In the public sector, adopting AI is not only a question of capability. It is a question of legitimacy. Citizens do not choose their government the way they choose a product, so the bar for how public institutions use technology is higher: decisions must be explainable, accountable and fair, and they must stay that way over time.
Two ideas now dominate the conversation: sovereignty and responsibility. They are related but distinct. Sovereignty is about control, where data lives, who can access it, which laws apply, and whether a critical capability can be withdrawn by a vendor or a foreign jurisdiction. Responsibility is about conduct, whether a system is used in ways that are fair, transparent and safe for the people it affects.
Sovereignty is easy to reduce to where the servers sit. That matters, but it is narrow. Real sovereignty is the ability to keep operating on your own terms: to understand what a system does, to change vendors without starting over, to meet legal and language obligations, and to refuse a use that does not serve the public interest. A government that cannot explain or modify a system it depends on has outsourced far more than infrastructure.
Responsibility, in practice, is less about principles and more about process. Who signed off on this use. What data trained it. How errors are caught and corrected. Where a human stays in the loop for decisions that affect someone's benefits, status or liberty. What the institution will tell the public when something goes wrong. These questions have answers only if they are asked before deployment, and revisited after.
The bilingual and federated nature of Canadian institutions adds a layer. A responsible system here works in both official languages, respects provincial and federal boundaries, and reflects communities that are not uniform. Tools built and tested elsewhere often miss this, and the gaps show up as quiet failures for the people least able to push back.
None of this is a reason to avoid AI in public institutions. It is a reason to adopt it deliberately: to treat sovereignty and responsibility as design requirements rather than afterthoughts, and to build the governance and capability to uphold them. Done well, that is not a brake on adoption. It is what makes adoption durable and trusted.
This is the balance we help public institutions strike: technology that earns public trust rather than spending it.