THE STANDING SENATE COMMITTEE ON HUMAN RIGHTS
EVIDENCE
OTTAWA, Monday, April 13, 2026
The Standing Senate Committee on Human Rights met with videoconference this day at 4:01 p.m. [ET] to examine and report on the impact of artificial intelligence on human rights and economic security in Canada, especially in relation to vulnerable groups and the international human right to work; and, in camera, for consideration of a draft agenda (future business).
Senator Paulette Senior (Chair) in the chair.
[English]
The Chair: Good afternoon, everyone. Welcome back.
I would like to begin by acknowledging that the land on which we gather this afternoon is the ancestral and unceded territory of the Anishinaabe Algonquin Nation. My name is Paulette Senior, and I am a senator from Ontario and chair of this committee.
I now invite my colleagues to introduce themselves.
Senator Robinson: Mary Robinson, representing Prince Edward Island.
Senator Karetak-Lindell: Nancy Karetak-Lindell, Nunavut.
Senator McCallum: Mary Jane McCallum, Treaty 10, Manitoba region.
Senator K. Wells: Kristopher Wells, Alberta, Treaty 6 territory.
Senator Arnold: Dawn Arnold, New Brunswick.
Senator Lewis: Todd Lewis, Saskatchewan.
The Chair: Welcome, senators, and welcome to all those who are following our deliberations today.
Our committee will be continuing its study on the impact of artificial intelligence on human rights and economic security in Canada, especially in relation to vulnerable groups and the international human right to work.
This afternoon, we will have two panels. In each panel, we will hear from the witnesses, and then the senators will participate in a question-and-answer session.
I will now introduce our first witnesses, who have been asked to make a five-minute opening statement. With us, by video conference, from Arteria AI Inc. is Shelby Austin, Chief Executive Officer and Co-Founder. Also joining us by video conference from The Dais at Toronto Metropolitan University are Mahtab Laghaei, Policy Analyst; and Viet Vu, Manager of Economic Research.
I will now invite Ms. Laghaei to make her presentation.
Mahtab Laghaei, Policy Analyst, The Dais at Toronto Metropolitan University: Good afternoon, Madam Chair, Madam Deputy Chair and honourable senators.
My name is Mahtab Laghaei, and I’m a policy analyst at The Dais, based in Toronto Metropolitan University. I’m joined by my colleague Viet Vu, Manager of Economic Research. The Dais is a public policy and leadership think tank where we focus on the intersection between technology and our economic and democratic systems.
Recently, we released a series of studies examining the impact of generative AI technology on jobs in specific sectors — public, financial and creative — adding to a decade of expertise we’ve developed in AI.
What our research shows, using job posting information in Canada, is that while AI will not cause mass unemployment, it may still cause complex disruptions unevenly across the economy.
This is good news, insofar as we use this research to prepare Canadians. This means making sure that policies support the right kinds of skills building early on to allow Canadians to find meaningful and sustainable employment and succeed in their roles in the age of AI.
I am also here today to talk to you about one particularly vulnerable group, which I am also a part of: young people. It feels as though young people have drawn the short end of the stick, coming of age just as economic instability takes hold. Many have attributed this crisis to AI being used to replace entry-level roles. But, as my colleague Viet wrote for The Globe and Mail, AI has not yet reduced the availability of junior roles.
Though this has not happened yet, we believe that because of AI, the tasks traditionally included as part of entry-level roles will change. As a result, employers will seek out different skills in early-career professionals. As AI is used as a tool across jobs, instead of expecting technical acumen, employers will look for judgment, a critical eye and a knack for AI experimentation.
Take the role of a junior personal finance adviser working in a sector our research finds is highly exposed to AI. Before AI, a junior adviser would draw product recommendations for clients on their own; today, AI systems use data collected on clients in the past and generate product recommendations to the adviser in an instant.
The adviser’s role has not disappeared. Neither has the desire for clients to connect with real humans. However, the tasks in the role have shifted, and the adviser is valued for their judgment and human skills to provide the best service for their client.
As a result, we need to prepare youth for this shift and update policies that shape their journey from education to employment to reflect changing skill demands.
First, we recommend that the government support the development of a national strategy for K to 12 AI literacy. AI literacy programs that are tied to a curriculum can ensure students develop an understanding of how generative AI works as well as its limitations.
Second, the government should continue funding student placements for internships and work-integrated learning opportunities through extending the Student Work Placement Program.
And, finally, when making any decisions on the future of work, the government must base policies on evidence, not the grand promises made by top AI companies. Their bottom line is to release messaging that helps pad the pockets of their shareholders rather than deliver a public good. Establishing capacity for generating and sharing ongoing labour market intelligence about AI’s impact will benefit Canadians.
Though I’ve come here to discuss AI and how it could impact the human right to work, the reality is that the future of employment for young people depends on strong support systems beyond AI skills readiness. This is only one piece of the puzzle.
As lawmakers, I urge you to look at the bigger picture of what it takes to build a society in which young people can thrive.
I thank you again for inviting me to be here, and I welcome any questions you may have.
The Chair: Ms. Laghaei, thank you very much for your presentation.
Shelby Austin, Chief Executive Officer and Co-Founder, Arteria AI Inc.: Okay. Good afternoon, honourable senators, and thank you for the opportunity to appear before you today.
Canada stands at an inflection point in the AI era. As a long-time founder and entrepreneur, I’ve had the privilege of building companies and creating many jobs for Canadians.
From my seat as co-founder and CEO of Arteria AI, I see how profoundly artificial intelligence is reshaping every part of our economy — faster than I ever imagined possible. I have always believed that AI would change the world, but now that change is a certainty and is not on the horizon; it is here.
My hope today is to share just a few observations and then turn to any questions you might have.
First, AI is not just another technology trend; it’s transforming the way we work, live and compete. Around the world, the United States and China are spending at an extraordinary scale to secure leadership in this domain. For obvious reasons, Canada cannot compete dollar for dollar — nor can we single-handedly set the global safety agenda. What we can do is focus on the inputs that drive outcomes: compute and, by extension, energy, talent and capital. With thoughtful policy, those enablers will accelerate adoption, fuel productivity and help ensure Canadians capture meaningful economic benefits from the AI systems we build and export.
Let’s start with talent. Canada remains a global leader in AI research. We punch well above our weight, but our challenge is retention. We need tax and immigration frameworks that make it easier for innovators to stay, build and grow here. This is not about protecting domestic talent from leaving; it’s about giving them every reason to stay.
Next are regulation and policy. Critical AI workloads — especially those tied to national security — should operate on Canadian-controlled infrastructure, with clear data residency and encryption requirements. Just as importantly, our researchers and businesses need affordable access to compute power.
Finally — and most relevant to us builders, I would suspect — today, government investment vehicles and grant programs just are not keeping pace or understanding the real challenges of scaling companies.
