The federal government has tabled a flurry of tech-related legislation and policies recently—and Canada’s minister of artificial intelligence has been at the centre of it.
From Canada’s AI strategy to new laws that govern privacy, deepfakes, social media, and online surveillance, Minister Evan Solomon is one of the key players in Ottawa defining what Canada’s digital future will soon look like.
BetaKit caught up with the minister for a wide-ranging conversation on why the government is pushing Canadians to adopt AI, how the government plans to invest directly in it, and why Canada needs to have sovereign digital alternatives to US tech.
This interview has been edited for length and clarity. For the full conversation, listen to The BetaKit Podcast.
On stage at the BetaKit Most Ambitious: Town Hall, you said your job was to make sure that safe, transparent Canadian AI is adopted. Can you explain why the federal government feels like it should be in the business of adoption of a specific technology?
When we launched our strategy, AI for All, the “for all” matters, right? That’s our principle: that this technology should be available and accessible and beneficial for all people in Canada.
We don’t want the digital divide, we don’t want just a few people to benefit from this, we don’t want just the early adopters. Distribution matters. We need to get the lift from this—the real opportunities that AI presents and the risks that it poses. Our foundational principle is that everybody needs to have an opportunity here, and we should also try to mitigate the risks for everybody as well. AI for All is about trust, opportunity, and sovereign control, and essentially the whole strategy can be seen through those lenses.
It is not unique for a government to make sure that access to technologies are equally distributed. There are technologies, and there are general-use technologies. So, for example, we’ve had a universal broadband thesis to make sure that we have connectivity for rural areas, so that’s making sure that all Canadians have access. We built an entire CBC network to make sure that everybody in Canada has access to programming and information. Every education system is built on making sure that every child has access to key technologies. At one point, we spent a lot of money making sure everyone’s got access to a pencil and a pen and an eraser.
It sounds like you’re saying that AI—and being specific with AI is tough, because we’re talking about generative AI, LLMs, algorithms—you’re considering AI to be foundational like reading and writing?
Yeah. We’re at a specific moment.
When I started my first company, Shift Magazine, it was at the birth of the internet. My business partner and I had the notion that we were at the beginning of something brand new that would be transformative. Now we’re at the birth of something else, a general-use technology that will have transformative impacts, and we’ve got to be open to all the opportunities.
The vertical axis is that 150,000 people are working in AI, within 800,000 people who are in the digital innovation space in Canada.
The horizontal axis is where AI starts intersecting with others: in manufacturing, you see industrial AI; in agriculture, where you’re seeing AI technologies in how farmers are applying fertilizer; in healthcare, across every sector. That’s why our government is seized with the notion that this is going to be a general-purpose technology.
We need a literate, educated, skilled workforce. We need to have a regulatory environment that protects kids and our personal information and our privacy, and we need to invest to make sure foundationally, things are built here. Either we build the infrastructure here or we rent it somewhere else. Either we build the innovation here with our SMEs and our entrepreneurs, or we buy it from someone else, and either we make the rules here or we follow someone else’s.
Those are the choices we’ve got as a country. Then we’ve got a shot at shaping the thing that’s happening so quickly.
You put numbers to the AI strategy: 90,000 AI-related jobs and work opportunities for young Canadians; 250,000 total new jobs through AI adoption by 2031. A $200-billion increase in GDP from productivity tied to AI. Getting AI adoption to 60% by 2034, from 12 percent. Are there any other metrics that you’re looking at internally? What’s the process to track this?
We did this for this question exactly. You can have a strategy and just not put KPIs and goals, and then you’re not accountable. We think Canadians deserve to have real goals of what we’re trying to attain.
How are we going to do it? We’ve got the principle of accountability, and the strategic buckets, and then there’s a series of tactical investments.
We are going to create job placements, working with organizations like Mitacs. We’ll be working with partners like Patty Hajdu’s jobs ministry and the Canada Summer Jobs placement. These are paid placements for young people in small and medium-sized enterprises that we are supporting.
As for where the 250,000 comes from? Those come from OECD numbers. All the modelling that we use is based on OECD numbers, based on adoption rates that we think are very attainable.
How are you going to know which of these levers are most impactful?
What you’re looking for is the overall impact. First of all, on AI and the economy, there’s lots of modelling. We have data points. When you read a Bank of Canada report, you can disaggregate on labour market: what investments create what kind of job multiplier effect.
I’ll give you one example. We have a government-sponsored cluster called NGen for manufacturing. They had a study of manufacturers that are using AI in their manufacturing process. They’re measuring machine downtime—how often the machines are down less, because the AI is catching things early. They’re saying, “wow, we have had a real return on investment on AI, our productivity is higher.”
