Last month, Pope Leo XIV released the Magnifica Humanitas—a statement of faith in the “grandeur of humanity” in the face of an ascendant artificial intelligence (AI) industry.
Yesterday, prime minister Mark Carney released what we might call the Magnifica Technologia—a statement of faith in the AI industry in the face of a skeptical public.
The new federal artificial intelligence strategy, AI for All, has all the trappings of a policy document. But it is, at its core, a declaration of belief. “For Canada to thrive in the era of AI,” it argues, “Canadians need to trust in its promise.” We must have faith.
Evidence receives short shrift in this doctrine. The adoption of AI will necessarily deliver “better work, stronger services, new ideas, broader prosperity and greater wellbeing,” it asserts. And while some of the AI-related concerns raised by Canadians are acknowledged—from privacy and safety to sovereignty and job loss—they are superseded by the adoption imperative. Canada must implement its AI vision “with urgency.” We cannot afford to slow down.
In this article, we identify seven key takeaways from the federal government’s new strategy. They cover a lot of ground, but what ties them all together is the government’s ideological commitment to AI adoption as a matter of principle even (or especially) when that principle is unsupported by evidence. Let’s get into it.
Takeaway #1: We’re not all talking about the same AI
Artificial intelligence is never defined in the strategy—an omission so fundamental that it is hard to imagine it was an accident. That ambiguity is the source of serious conceptual drift.
For example, the plan identifies five priority sectors for AI adoption—health sciences, energy, transportation, agriculture and manufacturing. In each case, the focus is on “practical, sector-specific applications,” which generally means purpose-built machine learning tools.
That stands in contrast to the generative AI tools that the public is most exposed to and that are driving large-scale investment in the AI industry (including in data centres). In fact, generative AI is mentioned by name only twice and chatbots are mentioned only once in the entire document, even though they are the AI applications that are by far the most publicly salient and in the most dire need of regulation.
What this amounts to is a cynical bait-and-switch. For example, the federal government is foregrounding the adoption of inoffensive, sector-specific applications to justify new data centres, even though those data centres are mainly necessary to power contentious generative AI models. Similarly, the government warns against overregulating AI so as not to “smother” innovation, even though sector-specific and consumer-facing tools are often entirely different technologies that can be regulated separately.
Takeaway #2: The federal government doesn’t understand why people don’t use AI
The strategy claims that its “north star” is building trust in AI. However, as the document makes abundantly clear, the true guiding principle of the strategy is AI adoption. Everything else, including trust, is a means to that end. The strategy specifically blames “low literacy and low trust” among Canadians as the biggest barriers to bridging the AI “adoption gap.”
The strategy offers various solutions. For example, a new National AI Literacy Initiative will target one million post-secondary students for training, equip 3,000 educators with AI learning kits, and offer free AI programs at public libraries. To build trust, the strategy offers milquetoast promises of safety and oversight (see below).
Digital literacy is a vital priority, but there is a big difference between critical literacy and technical training. In this case, the goal is to “build an AI-skilled nation” that uses AI confidently “at school, at work, at home, or in the community,” not to equip Canadians to push back against AI industry hype.
Moreover, whether or not these approaches succeed in building literacy and trust, they fail to address the many other reasons why individuals and institutions may be reluctant to adopt AI. Two in particular stand out for their omission.
First, the document does not grapple with the technical limits of AI systems. For example, generative AI is inherently prone to randomness and hallucination, which is a serious barrier to adoption in any context where accuracy matters. Literacy and trust can only go so far when the underlying technology is not suited to particular tasks.
Second, the government does not acknowledge the many moral and cultural objections to AI adoption. It is not only Catholics like Pope Leo who view humanity as sacred. For many, the arrogance and appetites of the AI industry—stolen training data, environmental impacts, automation of human judgment, etc.—are affronts that no amount of “literacy” will overcome.
Takeaway #3: AI regulation is coming—sort of, maybe, someday
The strategy’s safety pillar opens with an admission that the risks of AI are real and that the government must protect Canadians from AI-related harms. On privacy and consumer rights, the strategy proposes new privacy legislation to establish a “fundamental right to privacy,” strengthen individuals’ control over their personal data and restrict harmful practices such as surveillance pricing. The government promises to develop legal tools to combat deepfakes and protect elections and democratic institutions from AI-enabled misinformation. It also promises to test AI models and certify “trusted” AI products.
Unfortunately, that’s as far as it goes. The safety pillar is full of commitments but devoid of specifics. How the government will require AI companies to watermark their outputs, for example, or how the government will identify and prevent electoral disinformation, is unclear. It also fails to draw any hard lines. Unlike the EU, Canada has no prohibited AI uses list, no high-risk application categories and no binding restrictions on the AI deployments that carry the greatest potential for harm, including facial recognition, social scoring, algorithmic management of workers, biometric data collection, or automated decision-making in hiring, housing and benefits delivery.
Canada has been here before. The failure of Bill C-27 means privacy reform has stalled for years. The new strategy offers no indication of how this time will be different or how these commitments will withstand the same industry pressure that killed its predecessor. It’s also not clear how a renewed commitment to privacy can co-exist with the controversial Bill C-22 that would actively weaken privacy rights in Canada.
Takeaway #4: AI adoption is an industrial strategy for Canada
The strategy makes passing reference to the “betterment of society,” but the real impetus behind AI adoption is “economic growth, productivity, industrial capability, and geopolitical influence.” Five of the plan’s six pillars are explicitly focused on building the AI industry. AI for All is, unequivocally, an industrial strategy.
On this measure, it is one of the best examples of industrial planning that the federal government has produced in the past decade. The document checks all of the boxes of good industrial strategy, including a clear vision, measurable objectives, public coordination and domestic capacity.