On legislation, Canada should and must align with major jurisdictions, such that our firms are not disadvantaged, but should also avoid overreach. Efforts should be proportionate to risk and aimed at concrete harms like fraud and security — not theoretical scenarios that stall innovation without public gain. Canada should be known as a place where speed to market matters, backed by sensible governance that keeps the engine of innovation running safely.
Finally, I will speak on employment. The greatest risk from AI is not automation itself but failing to capture the productivity it unlocks. We are already seeing record levels of technology-linked job displacement.
We must act now to design retraining and redeployment programs that move people into new AI-driven roles quickly and effectively. If productivity gains take time to materialize or simply don’t arrive, our social support systems must be ready to bridge that transition.
Honourable senators, this moment is not only a challenge but a generational opportunity. Canada can carve out a position of real strength as a nation that scales innovation efficiently, invests confidently and equips its people to thrive in the next industrial era.
The choices we make today will define whether we are buyers or builders in the AI economy. Let’s choose to build and ensure that the benefits of this transformation are felt across every part of our country.
Thank you. I’d be pleased to answer your questions.
The Chair: Thank you for your presentations. We will now proceed to questions from senators. Colleagues, you each have five minutes for both your questions and the responses.
Senator Arnold: Thank you to our witnesses for being here. It’s nice to see you again, Mr. Vu. It is déjà vu.
Viet Vu, Manager, Economic Research, The Dais at Toronto Metropolitan University: It is nice to see you, senator.
Senator Arnold: My first question is for Shelby, and it concerns Canadian-controlled infrastructure. I’m curious if you’ve put any thought into what that could look like.
Ms. Austin: Yes, there are many efforts going on right now, senator. Thank you so much for the question.
When we look at what data sovereignty means, we need to have enough computing power for companies like ours; for Canadians, in general; and for researchers, in particular, to be able to exist and compete on an ongoing basis.
Then, when it comes to looking at more specific policies, we really need to focus on where we need them for security. That’s looking specifically at where we need data sovereignty infrastructure. For example, I don’t think it will be possible, especially for many of us who operate multinationally, to mandate that every company keep its data within the four walls of our country. That would likely make us uncompetitive. I do, however, think that where we have things that touch national security or health interests, perhaps, those specific data sources and/or broader ecosystems should be very carefully monitored to at least make sure they are operated within the country and with companies committed to keeping our data safe.
Senator Arnold: Thank you for that.
My next question ties in the presentation from The Dais, as well. You were talking about youth and need. I wanted to tie it into what they had said about education. What would you recommend to young people today who are looking at higher education? What would you suggest they study?
Ms. Laghaei: Thank you so much for that question, senator.
We actually recently had two round tables, one in Vancouver and one in Toronto, with folks across sectors — industry leaders — and we asked them a similar question. One of their main recommendations was to seek out opportunities that are interdisciplinary. Therefore, if you are pursuing an engineering degree, take some liberal arts classes, as well. What they want to see among their entry-level workers is a holistic profile. Some of the human skills that I discussed are critical thinking and judgment — ones that are cultivated and fostered in liberal arts and humanities areas.
That is one of the main recommendations that I’ve heard from industry, and I agree, as well, based on my personal experiences.
Senator Arnold: Thank you.
Senator K. Wells: My question is for both groups that are with us today. I’m going to ask you to do some homework for me.
This week, we have Evan Solomon, the Minister of Artificial Intelligence and Digital Innovation, coming to speak at Senate ministerial Question Period.
Given your work and backgrounds, what key questions would you want to ask the minister? In other words, what are your most pressing concerns for this minister and the federal government?
Mr. Vu: I’m happy to go first. Thank you, senator, for the question.
My question to the minister would be fairly simple. As much as Canada is a great country, we do not have unlimited resources in order to dedicate all the funding that would be necessary for any stream of programs, so strategic choices matter. As I think Shelby mentioned initially, the idea of sovereignty goes beyond owning every step of an AI’s value chain. Therefore, I think it matters, in terms of the new Pan-Canadian Artificial Intelligence Strategy the government has been working on, what the strategic choices look like. Colleagues of mine, Sean Mullin and Jaxson Khan at the Munk School of Global Affairs and Public Policy, recently published a paper that lays out the framework on how to prioritize between different sovereignty concerns, so the question of priorities is certainly top of mind for me.
Senator K. Wells: Thank you.
Ms. Austin: Thanks so much for the question, senator. Just to add to that, first, I think Minister Solomon and his working group have done an incredible job of engaging industry. It’s quite different from anything we have seen before. So, for starters and for whatever it’s worth from our perspective in industry, it’s a very credible effort. It’s trying to speak to Canadian builders and all facets of the ecosystem — really trying to touch upon that. Many of us have felt very heard throughout that effort, so that’s fantastic.
I would share a question around balance: Are we really looking at things through a prosperity agenda? Should we take a more balanced approach in terms of trying to balance — and hopefully you took this from my remarks — a prosperity agenda with a resource agenda? I think it’s really important that we look at very sensible balances.
Canada, given how it’s positioned in the world, isn’t going to put a national ban on AI — at least, I don’t think so. Similarly, I don’t think we can demand 100% sovereignty.
So, how do we be nuanced in terms of the dialogue we’re having with Canadians? It’s important that Canadians are able to converse deeply, especially for this next phase. I think there’s certainly a feeling from industry that Canadians don’t know what’s coming in terms of significant impacts to unemployment and to the technology industry. There’s a feeling that it’s a little understated.
Plans on how we’re going to educate Canada such that we can have a nuanced national discourse would be really interesting.
Senator K. Wells: Thank you for the great answers.
Senator Robinson: Ms. Austin, I wanted to go back: You said in your remarks that we need to give people a reason to stay. I was wondering if you could give us some examples of ways we can encourage people to stay.
Ms. Austin: I think we keep having false starts; we’ll start down a capital gains route and then take a gigantic left turn and/or something else. So, we really need to be thoughtful about how we’re reflecting to the builders of the country what our outcomes are. If we start with the outcomes we want to create, which are hopefully massive, amazing Canadian companies, we can, then, work backward from there and figure out the incentives.
We should have a tax policy that is reflective such that, if you can win big here, we’re going to celebrate your success and we will be super happy for you, which will be reflected in the spoils of what you’re building. That would be wonderful.
Similarly, we’ve had start-stop on immigration policy in terms of how we’re letting people back and where we’re going.
We need to be really mindful that we’re making a builder-friendly environment, because we say we want builders here — all kinds of builders, like me and a lot of other people who don’t look like the builders we have today. We need to make sure we have the right infrastructure. Similarly — and it’s certainly coming — we’re starting to see a transformation of Scientific Research and Experimental Development, or SR&ED, and other policies.