There’s a commitment in the strategy for the federal government to start taking equity stakes in Canadian companies. Some are uncomfortable with the idea of the government picking winners. Can you give any clarity on how you will approach making those investments?
We heard in this process that what people want is help on the three Cs—capital, compute, customers—and the two Ts—trust and talent.
We kept hearing “I can’t get investment here, I have to go to the US, or I have to go elsewhere.”
On the investment side, there’s $500 million through this new fund, the Technology Growth Fund.
Let’s get into that first $500 million. That’s supposed to give access to capital. That has the optionality of equity. There’s often a lack of a lead investor; we could play that role.
There’s often a lack of a lead investor; we could play that role.
We are in discussions of exactly what the structure will look like, but I will just tell you that there are options on debt instruments, so any fund you can lend, but we thought it’s important also to have equity optionality.
When you’re investing, governments don’t like risk because you’re politically volatile. If you’re a VC, you can afford to have misses if you have one hit. Governments, we have a responsibility. This is taxpayer money. We have to be careful.
RELATED: AI minister says Ottawa considering taking the lead on AI investment rounds
There’s another model in policy where we reward risk-taking. There are some suggestions in our strategy where we’re looking at ways to make sure that we’re rewarding risk in a better way to incentivize risk.
It’s not like you, as the AI minister, will be cutting the cheques, right? You’re going to go through more traditional mechanisms.
Yeah. We have to structure these things. Canadians want accountability.
We will be working, obviously, with partners, and the structures—we’ll be transparent about it.
Let’s talk about compute. How much of the Compute Access Fund goes to international hyperscalers? Do you foresee any mechanism to turn more of that capital toward Canadian companies?
The Compute Access Fund, was, candidly, designed by the last government. It was designed to give a 50 percent subsidy if you’re using an international provider, and 67 percent if you’re using Canadian.
The department told me we’ll probably get a few hundred applications, and it was just a firehose. SMEs genuinely need access to compute. The diligence process took a long time. We recognize that this is a really important part of the innovation supply chain, and we’ve learned some lessons.
Every company is probably trying to get access to this new technology.
And that’s good! Now we’ve realized, OK, demand is high, we’ve got to process these faster.
First order of principle: let’s make sure it’s not general admission anymore, that we have a better, more orderly fashion. Second is that we want this to incentivize Canadian compute. You will see a scoping of this to make sure that there’s an incentive to use Canadian companies.
RELATED: Nokia’s Ottawa expansion tests Canada’s definition of “sovereign AI”
But, as I like to say, sovereignty is not solitude. We have companies that we want to invest here. Foreign direct investment in Canada has gone up dramatically this year, but Canadian infrastructure, Canadian companies, that’s all first-order priority.
Anthropic was recently told by the US government to restrict foreign access to its two most powerful models. This came shortly after the Canadian government was granted access. [Editor’s note: After this interview was conducted, Anthropic was granted permission to reopen foreign access to its models.]
How is the government ensuring Canada has access to frontier models? Is there a plan in place to regain access to Mythos, or does it reinforce the need for sovereign alternatives, like those built by Cohere?
You gotta do two things at once. Let me reassure Canadians: we are in close contact with the US government all the time.
We are making sure our banks and our critical infrastructure are protected. Our Canadian Cyber Security Institute works very closely with US government security. That said, what’s happened with Mythos, what’s happened with Fable, for some was a clarification moment.
Sovereignty is not solitude. But sovereignty is choice. Sovereignty means that your most important information, data, is not dependent, is free from the coercion of others.
Cohere’s Nick Frosst likes to define sovereignty in terms of access and agency.
That’s right. Choice, access, agency. So, what does that actually mean, practically? We all want to use the best models in the world and the best tools, and we’re not trying to cut people off. So, if good models are coming from the US or France or Japan, Canadians want options.
You do need sovereign control over certain things—that’s why Canada is one of the four countries with a large language model. The US, China, and France have one, and we have Cohere. We have a huge competitive advantage here, and it’s a strategic advantage. Which is why we did an MOU with Cohere, and a sovereign tech alliance with Germany, which led to the Cohere Aleph Alpha merger. Which, as The New York Times said, gives an alternative to Silicon Valley models. Doesn’t mean it has to replace them, but it’s an alternative.
It’s an interesting diplomacy and business development combo.
It’s strategic autonomy.
And this is just the beginning. I will just say: watch this space.