To that end, the plan commits billions in public funds for the domestic AI industry, including $500 million in direct subsidies for Canadian AI startups, $1.75 billion in venture capital stimulation, $700 million in subsidized compute access and $130 million for commercialization programs. The strategy also introduces a $200 million AI Missions Program, which will support projects that “demonstrate meaningful improvements in Canadians’ lives” (in contrast to all of the other publicly-subsidized AI projects, it would seem). It’s a promising idea, but without strong guardrails these AI “missions” will likely lead to public-private partnerships rather than true public goods.
Most of the money outlined in the strategy was announced in previous federal budgets, but it still represents a material commitment to domestic industry. The more important question is whether it’s the right amount of money. Compared to the trillion-dollar valuations of leading American AI firms, such as OpenAI and Anthropic, Canada’s biggest domestic AI champion, Cohere, is worth only US$7 billion. It remains to be seen whether a few billion dollars in federal support can move the needle meaningfully away from U.S. control of the global AI industry.
There’s also the question of whether the benefits outweigh the costs. For all the talk of AI-powered growth, by the government’s own admission widespread AI adoption will only lead to a “0.3 percent to 1.1 percent annual labour productivity increase.” That’s hardly transformative.
Takeaway #5: The strategy is pro-worker in the same way it is pro-trust—as a means to an end
The strategy claims to have a pro-worker orientation, but it says little about how workers will have a meaningful role in deciding whether and how AI is introduced in their workplaces.
This tension is particularly visible in the strategy’s treatment of women and young workers. It acknowledges that female-dominated sectors face early disruption from AI and that entry-level roles are shrinking. Yet its response is largely to accelerate AI adoption through existing programs while creating a $50 million AI-powered job bank for the unemployed. The workers most vulnerable to automation are effectively being offered more automation rather than stronger labour protections or minimum hiring commitments. Similarly, short-term placement programs for young workers do little to address the broader erosion of entry-level opportunities.
A genuinely pro-worker AI agenda would ensure workers and unions lead the design and deployment of workplace AI systems, particularly where those systems affect performance evaluation and job security. It would also address how productivity gains are distributed. If AI increases efficiency while wages stagnate, skills erode and jobs disappear, the benefits of adoption will accrue primarily to employers and technology firms rather than the workers whose labour generates them.
Takeaway #6: Industrial capacity takes precedence over democratic control at every turn
The strategy pays considerable attention to building AI capacity, but sidelines democratic oversight. Canadians should have “a meaningful voice in the public debates that will define AI’s role in Canadian society,” according to the plan, but no process for that is described. This gap is particularly evident in its approach to international partnerships, health data and Indigenous data sovereignty.
The Sovereign Technology Alliance that Canada is pursuing with other countries is designed to scale up AI supply chains, not AI governance. It is presented as a vehicle to help allied countries expand AI capabilities and procurement opportunities. A governance first approach would look quite different. For example, the alliance could instead develop binding accountability and enforcement mechanisms and condition trade partnerships on the protection of workers’ rights, data security and environmental standards.
The strategy’s approach to health data raises similar concerns. It commits $200 million to building a consolidated database of Canadians’ health information, describing patient-generated data as a strategic national asset to be unlocked for innovators. There is no mention of whether patients will be consulted, whether consent frameworks exist or how data will be governed once it enters commercial pipelines.
The strategy’s Indigenous AI commitment focuses on language revitalization and cultural initiatives. These are welcome, but they are not a substitute for Indigenous data sovereignty. The strategy makes no mention of free, prior and informed consent for data collection about Indigenous peoples or their lands, Indigenous governance rights over that data, or any mechanism allowing communities to shape or refuse AI deployments that may impact them.
Takeaway #7: Some of the biggest AI risks and harms are ignored completely
“To foster trust,” the strategy declares, “we will protect Canadians from the risks and harms of AI.” As we discussed above, the protections outlined in the document are largely aspirational, but they are at least a step in the right direction on several issues. More damning are the risks and harms that go entirely unacknowledged. Three stand out.
First, the strategy fails to acknowledge the theft of copyrighted works to train AI models, nor does it acknowledge the demands of artists and writers for consent, compensation and credit when their work is used. Instead, the strategy slaps creators in the face by establishing a $50 million Creative Technology Program to encourage creators to use AI in their work.
Second, the strategy fails to acknowledge the environmental impacts of AI. Instead, it celebrates the addition of 850 megawatts of compute capacity by 2030 with a potential expansion to 2.3 gigawatts. While the plan touts Canada’s clean grid as a competitive selling point, most new data centres are being built using highly-polluting natural gas generators. The strategy entirely sidesteps noise, water use, e-waste, land rights, energy affordability and other implications for local communities.
Third, the strategy fails to acknowledge the risks to cognition and mental health associated with AI use. Cognitive offloading, AI psychosis and related phenomena are still being studied, but the potential threats that they pose to individuals and society are substantial. Young people face the most acute risks, but no demographic is immune to the potentially negative impacts of AI on cognition and mental health.
These are not the only issues that the new federal AI strategy ignores (there is also warfare, corporate power, human rights, ethics and more) but they are emblematic of the underlying ethos of the Magnifica Technologia. The federal government seems to truly believe that “prosperity and sovereignty in this era belong to nations that can leverage trust to adopt, build, and govern AI.” In the context of that vision, there is no space for fundamental concerns about AI itself.
Nevertheless, Canada has its long-overdue artificial intelligence strategy. Now comes the hard part. The federal government must deliver on its commitments to safety regulation and it must be prepared to pick a side when those commitments come into conflict with the demands of the AI industry.
It must also be prepared to respond to the evidence, even (and especially) when it conflicts with the government’s vision of AI adoption. Faith can only take us so far.