However, as somebody who has been building in Canada for a while, I would note those programs don’t always reflect our needs. We have a very strong early innovation ecosystem here. Then we have a very strong latest-stage ecosystem here, but we have critical gaps, both at the early stages, which we are long past but where, if you’re brand new, you get almost nothing; then, once you get a little bit bigger, there are some great policies and investment incentives to be here. Then, as you get a bit bigger — which is probably where Arteria falls these days — there’s almost nothing: You find yourself in a really interesting moment with those policies.
So, we need to be mindful about the stages. The Canadian government has obviously done a great job in terms of trying to invest in that innovation ecosystem. There is a lot of focus and effort being put there, which is unbelievable; we just need to make sure we are looking at it at every stage. We often focus on early stages and the latest stages, but the folks in the mushy middle are the ones who need the most help.
Senator Robinson: Super. That’s a great answer.
I think I heard you say that capital gains issues are there, as well. Am I correct?
Ms. Austin: Yes, last year, obviously, the drama on what we were doing with capital gains was quite concerning. When it took some time to dissipate, that was also quite concerning. Every entrepreneur had to ask, “Is it too much work to be here?” Were we going to drive away foreign investment in our companies if we continued to exist here, especially because many of us operate in multiple jurisdictions?
I think we want to ensure that those of us who so proudly and principally focus on building in Canada are encouraged to do so. We’re really so committed and so loudly banging the drum for Canada, and we just hope that we get to continue to have those options and that we’re not making choices as a country that would make it very difficult to choose to be here.
Senator Robinson: Thank you.
Senator McCallum: Thank you for your presentations.
In Canada, AI development and applications fall under both federal and provincial jurisdictions. For example, jurisdiction over privacy, data protection, health and human rights is shared between federal and provincial governments. Competition law and intellectual property law are within federal jurisdiction, and consumer protection law and property law are within provincial jurisdiction.
When we had presenters here, many were concerned about how AI may threaten to erode ethics, interdisciplinary skills, human rights and equity. How do you envision these governments working together on how they can best protect Canadians?
Ms. Laghaei: I can go first.
We have been talking a lot about sovereignty in the context of the strength of Canada and our strength in AI, but we have not talked enough about sovereignty in the context of Canadian values and ensuring that these AI systems have the values that we recognize and that are important to us embedded within them. That is something that both the federal government and the provinces can work on.
Something I want to introduce is also the idea of public AI and not only helping founders and companies within the country, but maybe there is an element for the government itself to invest in either large language models, or LLMs, or infrastructure where we can control those values and we have control and accountability within our AI systems.
That’s something that I definitely think requires federal-provincial coordination.
Ms. Austin: Just to build on that, first of all, for the first time in so long, Team Ontario and Team Canada are working together quite well on the AI agenda, so major kudos on that.
In terms of the bias discussion, we need to be a little wary here, because most of the major foundation or frontier models being used will not be developed by Canadians and are not being developed by Canadians. As a result, the truth is that we don’t have a lot of impact or control over what goes into those biases.
We should be careful, in terms of the regulatory agenda, to make sure that we’re not putting more burden on Canadian models than we would on others. Obviously, we want models that are fair, equitable, thoughtful and recognize all sorts of equities and then balance them appropriately, but I would be very nervous if we put additional requirements on Canadian-built models than on global-built models, because those will certainly not be impactful and may harm Canadian prosperity — and we’re certainly in the midst of a productivity crisis.
I hope it is okay that I’m balancing the discussion by saying that, obviously, removing bias from models is super important, but it needs to be done with a really clear understanding of what we’re trying to achieve and ensuring that we are also remaining on parity with global standards, which is the most important thing for any AI effort here.
Mr. Vu: Yes. To add to my colleagues’ comments, I believe that the role of standards will be important.
Take the example of interprovincial trade, the regulatory differences between provinces and the approach that we have taken to solve it. The federal government is, essentially, a standard setter that sets a minimum baseline of what needs to be met in order for trade between provinces to take place. Then, depending on specific pieces of provincial jurisdiction, each province may have to add further requirements on top but never be below that standard. Should they choose, as a bilateral agreement between provinces — as in the example of interprovincial trade setting — they can rely on federal standards for their provincial solutions. There is a huge role for standards.
Second, I believe enforcement clarity helps here. If we take our example and some of the recommendations that we have put forward in the design of an online harms act that will, hopefully, be considered by this body sometime soon, we see that many overlapping jurisdictions, even from the enforcement of online spaces, still require a single coordinator in order to facilitate the work between — let’s say — the privacy commissioner in investigating, perhaps, a privacy violation resulting through online interactions. So we think that some form of dedicated coordinating body to ensure that there is clarity on who is enforcing the regulation and who is enforcing the legislation also matters here.
The Chair: Thank you for your responses. I’m going to attempt to pose a question or two as well, because I really appreciated your presentations — particularly the work that you talked about, Ms. Laghaei, in terms of what Toronto Metropolitan University, or TMU, has done with the research and what it has uncovered. It doesn’t seem to present a doomsday sort of result, which I think is hopeful. That’s great.
I want to build on what I think Senator McCallum was getting at because we certainly heard from previous panels with respect to some of the concerns. In your presentations, the concerns seem less serious, if I can put it that way, and are more to do with how Canada approaches standards or regulations to ensure — as you said, Ms. Austin — that it doesn’t get in the way of innovation.
I am looking for some comments with respect to how we actually make sure, as we are developing and putting in place whatever the guardrails are, that we’re not leaving people behind — particularly young people and those who are most vulnerable — while also supporting opportunities that could lie ahead via AI.
Perhaps Ms. Laghaei could start first.
Ms. Laghaei: Sure. Before I pass it on to my colleague Viet, who I think can actually speak on this a bit more, I want to propose an opportunity that we have with the methodology — that we look at jobs and how likely they are to have tasks that are automated by AI.
When considering vulnerable groups, our research could be used to look at labour force representation among women or Indigenous communities to also understand whether their tasks and roles are more likely to be automated or to go in that direction so that we are better prepared.
I’ll let Viet also continue on that.
Mr. Vu: Thank you, Mahtab.
Senator, thank you for that question. I believe that a distinction should be made between a short-term fluctuation and, essentially, a long-term structural change in the labour market.
If this technology is being disseminated in the economy at a rapid pace, that is a problem. But what we see is that the pace of adoption of this technology continues to generally mirror those of other technologies. About 12.2% of Canadian businesses, by the end of last year, had adopted generative AI or AI in general in their operation.
Obviously, many more Canadians have tried out the tool, but maybe they have not adopted it in their day-to-day lives. If that’s the case, then we have policies that are already in place that can support a minor sort of labour market friction. The automatic stabilizers we have, such as Employment Insurance, come into play.