Back to the strategy, because you’re going to hold me accountable for this: what are we doing on sovereign control? We are making sure that we have strategically strategic autonomy. We are going to do that through investments; we are going to do that through alliances. We are using our critical parts of the chain where we have competitive advantages.
We’ve got to build this infrastructure here. That’s why we’re doubling the electricity grid by 2050. That’s why we have a critical mineral strategy. We invested in, for example, digitizing the core samples library, because this is key for critical mineral development. So we are building up sovereign capabilities and key leverage points in the supply chain to ensure that we have choices in the future.
A lot of tech-related legislation has been introduced recently. Can we do a bit of a lightning round on the bills? Let’s start with C-36, the Protecting Privacy and Consumer Data Act. You’re creating a super regulator. Why not just give the current privacy commissioner more power?
OK, there are different bills, so it gets complicated. There’s a bill called the Protecting Victims Act, that makes it illegal to share non-consensual sexualized imagery. In the Strong and Free Elections Act, we protect our elections from deepfakes. Then we launched the AI strategy. Then we launched the Safe Social Media Act, which restricts access to social media for kids.
So when we regulate social media, we’ve never had a regulator for that. Other countries do, so we needed to set up a new regulator anyway. We believe that since we’re setting that up, it makes more efficient sense to simply set up a regulator, a commissioner that will do the Safe Social Media, but also have a commissioner specifically there on privacy that will have real enforcement powers on how businesses treat information.
Sovereignty is not solitude. But sovereignty is choice. Sovereignty means that your most important information, data, is not dependent, is free from the coercion of others.
I will say this: the Privacy Commissioner himself, who’s an independent officer of Parliament, when we tabled our bill on privacy, he put out a statement saying he was very happy that this builds on his recommendations to treat privacy as a fundamental right. So I get the criticism, but even the Privacy Commissioner is supportive of these measures.
The privacy commissioner remains, and will still regulate the government. We’re not trying to build a big bureaucracy. This is protecting privacy and consumer data, but we’re making sure there’s not odious red tape on small business.
When it comes to the Safe Social Media Act, two things we’ve heard here is that, first, trying to keep kids off the internet doesn’t often work. The other component of this is that if you set up an ID verification process, it now requires maybe all Canadians to identify themselves online. Why not attack the algorithms and the social media companies directly through regulation?
I think the Canadian solution is a little different than the ban, no ban. Taylor Owen wrote a pretty great piece on it. [Editor’s note: Taylor Owen, the Beaverbrook Chair in Media, Ethics and Communication at McGill University, wrote for the Globe on how Canada’s legislation, unlike some other nations, offers platforms a pathway to allow youth to use their products if they can demonstrate safety.]
If we have safety by design, if there are things that are useful for education, if they’re safe, the regulator will say, there is a use for this. Think about this like driving. It was like, oh, some cars are dangerous; they blow up, or you need seat belts. We started designing cars that were safer, and mortality rates fell. No one was saying we should ban cars. We manage traffic every day for lots of reasons, and one of them is safety by design.
The driving example is perfect, because it speaks to how difficult it is to regulate the internet, because if you’re a 16-year-old and you haven’t passed your driving test yet, you just won’t have access to a vehicle. There’s gonna be people that get around it. Like you and I probably had a beer in your teens—I admit nothing on the record—but this has happened. But you can’t get the perfect in the way of the good.
On the age verification, they’re hoovering up lots of information anyway. This is why we need to have responsible uses of how to treat information. There’s easy ways to verify that don’t give sensitive information, there’s ways around it. This is not going to be a perfect solution. But we’re learning lessons from Australia, we’re learning lessons from other jurisdictions. We’ve tried to design a Canadian version.
We’re working with the companies. We don’t want this to be adversarial. We’ve seen the harms. We don’t want to stifle innovation; we don’t want to stifle good uses, but look, you’ve got to be blind not to have seen what’s happened over the last 25 years with social media. We have been slow to act on this, but I also feel that poor policy can be more pernicious than no policy.
I have 1,000 more questions, but I know we’re out of time. I really appreciate you taking the time to dig into this.
It’s a huge topic. In the last year, we’ve set up a new ministry. We’ve set up a quantum program for our quantum champions. We’ve got a new national AI strategy. We’ve got two pieces of legislation on the trust side. We’ve got investments coming out, everything from supercomputers to data infrastructure to compute access. There are new challenges and new questions, so let’s keep the conversation going.
Watch the full conversation from The BetaKit Podcast here:
Feature image courtesy The BetaKit Podcast