Six key takeaways from Canada’s AI strategy

Shingai Manjengwa is a data scientist, educator, and entrepreneur who has taught data science and AI to more than 500,000 learners worldwide. She leads AI education at Mila, the Montreal-based artificial intelligence research institute. Mila is one of the national AI institutes funded by the strategy; the standards she calls for apply to her own work. The opinions are her own.

Canada’s new artificial intelligence strategy, AI for All, outlines the government’s plan to drive widespread adoption, while also addressing increasing concerns around privacy and safety. As Minister of Artificial Intelligence and Digital Innovation, Evan Solomon, recently told PPF on the WONK podcast, the strategy aims for a pragmatic approach. “We’ve got to be open to the massive opportunities here, but we’ve got to be candid about concerns.”

The strategy revolves around six pillars and although AI is mathematics, the strategy itself is written in plain enough language, requiring little explanation. It presents a rough outline of how to reach best-case scenarios around AI adoption in Canada. But as the strategy gets rolled out, some areas will require special attention and raise important questions about the risks and potential impact of AI.

Safety, safety, safety

Pillar one is, appropriately, focused on “protecting Canadians and safeguarding our democracy.” The strategy commits $50 million to expand the Canadian AI Safety Institute, inside a package of more than $2.3 billion. Roughly 2 percent. That number should give us pause. The private sector has clear incentives to build most elements of an AI plan. There is money in compute, data centres, startups, even training. But in a capitalism-fuelled tech race, safety is the public good, the one thing only government may step up to fund.

The AI safety conversation is inconvenient, but it cannot be a footnote to an otherwise exciting economic strategy. If we lose control of AI agents with access to our critical infrastructure, nothing else in the strategy will matter.

Pillar one covers deepfakes and disinformation, but there’s much more to consider. Researchers have documented troubling behaviour in AI like deception and collusion. Frontier labs now warn of recursive self-improvement, AI that can build its own successors. And AI agents can now replicate themselves. Just two days before the strategy launch, University of Toronto and Vector Institute researchers published a proof-of-concept worm that uses a free, open-weight AI model to study each target, write a tailored attack on the fly, and replicate. It compromised nearly three quarters of their test network in a week, exploiting vulnerabilities published after the model’s training data ended.

MORE FROM PPF: On PPF’s WONK podcast, host Amanda Lang talks to Minister of AI and Digital Innovation Evan Solomon about driving adoption, building safeguards and working with the U.S. 

The good news is Canada has top researchers. Let’s fund them like we mean it. Serious money for safety research, creating a pipeline of Canadians as our front line of defence. Aviation, health care and energy taught us that we do not govern critical systems with optional principles and performative checklists. We govern them with deep research, Canadian experts, regulation, engineering, procurement discipline, accountability, and international cooperation.

Our allies are already moving. Germany announced a national AI Security Institute this week, in direct response to frontier AI cyber risk. The UK launched its institute with £100 million, roughly $170 million Canadian, more than three times what this strategy commits.

Socialize AI like we socialized electricity

Education and training are central parts of the strategy’s second pillar, looking at “empowering Canadians.” Just as society once had to socialize electricity — teaching safety, how it works, and opportunity, from the classroom to the factory — Canada must now socialize artificial intelligence. We must do so for the benefit of all, to the point where children may one day use it as easily and safely as turning on a light switch.

This means embedding AI education across all sectors and stages of life, ensuring that every citizen has the awareness, confidence and skills to live and work responsibly with intelligent systems.

READ MORE: PPF President and CEO Inez Jabalpurwala and Andrew Nevin, director of Brainomics Venture at the Center for BrainHealth, on how humans can thrive in an AI world

The strategy’s K to 12 commitment is to double teacher training to more than 3,000 educators. Canada has more than 420,000 educators in its public elementary and secondary schools alone. That is less than 1 percent of our single most important multiplier. One well-trained teacher reaches 30 students a year for a career. To those in the field, 3,000 reads like an error. Digital Moment, a single Canadian charity, has already trained more than 34,000 educators in digital skills and AI. If one nonprofit can deliver 10 times the national target, the national target needs another zero.

Up-skilling a population takes structured programs across six groups:

•       K to 12 learners and communities: early literacy and pipelines into machine learning and cybersecurity, with support for school communities, including educators.
•       Colleges, universities, and the trades: AI across disciplines, not quarantined in computer science.
•       Public sector professionals: structured productivity training, because adoption is also a governance question.
•       Private sector professionals: sector-specific programs, especially for small and medium-sized businesses, and the organizations serving our communities.
•       Special interest groups: advancing inclusion for Indigenous Peoples, Black and racialized communities, people with disabilities, newcomers, and rural and Northern communities, designed with communities, not for them.
•       General public and lifelong learning: supporting all citizens, including seniors, in understanding AI’s opportunities and risks, critical for democratic participation and digital resilience.

We will need targets for each group, and roadmaps with dates.

Adoption and job losses are the same bullet

Pillar three looks at empowering AI adoption in Canada, where as of 2025 only 12 percent of businesses reported using AI to produce goods and services. The strategy outlines a goal to increase adoption to 60 percent by 2034, along with the creation of up to 250,000 new jobs by 2031. A fivefold adoption increase is also, mechanically, a labour market event.

