Last month, Canada played a central role in world politics when Prime Minister Mark Carney delivered a consequential address at the World Economic Forum annual meeting in Davos, outlining Canada’s response to what he described as a rupture of the world order.

In what many interpreted as a ‘eulogy’ for the liberal international order, Carney argued that middle powers like Canada can no longer rely on the multilateral institutions and the economic and security agreements of the old, rules-based order. Instead, they must build strength at home to act with credibility abroad, pursuing a foreign policy that is both principled in its commitments to its core values, and pragmatic in recognizing the need to build coalitions with partners who may not share every one of them.

To illustrate what this ‘value-based realism’ demands, Carney invoked Czech dissident Václav Havel’s idea of ‘living in truth.’ In Havel’s famous parable, the shopkeeper who hangs Soviet slogans in his window, which he privately knows to be false, helps sustain a system whose power rests on the performance of its rituals, rather than the belief in it. Carney extends this to international relations, arguing that middle powers now face a similar choice to continue performing the rituals of the rules-based order, or to ‘live the truth’ by calling out coercive great powers and by building institutions that align with their interests and values.

The prime minister’s speech was no doubt remarkable, which makes it all the more essential for his policies, both foreign and domestic, to match the standard he sets. Artificial intelligence presents him with a difficult test. Is his government’s AI approach ‘living in truth,’ or performing the rituals of great power rivalry? The signs suggest the latter.

The United States has framed its AI policy as “a race to achieve global dominance,” with winner-takes-all stakes. China has responded in kind.  Experts caution that the eventual winner of the AI race would have a ‘decisive military advantage’ that “could engineer an existential catastrophe.”

For middle powers like Canada, Carney’s value-based realism would suggest charting an alternative course: refusing to participate in a race that threatens democratic values, the environment, and human rights to the imperatives of geopolitical competition.

Yet Canada’s approach has thus far followed the speed and logic of great power rivalry. In the fall, the government launched a 30-day public consultation on its AI strategy. More than 40 civil society organizations and 120 experts and advocates rejected the process outright in an open letter, calling the accelerated timeline evidence of “serious disregard for the Canadian public’s known and wide-ranging concerns” about AI’s demonstrated harms. They noted the “contrived urgency” and the absence of human rights or civil liberties representatives on the government’s AI Strategy Task Force. The rushed consultation process invokes the language of values and sovereignty while adopting the substance of the arms race. 

Canada therefore faces a fundamental choice about how to position itself in AI development. The temptation is to accept the terms set by Washington and Beijing, to compete in their race for the fastest computational infrastructure and the most dominant market position. But as researchers from the University of Ottawa’s AI + Society Initiative observe, “Canada simply can’t outcompete American companies with vastly greater sums to invest.” Canada lacks the resources, market size, and tech ecosystem of the superpowers. Trying to compete on their terms means inevitable subordination, eventually becoming exactly the kind of dependent middle power Carney warned against. Racing to keep pace is just another way of hanging the slogan in the window.

But Canada can compete on different terms: green, ethical, and sovereign AI. This means pursuing, as researchers Anne Pasek, Jiaqi Wen and Kelly Bronson argue in Policy Options, AI systems that “can outcompete general models in their designated tasks, while also greatly enhancing the opportunity for meaningful community consultation.”

This approach would position Canada to lead the kind of coalition Carney described in Davos. His framework of ‘variable geometry’ – different coalitions for different issues based on “common values and interests” – already has natural partners emerging. The European Union has implemented its AI Act after years of multi-stakeholder processes. Nordic countries are pioneering sustainable AI infrastructure. Middle powers are increasingly feeling squeezed by the U.S.-China AI race and seeking alternatives.

Those who, in the spirit of pragmatism, believe that the government must choose the pragmatic path over the principled path, fail to see that these two are complementary, not competing, to one another. The principled path is the pragmatic path. Policy scholars demonstrate that while middle powers “face catastrophic risks” in an AI arms race individually, collective action can “deter dangerous AI R&D by any actor, including superpowers.” Far from being a prisoner’s dilemma where defection is inevitable, it is rational to seek cooperation to achieve a ‘trust dilemma,’ where states prefer mutual restraint to racing.

A middle-power coalition would offer the world a sustainable and accountable alternative development model to the AI the arms race approach of coercive great powers. Countries and companies are increasingly seeking trustworthy democratic partners, as public trust in and the “social license of AI developers” erodes. Canada could “usefully distinguish itself from its neighbour by cultivating an industry that does not compete in a winner-takes-all race to monopolize the market, but instead offers ethical integrity and sustainability as core product features.”

This alternative consists of three pillars: the environment, safety, and democracy.

Training and running AI models and systems relies on large-scale data centres. These are physical facilities that concentrate immense amounts of computing power in single locations. This computing demand requires extensive electricity consumption, making AI one of the fastest-growing sources of new energy demand in advanced economies. Where electricity grids are already congested, this demand is increasingly met through fossil fuels.

The environmental impact of AI is already evident. According to the International Energy Agency (IEA), data centres already account for about one percent of global electricity consumption. The IEA projects emissions from data center electricity will rise from 180 megatonnes of CO2 in 2024 to between 300 and 500 megatonnes by 2035. In hubs like northern Virginia, where data center growth is geographically concentrated, household electricity costs have increased by as much as 267 percent over five years. Where clean sources of energy cannot be built quickly enough, utilities have extended the life of coal plants to keep pace with demand.

