A group of Canadian foreign-policy experts is urging Canada and the European Union to outline a new Middle Power Charter at next week’s Canada–EU Summit in Montreal.
They compare the idea with the 1941 Atlantic Charter, in which U.S. President Franklin Roosevelt and British Prime Minister Winston Churchill set principles for an international order, including peoples’ right to choose their own governments, free trade and freedom from fear and want. Eighty-five years later, the proposed charter would unite democracies determined to defend their independence, tackle problems none can solve alone, and co-operate on technologies tied to national security, economic prosperity and self-governance.
The argument is that a new technology order shaped by artificial intelligence requires a model of technological sovereignty: each nation would cultivate critical digital capabilities and co-operate closely with trusted partners. Canadian hospitals, businesses and governments are adopting AI systems supplied mostly by American companies, and citizens are beginning to use personal AI agents such as Meta’s MUSE. The piece asks who would own the knowledge those systems accumulate, who would control access and who would set the terms.
It warns that an American AI company, perhaps under U.S. government pressure, could suspend service to a Canadian institution in a cross-border dispute, and that as AI displaces knowledge work Canadian companies could pay licence fees to providers rather than wages to employees. A data centre in Canada would not settle the issue if software, operating decision rights or legal authority remained abroad. The authors say Canada needs enough computing power, secure storage, reliable models, skilled people and domestic control to keep key services running, with sensitive public information subject to Canadian law so hospitals, the financial system, government and critical infrastructure can operate on Canadian-controlled systems in a crisis.
They propose an AI Alliance for Middle Powers including Canada, Europe, Japan, South Korea, Australia and others. Each country would maintain its own digital foundation while sharing know-how, large-scale computing, research, networks and services under common technical rules, standards and licences. Applications could move between trusted providers, sensitive data and key operations could stay at home, and Canada could still function if an ally changed course or the alliance collapsed.
The partners could also help their AI companies grow. Canada has Cohere; France has Mistral. Universities, entrepreneurs and public laboratories across these countries are described as having much to contribute. Joint access to computing power, investment and government purchasing could give promising companies a chance against industry giants elsewhere. The authors call for competition rather than a government monopoly, and for openness built around standards, open-source models that organisations can inspect, adapt and run themselves, research tools and verifiable public data, including help for developing countries.
On governance, they list bias, privacy threats, disinformation, job displacement and concentrated power, and a further risk of recursive self-improvement in which AI improves its own capabilities, potentially beyond human control. They say middle powers cannot stop U.S. or Chinese labs or dictate rules for humanity, but can apply shared principles at home, license and inspect powerful systems, require disclosure of dangerous capabilities and failures, and use their collective weight to seek broader agreements. Humans should retain authority over decisions involving lives, weapons and essential services; agents should have limits, records and a way to be stopped, and should have no legal personhood. Countries would need limits on dangerous uses and emergency procedures for an AI crisis, with the UN able to convene wider talks while national laws and co-operation give agreements force.