Andrew Marble
marble.onl
andrew@willows.ai
November 5, 2025
Since Canada has been throwing around quotes from US Presidents recently, there are two I find very relevant to Canada’s AI industry:
Government is not the solution to the problem: government is the problem.
— Reagan
ask not what your country can do for you — ask what you can do for your country.
— JFK
Canada’s budget was announced yesterday and included the promise of a made-in-canada “AI Tool”:
… will develop a made-in-Canada AI tool that can be deployed across the federal government. Shared Services Canada will partner with leading Canadian AI companies to develop this internal tool. By supporting innovative research to strengthen public services, this work will protect our digital sovereignty, keep government data and information safe in Canada, and create opportunities for the Canadian technology sector.
The most charitable reading I can give is that there are national security reasons why the nature and purpose of this “tool” are not mentioned (it is to be built in partnership with the Department of National Defence and the Communications Security Establishment).
Regardless, in practice the outcome will almost certainly be overpaying some politically connected companies for an inferior product, in the name of sovereignty and being “Made-in-Canada”.
Sovereign capability is important, and I have been thinking about it a lot recently. Though no indication of what the “tool” the government will build is given, having AI models, specifically large transformer models like GPT et al., and the capability to train them, is a key aspect of sovereignty. At the same time, I see little value in having our own home-made tools if they are not competitive.
If we are concerned about autonomy, we would be better to just download new releases of the best open models as they become available and archive them. These models can be run locally without reliance on foreign tech companies, and choosing the best would mean we have world class models at our disposal. On the other hand, building a made-in-canada model that is not among the best, would leave us with an inferior product that nobody will use and that would leave us disadvantaged should the need for autonomy arise.
If the above sounds like a false dichotomy (use foreign models or use inferior ones) then we need a way to properly measure the competitiveness of anything we build ourselves and make this our focus. I would suggest there is an easy test for whether we’ve built a good AI model: other people and organizations use it. It’s not enough to be an also-ran, we need to focus on making something globally competitive if we want to pursue sovereignty. The US and China know this and both have industry-led, globally competitive AI and strong sovereign capabilities. If our goal isn’t at least to place amongst the top handful of globally relevant models, we might as well not bother.
Measuring success thus comes down to measuring downloads, mentions, and incorporation into other projects. If we can make something other people want, we will have demonstrated that we have a sovereign AI capability, with both the work product (the model) but also the expertise to remain cutting edge. While obviously difficult, with the right focus building a popular and powerful model is still a realistic goal. There are lots of unmet needs and lots of unaddressed issues that provide opportunities to make models that are both generically "better" and better address factors that make them popular to use.
I would argue that such a system will never come from the Canadian government (regardless of the party in power), either as a direct undertaking, or by the government commissioning a third party (which will inevitably suffer the fate of all government procurements and optimize for the wrong things). Worse, the market distortion caused by government efforts in this area ends up drawing off talent and attention and distracts industry from developing real market competitive capabilities.
Building a Canadian AI capability is a challenge for the Canadian private sector to take on, dually motivated by helping create a vital sovereign capability, but also by the profits and recognition that come from building a world class AI model. The sooner we (the Canadian AI industry) stop looking to government for support and start building, the sooner we begin to realize real market competitive, sovereign AI capability.