The AI strategy is quietest on something Canadians are deeply worried about. In the next year, there may be 10,000 or 10 million job losses attributable to AI. Amid conflicting reports, we must prepare for a range of scenarios.

The ideal policy funds credentialed pathways from displaced workers into Canada’s highest-stakes priorities, all starved for people. The care economy leads the list. These are sectors where we may choose humans over AI solutions, such as looking after our elderly, our youngest, and our sick. AI can draft the care plan; a person should deliver the care. An aging population guarantees the demand. Also on the list, climate work, the skilled trades for housing and critical infrastructure (including the data centres the strategy wants built), food security, education, and public services.

AI job displacement policy is also immigration policy, housing policy, food security policy, and the protection of our most vulnerable, all at play at once. A visibly weak job market, like the one AI could create, pulls on all of them. That is why the strategy should also include AI adoption in the nonprofit sector itself. The same tools driving productivity in industry could multiply capacity in food banks, settlement agencies and employment services. If adoption goes well, that multiplies public benefit. If displacement is higher, our social systems are already strong. Either way, it pays. The strategy mentions nonprofits, mostly as hosts for youth job placements but the shock absorber deserves support before the shock.

There is no productivity gain without data

This is true for government departments, companies and now for a country. A strategy built on a 60 percent adoption target needs a reliable source of truth on what AI is actually doing to Canadian productivity, wages and employment.

Statistics Canada already does excellent work; its surveys are where the 12 percent adoption figure comes from. We can build on that with regular public reporting, and put Canada’s deep bench of economists, PhDs and researchers on it. The UK just launched an AI Economics Institute, chaired by Nobel laureate Simon Johnson, with data-sharing agreements from the major AI developers. Canada already has the experts and the institutions, we need the mandate, the discipline, and a data-driven culture. If we cannot see the performance, we cannot manage it, and we certainly cannot claim it.

Public money should ask for something back

Whatever we put AI into gets amplified. If we are efficient, AI can make us more efficient. If we are inefficient or incompetent, AI can scale our inefficiencies and incompetencies in new and creative ways.

Public investment is no different, which raises three uncomfortable questions. Do we back the best and the brightest, or the best and the brightest with ‘Canadian experience’? That second filter has kept many capable newcomers out of jobs. How do we make sure Canadians use the best systems while supporting Canadian-made technology? How and when do we decide something is not working and give the money to someone else?

At the heart of the AI strategy is a commitment to building a “sovereign AI industry and research community” and “scaling Canadian champions”. This is the focus of the fourth and fifth pillars.

Canada must move to a real performance culture, with measurement, consequences and reallocation when needed. That may mean taking funding away from friends. AI will expose us either way.

In one case, Ottawa committed up to $240 million to a single company’s compute project in Ontario, built and operated by a U.S. firm. What is Canada getting back? Jobs, R&D and intellectual property that stay here? Name it, and report on it. Will the government take equity stakes in the companies it backs? The new $500 million Canadian Tech Growth Fund explicitly allows it. If public money helps build multibillion-dollar companies, Canadians should share the upside.

There is the speed of trust and there is the speed of government procurement. One ask of everyone, every RFP for public AI money should include a section on how the applicant will support AI skilling and inclusion, the impact their work may have on job displacement, and any potential mitigations. At the least, we collect useful data and at best, we normalize the principle that public AI money comes with public responsibilities. The scale demands it, Canada has about 42 million people and a labour force of more than 22 million. Reaching them is an all-hands-on-deck project, and we all have skin in the game.

Contribute, don’t reinvent

Pillar six focuses on “partnerships and global alliances.” It declares that “a coalition of aligned democracies, who pool research, talent, compute, and procurement power, would offer a credible alternative to the dominant market actors that increasingly define the global AI landscape.”

As a middle power Canada lacks the market size to make the world follow a made-in-Canada rulebook. “We are sovereign, we will write our own laws” sounds patriotic, but a standard no one beyond our borders must follow is a compliance burden on our own companies.

Meanwhile, the EU has done the work. Its AI Act is in force. The code of practice mandates real, evidence-based audits, and the frontier labs are engaging. That is a working standard. Canada, by comparison, is late to the party; our own AI bill, AIDA, died on the order paper. Should we build a new regulatory dialect from scratch when the most like-minded democracies have built one the market already respects?

The Prime Minister has made this case himself. At Davos he said Canada cooperates with like-minded democracies so we are not forced to choose between hegemons and hyperscalers, because if you are not at the table, you are on the menu. Perhaps the path forward is to contribute to the standard with weight behind it, and adapt for our local context. Canada can do that, and fast.

Conclusion and homework

These questions are linked, and they are all about people. The strategy is called AI for All. From those of us winning prizes onstage and raising millions, to those of us cleaning the conference venue and serving the food. We have the strategy; now comes the hard work of “how”, with people at the centre. Very many Canadians are just hearing about AI, so here is the homework, discuss this strategy at the dinner table, with a neighbour, in line for coffee, at the departure gate, or at your next social event. Whether it is safety, literacy, or jobs, it starts with education, education, education. We will need to educate ourselves, our communities, and even our models.

Acknowledgement 

Many people worked hard to get Canada to this point. The AI Task Force, public servants, translators, reviewers, and ecosystem partners, and that work deserves acknowledgement.