But I do understand, especially for youth, that there is a huge amount of anxiety, especially in the past year, as they have struggled so much to find a job right after graduation. To those folks, I will say to not despair because the market will likely recover. It is about ensuring that whatever skill you are developing right now is not responding to the needs of today but to the needs of tomorrow.
And they have great resources both online and in person, perhaps through a mentor, as well as through many of the incredible continuing education programs that we ourselves sometimes offer in order to ensure that their skills remain relevant.
The Chair: Thank you.
Ms. Austin, I have a related question for you, and you may have something to say. I have one minute.
Would you agree, based on what Ms. Laghaei said, that it is a humanities sort of education along with, let’s say, hard sciences that would be important? Is that what the industry is looking for?
Ms. Austin: First, I think that’s a great point. The humanities are certainly going to be critical going forward. It’s our humanity that will guide us, and certainly critical thinking, curiosity and other skill sets of that ilk will be critically important for this next wave.
I would go a little further than my friends here in terms of what they think about the unemployment crisis. Certainly, Goldman Sachs predicts 300 million jobs will dissipate. I think that’s probably conservative. There is not a doomsday scenario. I think that’s a challenging phrasing and would like to think that we can take steps here that will hopefully put us on a slightly different path, which is why I have been focused on productivity and prosperity.
However, I think if technology turns out to be the example and not the exception, we will see a dramatic shift in work, and we wouldn’t have anticipated that even when that study was conducted at the end of last year.
I think if we look at the results and adoption after models Claude Opus 4.6 and GPT-5.3-Codex were released at the end of last year, we will see that adoption has skyrocketed tremendously.
As a result, we have quite a lot to think of when it comes to — and this was reflected in my remarks — bridging that transition to productivity. So if we can’t set the right policy to get us on the right track, we need to look at social safety nets in earnest.
I read the remarks of my fellow panellists from the previous sessions and think they went further in their views than I might in mine. Again, I think the nuance is incredibly important here. For the first time in the past quarter or two, I have a bit of discomfort when it comes to just how far I think unemployment will reach and how unprepared we are, as my friend shared, with respect to young people in particular.
I share those concerns and think we need to be thoughtful about where we go from here and what policies we’re writing to ensure we don’t get to some of those doomsday scenarios.
The Chair: Thank you very much. I appreciate the thoughtful response.
We will move to the second round.
Senator K. Wells: During some of our previous hearings, we heard from individuals who were quite concerned about AI superintelligence as a significant threat to not only human rights in general but to humanity. I’m wondering about your own thoughts or perspectives on that and if there’s something you feel that we should be doing to harness or control the rapid development of AI, for many of the reasons that you have already described.
Ms. Austin: I can start. Thank you for the question. I certainly cannot claim to be an expert in superintelligence. What I am most concerned about is the economic output and the financial crisis that are almost certainly coming our way in the next 24 to 36 months. I think it’s a shorter time horizon than people think. That is my greatest concern, more so than the sort of broader superintelligence concerns.
And I really do think we have to be focused — and I’m a broken record here — around job creation and building good policies. We know well how to do that. In terms of protectionist policies, I might be more open to them, honestly, if I thought they would be effective. However, considering where we sit geopolitically, I think it’s impractical to consider those strategies; certainly, we can discuss them.
As a result, we need to ensure that we are setting policy, again, on parity and ensuring that if Canada can’t be successful in being a safety leader — which is perhaps not sensible given the geopolitical landscape — we have to be focused on job creation, investment and ensuring that we capture GDP in this moment to minimize the unemployment challenges that are most certainly coming our way.
Mr. Vu: Senator, I believe the way that I’m going to start my answer is by recognizing that technology is developed because it does things better than humans can. If we develop a technology that performs worse than a human, then it will not be an economically useful technology.
So, just like that, artificial intelligence and generative AI do some things better than humans can, but I don’t think — and I believe my colleagues would agree — that framing this conversation around artificial superintelligence or artificial general intelligence will help us get to the correct policy solutions.
Let me give an example of a very recent piece of news. Regarding Anthropic’s new Mythos model — and the fact they have delayed the public release of this model because it was judged to cause a significantly higher risk from a cybersecurity standpoint — we know the cybersecurity risk. Canada has developed Canada’s National Cyber Security Strategy. That is a risk that we understand and can respond to. That has to do with the model’s capability in one specific domain that is of national interest.
So in talking about the threat of an ever-smarter algorithm, we have to start by defining what is strategically important to Canada in terms of defence, economic security, human rights and action — general policy areas that respond to those specific and strategic areas — so that when the technology comes along that can supersede it, we are well prepared to respond from a policy standpoint.
Senator K. Wells: Great. I will switch tracks a little with this next question. I want to dig deeper into what you talked about: the importance of K to 12 education, addressing AI and developing literacy tools. Can you drill down and talk a little more about what that would look like on the ground in classrooms? What are sort of the exact skills that you are hoping young people will develop through a formal educational process?
We can talk generically about critical thinking, but I am looking for the practicality of what that curriculum might look like. Given the fact that the curriculum is not a federal initiative but one from the provinces and territories, it may be hard to get unanimity across the country.
Ms. Laghaei: Thank you so much for that question, senator. I look at the idea of AI literacy in three ways. First, it is about preparing students; second, it’s about preparing educators and teachers; and, third, it’s about preparing the overall system, the school boards, for this moment.
I can give you two examples of AI literacy that I think help build that picture. So what are we talking about? At the kindergarten level, an AI literacy activity could be something that talks about our emotions, maybe a lesson plan. How do humans interact with one another? How does one thing that someone says incite another emotion in another human? Does a refrigerator also have an emotional reaction? No.
What does that lesson plan do? It is trying to teach the idea that generative AI or technologies are not sentient beings. That’s a very crucial preliminary idea that we need to build among youth. In my own experiences talking to people younger than me, they talk about generative AI as though there is another human on the other end.
These are really basic principles that we want to build early on.
On a Grade 12 level, the topic could be misinformation and disinformation, and how the accessibility of AI tools is impacting our democracy, civic infrastructure and some of our core values. There are a lot of ways to integrate AI literacy — not just within that specific class, but through math, language arts and history — in order to prepare students, and not just for working with AI but for communicating about AI with their families and using it on a day-to-day basis to improve their own livelihoods.
I hope that painted a good picture of what we’re talking about when it comes to AI literacy.
Obviously, yes, curriculum development is a provincial role, but there are so many federal programs that are critical in helping build digital literacy. One of those programs is CanCode, although over the past few years, CanCode has been specifically for digital literacy. There are a lot of organizations that are now teaching AI literacy. We are one organization that provides curriculum-backed AI literacy programming.
The federal government, in voicing that a national K to 12 strategy is a priority, can help build those programs and invest in them more so that, on local and regional levels, teachers, school boards and students can be more prepared.