In Davos, Carney reaffirmed Canada’s commitment to sustainability to like-minded allies. Yet the federal government has provided no guidance on how provinces should allocate scarce clean energy or manage the emissions impact of data center expansion. Alberta’s AI strategy encourages operators to ‘bring your own power’ through dedicated gas-fired generation. This approach, if fully implemented, could double the province’s electricity sector emissions and reverse gains from the coal phase-out. Similarly, Ontario projects that data centers will account for 13 percent of new provincial electricity demand by 2035, with no coordinated plan to ensure this growth aligns with climate commitments. Therefore, while electricity distribution falls under provincial jurisdiction, it is imperative for federal coordination to ensure electricity grids are not strained and the country remains on track to meet its 2030 emissions reductions targets.

This imperative is consistent with Canadian public opinion. The summary report of the federal consultation on AI outlined that “environmental sustainability is a major concern for many, with calls for strategies to mitigate energy consumption and water usage in data centres. Respondents urged Canada to develop a national AI infrastructure roadmap that prioritizes sovereignty, sustainability and public benefit.”

Countries like Finland have become attractive markets for green data centre investment. Microsoft’s hyperscale data center in Espoo, for example, is being developed through a coordinated partnership with municipal authorities and the local utility to recover waste heat and supply it to 100,000 homes, accelerating the closure of a coal-fired heating plant. Canada has the clean energy advantage, with 85 percent of its electricity already non-emitting, and a cold climate to lead in sustainable AI infrastructure. It lacks only the political will to impose the standards such leadership requires. Sustainable infrastructure becomes an organizing principle for coalition partners seeking a clean alternative to fossil-powered data centers.

Safety and Information Integrity

The arms race logic also prioritizes speed over safety, with particularly concerning implications for children and the information environment that sustains democracy. In October 2024, the G7 data protection and privacy authorities, including Canada’s Privacy Commissioner, issued a joint statement identifying “manipulation and deception” and “AI-based decision making” as areas of particular concern for children. The statement highlighted how AI companions could exploit children’s tendency to form emotional bonds and disclose sensitive personal information, and how deepfake technology enables the creation of sexually explicit imagery. Children’s personal data is being scraped from public sources to train AI models, often without meaningful consent or oversight.

Similarly, generative AI tools have fundamentally destabilized the information environment. When OpenAI released its Sora video generation tool in late 2025, users immediately deployed it to create “strikingly realistic videos of ballot fraud, immigration arrests, protests, crimes and attacks on city streets.” Computer scientist Hany Farid warned, “I worry about it for our democracy. I worry for our economy. I worry about it for our institutions.” The proliferation of such content creates what researchers call the ‘liar’s dividend’: as fake content becomes more convincing, people lose the ability to trust authentic evidence, allowing bad actors to dismiss genuine documentation as fabricated.

‘Living in truth’ would mean establishing strict mandatory safety testing before deployment of AI systems likely to affect children or public discourse, age-appropriate design requirements, and leveraging Canada’s middle power tradition, as the nation that championed the landmine ban and the International Criminal Court, to lead international norm-setting around AI applications that threaten human rights. The arms race demands speed at any cost. Democratic values demand we count the costs before we pay them.

The rushed AI consultation reveals a third domain where policy diverges from principle: democracy. The survey instrument had “predetermined framing,” with questions focused overwhelmingly on economic competitiveness and investment attraction. Questions about safety and human rights were framed as secondary considerations. ‘Living in truth’ would mean establishing consultation processes that provide Canadians sufficient opportunity to express their views and concerns. 

A fair survey, however, would alone be insufficient to meaningfully collect the input of Canadians on the governance of such transformative technology. Democracy scholar Aviv Ovadya argues that it is necessary to reinvent democratic infrastructure to “enable democracies to better confront the challenges arising from AI.” He proposes various interconnected democratic innovations, such as randomly-selected deliberative citizens’ assemblies that are provided with the resources to grapple with technical complexity while maintaining democratic legitimacy.

Some might dismiss such proposals as impractical. Yet civic society organizations are already demonstrating their feasibility. McGill University’s Centre for Media, Technology and Democracy is leading Gen(Z)AI, a series of forums bringing together 100 young delegates who reflect Canada’s demographic diversity to deliberate and provide policy recommendations on concrete AI policy issues. As a participant in this process, I have seen firsthand that the institutional capacity to engage Canadians meaningfully already exists. What is missing is the political commitment to ensure such deliberative processes actually inform policy rather than serve as information for the government to selectively adopt or ignore.

Democracy is not a 30-day sprint. Nor can it be left to tech companies racing for dominance or politicians chasing short-term gains. The protection of Canadian democracy should be non-negotiable. It should be the foundation for building coalitions with partners who share that commitment.

Havel’s shopkeeper had a choice. So does Canada. We can keep hanging the slogan in the window, by performing the rituals of great power competition, rushing to keep pace in an unwinnable race. We can accept coal plant extensions and double emissions. We can sacrifice democratic governance for speed, reducing public input to a 30-day formality. We can subordinate safety to competitive imperatives, treating our children and information integrity as acceptable costs.

Or Canada can take down the sign. We can build a competitive advantage and lead a coalition of middle powers who refuse the AI race and offer the world an alternative: green, ethical, sovereign AI development. This is the principled and pragmatic opportunity before Canada. It is principled in refusing to sacrifice values for speed. It is pragmatic in recognizing that Canada cannot win by competing on the superpowers’ terms, but can win by setting different terms entirely. In Davos, Carney declared: “Friends, it is time for companies and countries to take their signs down.” It is time for Canada to meet the standard it set.

© OpenCanada