Canada Is Still Asking the Wrong AI Question

Geoffrey Hinton’s appearance before the Senate’s Standing Committee on Social Affairs, Science and Technology should have been a prompt for a bigger national conversation. The committee met on February 26, 2026 - to examine - “the impact of artificial intelligence in Canada,”

That is all good. Necessary, even.

But Canada still has a bad habit in digital policy: we keep asking the important question one level too low.

We ask how to regulate AI.

We ask how to manage harms.

We ask how to prepare workers.

We ask how to stay competitive.

Reasonable questions, all of them.

They are just not the biggest one.

The bigger question is this: - who controls the intelligence infrastructure Canada is about to depend on?-

That is where this starts to look very familiar. In cloud, we spent years pretending sovereignty was mostly a geography problem. Put the server in Canada. Put the contract under Canadian wording. Add a tasteful maple leaf to the architecture diagram and everyone can go home feeling patriotic.

But sovereignty was never really a postal-code problem. It was a - control- problem.

Who owns the platform?

Who holds the operational leverage?

Who controls the key layers?

Who can be compelled?

Who can interrupt, inspect, deny, influence, or price access?

AI is the same story, only bigger and faster.

If the models are owned elsewhere, the compute is concentrated elsewhere, the cloud layer is controlled elsewhere, and the commercial incentives are set elsewhere, then Canada may be participating in AI without being sovereign in any meaningful sense. We may be using the future quite enthusiastically while renting all the important parts.

That is the piece our policy conversations still tend to blur.

Even the committee’s framing — to study the - impact- of AI in Canada — hints at the issue. AI is treated as something that arrives here and happens to us, like weather with venture capital.

But AI is not just an external force to be managed. It is becoming infrastructure: for decision-making, research, administration, productivity, security, communications, and eventually for how institutions themselves operate. And when infrastructure becomes strategic, the first question is not just what harms it may cause. The first question is who holds the levers.

That is why I think too much of the Canadian discussion still misses the bigger picture.

The core issue is not simply whether AI will displace jobs, generate bias, spread misinformation, or need regulation. It will do some combination of all of that. The deeper issue is that AI is concentrating power in a stack that includes models, compute, cloud, capital, and distribution. If Canada does not think seriously about control across that stack, then “AI policy” risks becoming a very polite form of dependency management.

We have seen this movie before.

In cloud, many institutions convinced themselves they had solved sovereignty because the data lived “here.” Meanwhile, the meaningful leverage still sat elsewhere. Ownership sat elsewhere. Legal exposure sat elsewhere. Platform power sat elsewhere. The server had a Canadian postal code; the power did not.

Now we are in danger of doing the same thing with AI.

We will tell ourselves that Canada is active because we have hearings, strategies, advisory councils, pilot projects, procurement language, and responsible-use frameworks. All worthwhile. But none of that, on its own, answers the sovereignty question.

Because sovereignty in AI is not about whether Canadians use it.

It is about whether Canada can meaningfully - shape, constrain, verify, and, when necessary, operate independently of- the systems that will structure public and economic life.

That has implications for what public leadership should actually do.

A serious national response cannot just be tactical. It cannot be limited to “encourage adoption,” “coordinate policy,” or “govern the data estate.” Those are administrative functions. They matter, but they are not strategy.

Strategy means asking where dependency is accumulating.

Strategy means building technical capacity inside the state.

Strategy means treating auditability, procurement leverage, model evaluation, and operational competence as national capabilities, not side projects.

Strategy means understanding that sovereignty without control is mostly branding.

That is the AI question Canada should be asking.

Not just: How do we respond to AI?

But: What kind of country are we if the core intelligence layer is owned, operated, and governed elsewhere?

The Senate hearing was timely and worthwhile. But the real challenge is larger than AI harms, larger than innovation policy, and larger than one committee study. It is about power.

And in digital policy, power has a nasty habit of hiding behind neutral words like platform, service, ecosystem, and adoption until one day you realize you do not really control the thing you now depend on.

A maple leaf on the briefing binder will not fix that.

Originally published by Ross Norrie, founder of SkyeConnex, on LinkedIn.

Published March 8, 2026 · More from the SkyeConnex blog