The Chair: Thank you very much. We must move on, and I’m afraid I don’t have five minutes to spare. Therefore, it will be three minutes for each question from now on.
Senator Karetak-Lindell: I’m listening to all this, and, sometimes, I get quite confused. I think of what people are saying and how to apply it to my community, at that community level. I don’t see them connecting together.
We have talked about how there is no human element in AI, as much as some people think there is.
I’m trying to figure out how we can make some of these programs become more culturally sensitive. I know that AI provides opportunities in health, education and some of the services that we can’t always access in our Northern communities. However, if we have trouble with existing government programs in terms of understanding us as people and how we do things, how is AI going to understand us, our needs and our cultural ways of doing things?
I’m very worried that if we have more programs done through AI, the gap will grow even wider with respect to our ways of doing things being understood by, let’s say, government agencies using AI within their workplaces. I read an example of someone whose application was dealt with through AI, and it gave her a refusal because it didn’t look for the right things in terms of what jobs she might qualify for. I’m seeing the same thing with services being offered through AI instead of humans. We already have difficulty being understood by humans in the government system.
How would you deal with that kind of thing? Anyone can answer if they would like to.
Ms. Austin: Okay, I will give it a quick run.
First, thank you for that. Let me offer you two perspectives. Number one, the way government uses AI should be risk-tiered. Someone’s health or wellness shouldn’t necessarily be determined AI-to-AI because it can become quite challenging to get real nuance across, as you said.
From a community perspective, let me offer how I have rolled out AI within my own community: the folks with whom I work and live now. I will assure you that within this moment of agentic AI, there is more room for you to infuse your culture within the system than there ever has been before.
I would encourage you to get your community to play together with the intelligence, because you will likely find that you might be able to infuse a real sense of value within the agents themselves. As we move beyond frontier models into agentic models, I think you will find you will be able to convey, in much simpler ways, how to infuse your values, your thinking, your community and the nuances and information behind those into the system itself.
I offer that as a bit of hope there. What we’re seeing is really an ease of conveying what you want within the system. Therefore, rather than shying away from it, I would encourage you to take a different approach and play with it. I really think you will find that, if you are able to get into some of the models and build your own agents, you might actually see a really beautiful reflection of your own values and community within those models. You might actually take the opposite approach and find it is different than you thought and quite helpful for you to expand and sort of scale your community more broadly with those tools.
Senator Karetak-Lindell: Thank you.
Mr. Vu: Do I have time?
The Chair: No, I’m afraid not. I must move on, but the next senator might spare some time. We will see.
Senator Robinson: Okay. We could suggest that any further answers could be submitted in writing if we run out of time.
The Chair: Yes, if there is additional information, please feel free to share that with us.
Senator Robinson: I want to ask Ms. Austin this: What indicators do you see that suggest diffusion will help in 36 months? Please take a minute to talk about that.
Ms. Austin: First, I might recommend that everyone on this committee read Matt Shumer’s essay “Something Big Is Happening.” It was a viral piece that spread throughout the technology community. It really reflects what I consider my personal thoughts and many of my colleagues’ thoughts across the industry. Effectively, we are seeing 14,000 jobs gone from Amazon and 30,000 from Oracle. In our own organizations, in the middle of what the world is calling “the SaaSpocalypse,” you have to become fit as a fiddle really fast.
We were writing about 30% to 50% of our code before the holiday break, and we’re now writing, in the first instance, less than 10% of our code. That’s a dramatic shift. These are the leading indicators we see that really show us that something big is happening, as Matt Shumer would say.
I just can’t stress enough that technology is really different now than it was three months ago and even more different than it was six months ago. It is almost like The Matrix, where you can’t “unsee” once you are able to see. The more you can get close and spend time with people in the industry who are at the forefront of this and who employ tens and hundreds of thousands of people, I’m telling you it is completely different, and it is moving faster and every day. We all feel it —
Senator Robinson: I’m sorry, but I don’t want to lose all my time. I have one more question. I’m really interested in what you are saying, but I want more.
What sectors of the Canadian economy do you think are most vulnerable or, on the positive side, will see the greatest benefits to disruption? Coding is definitely on the bad side, but I’m thinking that agriculture might be on the good, for instance.
Ms. Austin: Almost everything could be on the good side. Let’s be positive for a second. Honestly, if we do this right and really transform our organizations, why shouldn’t we be on the positive side of this? No one had ever heard of a social media manager until a few years ago, as well.
Certainly, I think my friend would share the view that, in historical industrial revolutions, we have seen a hundred years of economic history that suggests we will create more jobs than we will lose. The challenge with this one is that we are trying to do an industrial revolution in 3 to 5 years instead of 60 to 100 years, which is what creates the challenge. For other industries, tech aside — I certainly hope tech is the exception, but I worry it is the example.
Senator Robinson: Thank you. I think my time is up.
The Chair: We have a minute, so I’m going to seize the opportunity. Regarding Senator Karetak-Lindell’s previous question, I’m trying to understand: In response to her, Ms. Austin, you mentioned that now is the time to actually play with the tools to get people familiar and comfortable. Do you see that window closing? Is that why now is the time?
Ms. Austin: Before, it was really hard for people to sit down and try and think, “How would I build a community agent that helps facilitate discussions within my community?” Or, “How can I make a community agent that might liaise with other people about our community?” That used to be a really difficult task.
Now we’re finding that there’s great ease and conversancy. With the technical literacy that my friends talked about and with AI literacy, I can’t state enough that if you play with the technology and really get into the technology and sit down with it — in my organization, it wasn’t until we cancelled everything and sat there and “vibe coded” together that we really understood what was happening in this moment.
I encourage everyone to really come at it from their own perspective and see where this technology fits because I think it can seem very different from far away than it does up close —
The Chair: So you’re not saying that there is a closing window? Do you just think that this is an opportunity?
Ms. Austin: No, I think that there are moments when AI is moving so quickly that we would, sort of, use the phrase corporately, “Get on the bus.” We think now is the time: Get on the bus before it leaves without you. I think there are real advantages to moving quickly.
The Chair: Thank you so much.
That brings us to the end of our time with you today. Thank you for your very thoughtful presentations. We really appreciate you being able to contribute to our study of this important topic today.
I will now introduce our second panel. Our witnesses have each been asked to make an opening statement of five minutes. These will be followed by questions and answers.
With us at the table, from the Canadian Union of Public Employees, we have Ms. Sarah Ryan, Acting Research Director. Thanks for being here. With us by video conference, from the C.D. Howe Institute, we have Ms. Rosalie Wyonch, Associate Director of Research. Finally, at the table with us, from the Prince Edward Island Federation of Agriculture, please welcome Mr. Donald Killorn, Executive Director. Thank you all for being here.
Sarah Ryan, Acting Research Director, Canadian Union of Public Employees: Thank you very much.
