Evan Solomon, Canada’s Minister of Artificial Intelligence, just signed a memorandum of understanding (MOU) with Cohere to bring AI into the federal public service. Cohere is the only Canadian firm building large language models (LLMs) for widescale, industrial use. And that matters to the Carney government. It needs sovereign AI to build a sovereign economy and that calls for Canadian LLMs. Cohere thus fills a critical gap in the government’s plan—which makes this MOU well worth watching.
Why Cohere is a Good Bet
AI underlies the Carney government’s economic strategy. As Solomon says, “AI is not just a tool. It’s a revolution.” It’s the new infrastructure that will drive productivity, competitiveness, and national security for decades. But if your goal is also economic sovereignty, you need the right kind of AI infrastructure, and it’s Solomon’s job to begin building it.
The public service is a logical starting point. That’s where government meets citizens day to day. It’s also where inefficiencies are most obvious, and where Ottawa has enough control to run experiments. In effect: the public service is Solomon’s sandbox. If he can get AI working there—streamlining paperwork, cutting wait times, automating complex tasks—it will set an example for businesses across the economy.
Cohere’s role is pivotal. Ottawa has already invested $240 million in the company. Now, with Bell as a partner, Cohere could use the public service to help build what Carney and Solomon really need: a sovereign AI stack. A stack is just the layers they need to make AI run—from data centres and fibre networks at the bottom, to large language models in the middle, to the applications businesses need on top. But if any of those layers sit outside Canada’s control, sovereignty starts to bleed away.
There’s more. A sovereign AI stack must reflect Canadian values. That’s critical. Under Trump, the US is racing toward deregulation, while in Europe, heavy regulation stifles growth and innovation. Solomon wants Canada to take a middle path, which he refers to as “light, tight, right regulation.” The goal is to keep Canada competitive, while ensuring clear rules for privacy, security, and accountability.
That’s also where the Bell–Cohere partnership comes in. Bell brings the muscle—infrastructure, networks, and enterprise services. Cohere brings the brains—its LLMs and the North platform for building AI agents. Together, they form the basis of a sovereign stack, that is, AI that is governed by Canadian law. If Cohere delivers real gains in the public service, Ottawa will have something concrete to show businesses. Sovereign AI will look less like a slogan and more like a strategy.
Risks and Pitfalls
None of this will be easy. Lots of things could go wrong. The first hurdle is sheer scale. The US has at least five model builders—OpenAI, Anthropic, xAI, Google, Meta—each of which dwarfs Cohere. OpenAI alone says ChatGPT has over 700 million weekly users and is spending hundreds of billions on infrastructure, with talk of trillions to come. By contrast, Canada is trying to build a sovereign stack with one domestic player and a small fraction of the resources.
Ownership poses another challenge. Canada has a long history of promising tech firms being bought out. While Cohere’s co-founder, Aidan Gomez, insists Cohere is “not for sale”—and has even called acquisition a “failure”—resisting lucrative offers will be difficult. If Cohere succeeds, big competitors like OpenAI and Google will have every reason to want to acquire it, depriving Canada of its only foundational AI builder.
There is also political risk. Carney has promised efficiency but, as the Globe and Mail recently pointed out, many will expect that to generate cost savings by eliminating jobs. That will make the politics of AI-driven change in the public sector far more challenging.
Finally, there’s a risk of symbolism running ahead of substance. An MOU is not a contract. It raises expectations but doesn’t deliver outcomes on its own. If early projects stumble, people will quickly conclude that sovereign AI was just another slogan. Once the confidence is gone, it’s hard to get back.
Conclusion
The Cohere MOU isn’t just about automating processes or reducing wait times. It’s a test of whether Canada can stake out a sovereign place in the next wave of digital technology. Carney has tied his government’s economic renewal to that goal. Now Solomon has to show that we can build an effective Canadian system—one that is visible to citizens, credible to investors, and secure against foreign control.
The good news is that the government is doing exactly what it should be doing. It can’t avoid the risks, but it does have to give Canada a fighting chance. This agreement is a step in that direction.
If it succeeds, the message will be clear: Canada can still be a builder in the age of AI, not just a buyer. If it fails, the country risks sliding back into dependence, watching others write the rules of a technology that will shape everything from public services to national security. That is why this modest-looking MOU is more important than it seems. It’s a measure of whether Canada can stand on its own feet in a world of AI giants.
Don Lenihan PhD is an expert in public engagement with a long-standing focus on how digital technologies are transforming societies, governments, and governance. This column appears weekly. To see earlier instalments in the series, click here.