulatory design, computational models of human norms and the development of systems that can respond to human values. She is a law professor and AI governance scholar at Johns Hopkins University, a faculty member of the Vector Institute and a Schmidt Sciences AI2050 senior fellow.

That range of expertise can help the council see problems from several angles. Researchers may identify technical limits. Entrepreneurs can describe barriers to scaling. Investors can assess capital requirements. Public servants can explain how rules work inside government. Governance specialists can test whether a proposed system fits legal and democratic expectations.

The structure still needs a clear route from discussion to action. Council recommendations should identify the department responsible, the decision required, the resources involved, the expected result and the timetable for review. Those details turn expert advice into something a minister can approve, fund and defend.

AI safety and democratic protection are central parts of the council’s mandate. Those terms can become vague unless they are tied to specific uses and harms.

For public agencies, the council could examine how systems are tested before deployment, how records of automated decisions are preserved and how affected people can seek review. It could consider standards for human oversight, cybersecurity, privacy and the handling of errors. It could also study how agencies disclose their use of AI to employees and the public.

Democratic safeguards will require attention to information quality, political manipulation and the use of AI in civic institutions. The council can help identify risks, yet federal departments and ministers will still have to set enforceable rules and accept responsibility for them. An advisory body can improve the quality of decisions; it cannot replace legal authority or public accountability.

Hadfield’s background gives the council a useful bridge between technical design and institutional rules. That bridge will be valuable when the government faces choices where speed, privacy, transparency and security pull in different directions.

A small team within the Privy Council Office will support the council. Responsible ministers and federal departments will remain accountable for policy decisions and implementation. That division is important because it places responsibility where Canadians can see it: with elected officials and the public institutions authorized to act.

The arrangement also creates a risk of diffusion. A council can issue recommendations, departments can study them and ministers can cite complexity when decisions stall. The government will need a process that records which advice it accepts, which advice it rejects and why.

Regular public reporting would help. Reports could describe the council’s recommendations, government responses, spending commitments, delivery milestones and unresolved risks. Sensitive information may require protection, especially around national security and commercial data, yet the basic direction of policy should remain visible.

The council’s independence will be tested through that process. Independent advice has value when members can identify weaknesses in government plans and when the government responds with reasons. A panel that only confirms existing announcements will add little to public decision-making.

The council’s early work should focus on a small number of decisions with visible consequences. A national inventory of critical AI infrastructure could clarify existing capacity and gaps. A federal procurement framework could establish expectations for testing, data governance, security and supplier accountability. A workforce plan could connect training funds to regional employer demand and the government’s target for young Canadians.

The council could also recommend a common risk framework for federal AI systems. Different departments may face different risks, yet consistent terms for testing, documentation, human review and incident reporting would make oversight easier. Such a framework could give departments room to innovate while preserving a shared baseline.

The strongest early recommendations will probably be specific enough to fund and review. They will identify who must act, what success looks like and when the public should expect evidence. Broad ambition has already been supplied by the AI for All strategy.

Canada now has a national AI council with the expertise to discuss research, investment, infrastructure, safety and governance at the same table. That is a meaningful institutional step. The council’s real value will emerge through the choices it helps the government make.

Sovereignty will be visible in the country’s ability to maintain access to critical capabilities, support competitive Canadian firms, train workers, govern high-impact systems and protect public trust. The government has placed large numbers beside those aims. The council’s job is to connect those numbers to capacity that lasts.

Carney has said the council will advise on deploying AI, managing its risks and using the technology to improve Canadians’ lives. The next stage will show whether those words become a working system of investment, oversight and accountability.