I will be focusing today on AI and the human rights of workers.
The Canadian Union of Public Employees, or CUPE, represents over 800,000 workers across Canada in health care, municipalities, education, libraries, airlines, public utilities, social services and child care.
It’s important to pay close attention to the impact of AI on workers, given the high level of control that employers have over workers and the workplace. This power imbalance means we need safeguards to protect workers’ human rights.
AI has risks for workers in four main areas: job loss and work restructuring; surveillance; management by algorithm; and bias and discrimination.
If we look at jobs, the International Labour Organization, or ILO, found that women-dominated occupations are twice as likely to be exposed to generative AI, and 16% of women-dominated jobs are at high risk of automation compared with 3% of male-dominated ones. This is because women are concentrated in clerical, administrative and business support roles, which are more prone to automation.
Workers don’t have equal access to training and upskilling. These opportunities are essential during tech transformations, yet research shows that workers in full-time permanent jobs are more likely to receive training than workers in part-time and precarious work. Many workers face barriers in accessing training because of multiple jobs, shift work, high costs and family responsibilities.
Digital technology can increase electronic monitoring of workers. A new group of tools called “bossware” uses unprecedented electronic surveillance. Employers can monitor keystrokes, listen to conversations and track employee movements, emotions or attention. Workers who are low income, younger, have disabilities or are Black or racialized are more likely to be monitored.
AI is reshaping the relationship between employers and workers. Algorithmic management systems can screen job applications, conduct interviews, rank and recommend candidates, assign tasks and allocate work, set schedules or staffing levels, evaluate performance or productivity, assess training needs and even trigger discipline. These algorithmic management systems can make decisions or recommendations that violate workers’ human rights.
Bias and discrimination may be baked into the data that feeds AI systems and the algorithm that processes data. Many algorithms are proprietary secrets and are not transparent to workers, their unions or even the employer.
Nearly 90% of companies worldwide use some form of AI to screen job candidates. These AI hiring tools can make biased and discriminatory decisions or recommendations. In the U.S., workers are suing, saying AI tools discriminate against them based on race, age and disability.
Not surprisingly, research shows that invasive workplace monitoring and algorithmic management can intensify work, increase stress and harm the psychological well-being of workers.
So how can we protect workers? It’s hard for individual workers to take on AI. We need systemic solutions that minimize risks to workers and put the onus on manufacturers, vendors and employers.
CUPE is calling for laws that prohibit employers from making significant decisions about workers based on or supported by output from algorithmic management systems. This includes decisions like hiring, promotions, discipline, wage setting and termination.
Employers should be required to disclose information about the AI systems’ training data and algorithms. Employers should have to conduct bias and discrimination audits before implementing algorithmic management systems, and AI developers, vendors or deployers must disclose any incidents or risks related to bias and discrimination.
We also need a strong social safety net with a strengthened Employment Insurance system.
AI systems can affect the livelihoods and human rights of workers. We need laws, regulations and public programs that protect workers.
Thank you very much.
The Chair: Thank you, Ms. Ryan.
Rosalie Wyonch, Associate Director of Research, C.D. Howe Institute: Thank you very much. I think my remarks will slightly contrast with those of my co-panellist. I’d like to frame AI as a technology that is already beginning to shape decisions and our labour market with potentially uneven consequences that are not fully understood. I would also like to highlight that it is a potential source of improved productivity that can contribute to higher GDP per capita and provide associated tax revenues for growing government programs.
Fundamentally, AI itself is neutral, but the data used to train it or the fine-tuning of the models can contribute to reinforcing, neutralizing or reversing existing biases.
My remarks will provide background on the Canadian AI adoption landscape, discuss the research on the effects of automation through the lens of vulnerable groups and explain how the use of AI can impact decision making with or without user awareness.
First, Canadian data shows that about 12% of businesses are currently using AI, and most report little change in employment following adoption. More than two thirds of businesses report either no plans to adopt AI or uncertainty around doing so, and 6 in 10 businesses think that AI investment is not important or relevant to their business. Comparing previous and planned use of AI shows adoption could be slowing in Canada, with only 2.3% of businesses planning to use AI this year that weren’t previously.
In terms of employment effects, among the firms that have used AI, 89.4% reported no related change in employment, and 70% of firms planning to adopt AI also expect no change in employment. Really, what we see is that there is minimal impact, at least so far, in businesses that have adopted AI, and that the vast majority of businesses in Canada don’t currently have plans to adopt it or don’t think that it is relevant to their business.
If the past is any guide, a continuation of gradual changes can be expected in the demand for skills in the labour force. This is a natural market reaction to technological change. There is unlikely to be a drastic shift in employment due to automation in the near future, although some industries and types of occupations will be more disrupted than others. It is very unlikely that employment in occupations that are susceptible to automation will be completely replaced by AI over the next few years. This process takes many years, and labour markets adapt dynamically over time.
Occupations that are high in abstract, complex decision-making skills or with a strong focus on creativity, critical thinking and interpersonal social skills still have a relatively low risk of being automated, and demand for those employees is likely to increase in the near and medium term.
My past analysis of automation susceptibility, looking at individual characteristics, indicates that Black and Indigenous Canadians are employed in occupations that are more highly susceptible to automation compared to the population average. It is likely that the relatively higher susceptibility is related to their worse average employment outcomes, more so than to the automating technologies themselves.
Men, women and immigrants, however, generally face a similar average risk from automation. Women are more likely to be in roles highly susceptible to automation, but it’s not a statistically significant difference in Canada.
Similarly, age and education levels are related to the susceptibility of automation. Younger people and those with less education are more susceptible.
Overall, these differences are not large enough, at least currently, to warrant targeted pre-emptive policies specifically to prevent technological unemployment for particular groups.
Turning to AI, bias and decision making, I will be drawing from a recent working paper that uses major large language models to evaluate 12,000 real job advertisements from the United States. The findings suggest that the major LLMs either inverted or erased job application callback biases when directed to make decisions. Gemini was neutral, while ChatGPT, Grok and Llama actually prioritized Black candidates over White ones compared to the benchmark, which is what we see with just humans or when the AI is instructed to prioritize historical accuracy.
This study shows that AI model producer choices in fine-tuning the models impact output results, trading off between accuracy and perceived harm reduction or harmlessness. While using different LLMs for applicant selection will vary between model and producer choices, all the models were equal to or improved on the status quo of human bias.
Similarly, the context of the instructions affects the results. As businesses implement AI automation, due diligence is required if the application is going to make decisions relevant to human-rights-protected categories. The relative bias of LLM models differs across them and will change with updates. However, all the models tested do at least as well as people, so I think that this could be a case where we shouldn’t make perfect the enemy of good. While there is potential for bias, so far, the evidence suggests that they could potentially be an improvement.
Also, most businesses in Canada don’t currently have plans to adopt AI. Those that do expect minimal impacts on labour. The disruption in the labour market is likely to be gradual but to affect some industries and occupations more than others. Particularly, we should pay attention to Black, Indigenous and younger people and those with lower educational attainment, as they are more likely to be employed in occupations that are highly susceptible.
Finally, the use of AI model selection does affect outputs relative to decision making, but it has just as much potential to reverse and neutralize bias as it does to enhance it. Ongoing due diligence by businesses and in the form of research is required to monitor these effects as they change over time. Thank you.
The Chair: Thank you, Ms. Wyonch.
Donald Killorn, Executive Director, Prince Edward Island Federation of Agriculture: Thank you, Madam Chair and honourable senators, for the invitation to join you today. My name is Donald Killorn. I’m the Executive Director of the Prince Edward Island Federation of Agriculture, or PEIFA.
Agriculture makes up 7% of the economy in Canada and almost one third of the economy in Prince Edward Island.
It’s expected that in 10 years, humans will make up a small fraction of all the intelligence on earth. At the P.E.I. Federation of Agriculture, we are currently using AI to build intelligence tools for farmers, ensuring that they control this new intelligence layer. Canadian farmers are managing $1 trillion of capital — land, equipment and inventory. They’re doing this using many different systems and sources of data. This data is difficult to synthesize and communicate. Artificial intelligence will help farmers know when to irrigate, when to apply fertilizer, when to plant and when to harvest. The reliability of the data will help farmers manage risk and improve their terms with lenders and insurers, each of whom will be grateful for the improvements to risk intelligence in agriculture.
At the P.E.I. Federation of Agriculture, we envision using the intelligence layer to unlock tokenized ownership, decentralized finance, parametric insurance and irrefutable traceability that confirms environmental goods and services, including soil carbon credits.
The median age of Canadian farmers is approaching 60. Tokenized ownership of land will assist with succession planning, facilitating fractional ownership structures to ensure that the next generation of farmers can enter the industry. This will combine with decentralized finance, which will bring lenders to farms in many small amounts.
The future of AI is a digital world of agents deploying capital in small increments across an infinite number of investments, seeking out reliable yields for capital that Canadian agriculture will deliver for them. Parametric insurance will provide direct and immediate payout of insurance contracts, and traceability will help satisfy food safety requirements and ensure that we can overcome the non-tariff barriers to trade that hamper Canadian agriculture in our efforts to expand markets for our producers.
Artificial intelligence will ensure farmers’ contributions to climate change mitigation and biodiversity are available to the marketplace. Attempting to sequester the 35 megatons of additional and permanent carbon that Canadian agricultural soils are capable of storing only happens with the help of this external intelligence layer.
Data collection, analysis and management create capital intelligence, and capital intelligence improves risk management. We have positioned our agricultural intel data platform as a leader in the effective governance of farmer data. Trust is the real product. Trust in governance will be critical in managing the impact of artificial intelligence, not just in agriculture but throughout the economy. Artificial intelligence is a force multiplier, not a replacement for expertise. It helps capital managers, like farmers, do more with less.
This will be true across the whole economy. We will need fewer people to accomplish the same amount of work. To navigate this disruption, governments will need to become more transparent managers of capital. To manage the risk of AI, I recommend Canada look to other jurisdictions for examples of managing governance, incorporating inclusive wealth into its economic reporting and ensuring that tax structures are designed to reduce inequality in the face of this disruption.
I point to Estonia as the leader of e-government. Their “only-once” principle, where citizens provide data only one time and then it’s used across various government systems, ensures transparency because citizens can see who has accessed their data. This algorithmic transparency increases the visibility of automated systems, placing a focus on trust. Trust is the product.
Canada must also implement a measure of inclusive wealth that equips us to make decisions about how to balance our capital between industrial capital, social capital and natural capital. The UN Environment Programme, in 2018, released an inclusive wealth report that showed that since 1992, the global stock of natural capital had declined by nearly 40%, while industrial capital had doubled and social capital had increased by only 13%.
In 2016, the International Institute for Sustainable Development, or IISD, said that Canada is better positioned than any other nation to show leadership on the measurement of inclusive wealth. Statistics Canada measures more of the components of comprehensive wealth than any other statistical agency. I echo the call for the federal government to fund Statistics Canada to carry out the needed changes to its program and begin regular reporting on comprehensive wealth. Doing so would better equip leaders to make balanced decisions.
Finally, we need tax system reform. Tax needs to shift from income to capital. A small tax on industrial capital is more equitable than a progressive income tax. French economist Thomas Piketty has shown that return on capital has been greater than the growth of GDP for the past 50 years, resulting in the worst inequality we’ve ever seen in the modern world.
To try to limit the impacts of this change on capital, I echo MaRS Discovery District’s call for Canada to adopt the U.K.’s enterprise investment tax credit models. These are designed to increase investment in seed and Series A companies. The schemes use three core tax incentives that Canada can replicate. They provide income tax credits to investors in innovative start-ups, they allow losses from start-up investments to be directly offset against income and, perhaps most importantly, all shares purchased in innovative start-ups through the schemes are exempt from capital gains tax.
The question of will we or won’t we have AI is like standing on the tracks when the train comes through. We are engineering intelligence. It will make up the bulk of the intelligence in the world in a short amount of time. We give it context. We teach it semantics. We apply a reasoning model. We get results.
At PEIFA, we’ve created intelligence for farmers. It runs on an ontology that farmers define. It is a force multiplier, not a replacement for expertise. Help us build expertise. Incentivize critical thinking and the education required to implement these systems. Remember that education has been, and continues to be, the most reliable investment a society can make. This disruption will not change that.
We need this body to lead the design of a governance protocol to measure and deploy capital according to the wishes of the electorate. We need a reborn democracy that can once again bring capital to heel and ensure that our country has the social and environmental capital needed to continue to support the growth of our wealth.
I will leave you with something that Mr. Marshall McLuhan, who saw in the lightbulb the question that we face here today, said: “First we build the tools, then they build us.”
I thank you for your time and look forward to your questions.
The Chair: Thank you, Mr. Killorn, for your presentation. We will now go to questions from senators.
Senator Robinson: I have a question for Mr. Killorn. I am wondering if you could give us an idea of whether you think AI will augment jobs in agriculture.
Mr. Killorn: I certainly think it’s good for farmers. I think that farmers are very quick to adopt automation technology because they struggle with bringing the labour they need to the farm.
In Prince Edward Island, we have seen tremendous capital investment in robot milkers and optical potato graders. I am sure that across the Prairies, we have seen similar investments in automation.
In the future, I think we will continue to see automation in-field; however, we will see an increase in high-paying jobs in agriculture. So the losses in the fieldwork will be offset by the tremendous opportunities to manage data on-farm and to design and deploy these robotics. Really, the innovation agenda for agriculture should lead to a significant increase in the economic impact of the sector in Canada.
Senator Robinson: So I have to keep training my brain, kind of like how you do AI, to differentiate between robotics and AI. When we look at the impact of AI and agriculture, my mind goes to how we can take data and process it at a much greater rate, and then how that will impact employment on-farm — you spoke a bit about that — and what it will mean for food production.
In the end, you talked about tokenized land ownership, aging farmers, looking at succession and all of these pieces; it is a very big picture. I’m wondering specifically about AI’s impact on a farmer’s ability to be profitable.
Mr. Killorn: Yes. It’s a very big picture. It’s a big value chain in agriculture. It’s a huge part of Canada. It’s a tremendous opportunity for growth.
Someone once asked me what we can do to impact profitability on-farm, and this is the first time I’ve felt as if we have an answer.
I’ve gone all over the world to discuss the importance of the governance of on-farm data, even before we saw the release of Claude Code. ChatGPT was here, but until Claude Code was released in January, the frontier models were just the next iteration of web search engines.
But we never let that dissuade us from focusing on governance, because once the data is malleable and easy to manage and analyze, all that matters is how we govern these systems.
Profitability for farmers is on the agenda now. We’re building inside the federation, and we see so many interested parties trying to get to the farmers and get their data because they know the value of it. As such, we’re grateful to be able to build for the farmers and ensure that they maintain control of this data and the ultimate intelligence layer that it facilitates.
Canada currently has the worst food inflation in the G7, which is pretty insane for a country that produces so much safe and healthy food. I believe artificial intelligence can help with that — can help with more efficient production of food and, ultimately, the profitability of Canada’s farmers.
The Chair: Thank you.
Senator McCallum: Thank you to the presenters.
When you look at AI models for farmers and that they will help with when to plant and to fertilize, basically, they become the farmer.
So farming is an area that is more than agriculture. I’m relating it to my ancestors and their trapping and fishing. It includes intergenerational knowledge transmission between and within the family; kinship building; and continuous, lifelong learning in understanding our world — that is, the seasons, the weather, environmental dangers, conservation and soil health.
What will be the future of humanity if AI becomes the ultimate farmer? Will farmers become kept people? What will they do?
Mr. Killorn: That’s a great question. I want to start by saying that the best lesson I ever received in sustainability was when Chief Akagi told me that we don’t pick all the mushrooms; we just pick the mushrooms around the edge. And the best example of data management I’ve ever seen is when Chief Akagi explained to me the story of the sunrise, and I realized he was actually telling me about the end of the ice age as experienced by the Wabanaki people.
A previous presenter, Shelby Austin, said that things are happening fast. We’re having an industrial revolution in five years. We are having industrial revolutions more and more frequently, and they are happening faster.
There are a range of outcomes right now; some are dystopian and some are utopian. We’re here today to talk about best outcomes for Canadians and our Indigenous neighbours and how we use the power of governance to achieve these outcomes. It has to begin today, putting groups and using governance at the centre of decision making around how we deploy these tools.
So, the story is just beginning to be told. We have capital-sucking monsters in the United States and China who are going to build $200 billion worth of data centres in the next year and consume 70% of all new grid connections. This thing is here, it is extremely disruptive and, as Canadians, we continue to be able to learn a lot from our Indigenous neighbours about how to plan for the future and how to adopt an abundant mindset.
People in technology want to espouse an abundance mindset to bring us closer to a utopian outcome, and our Indigenous neighbours in Canada are born with an abundance mindset; they are not trying to learn it in Silicon Valley.
Our Indigenous neighbours are incredibly important. I was thinking today, so much of what I’m trying to share is — we’re in a river and we can’t stop the river. Sometimes, we build structures in the river to improve flow and point the river where we want it to go. We need to do that with all the new wealth that we’re going to generate.
And that made me think, of course, of our agreements to paddle our own canoes. And we hope that you will help us prepare for this next phase of economic activity in Canada.
Senator McCallum: Do the others want to add to that?
Ms. Ryan: There is a good question there. You were mentioning the build-out of data centres and energy and environmental costs. I think that is a really important part of large language models: the energy that they require and the public and municipal water base that’s used in them.
That should not be lost when we’re talking about this because it affects future generations. It affects the land and our ability to achieve climate change goals as well.
A warming world will have all kinds of impacts on workers and the public and the most impact on folks who are equity deserving.
Senator McCallum: Thank you.
Senator Arnold: My head is exploding a bit with all of this information from three very different witnesses. As you just said, there is such a range of outcomes. We have heard from many different witnesses — I am actually on three different Senate committees that are doing AI studies, so there is a lot. The range is so wide with respect to what we’re hearing.
My question is to Rosalie. If only 12% of Canadians are using it and two thirds have no plans to use it, the data you use to tell your story isn’t all that concerning compared to some of the others that we have heard. Is this a risk for our country if we don’t embrace it more?
Ms. Wyonch: I would say it is a risk. Canada is still a world leader in AI. I want to say that nowhere in the world, actually, is — there is a really high proportion. The data is very uncertain. If you go to McKinsey, they will say it is 88%, but when you use the OECD — with more reliable, national, standardized statistics — adoption within businesses is actually quite low.
As individuals, Canadians are some of the highest-volume users of AI in the world, but our businesses are about average or falling slightly behind. In the global rankings, our ranking has actually been falling over time. So, our global competitiveness in the AI industry has been an area of concern — or should be — as we are sliding. If AI is the next general-purpose technology, its potential to grow GDP and improve living standards is at least as large as the potential for it to be detrimental.
So, I would caution against additional barriers to adoption and experimentation because this is very much a fast-growing, fast-moving industry.
I would say, yes, it will be a risk to Canada if we do miss the AI wave and it turns out to be the next general-purpose technology, but I would also caution against over-indexing on it, because all of this potential is just that: It is potential. There are the downsides. A data centre will have a short-term GDP impact in its construction, but it will have almost no ongoing impact after it is built and will support very few jobs. On the adoption side, however, if we get broad adoption throughout the economy, that has the potential to have real productivity and GDP impacts.
It is really about balance. We need to put guardrails and regulations to manage the large risks that we don’t want to happen in this country, but we want to keep them minimal so we don’t hamper development and can maintain global competitiveness.
Senator Arnold: Thank you for that. We have just heard what the impact on agriculture in P.E.I. is. In your research, and using your crystal ball, what are the largest opportunities ahead?
Ms. Wyonch: The largest opportunities are probably not yet known to anyone. As we just heard, Claude Code was launched in January, and it has completely changed what we think the potential could be. To perhaps illustrate why I think we simply don’t know yet, why businesses are going to have to adapt and why we will have to reorient business processes around this technology and reimagine things to really get that potential for growth — as an example of “we don’t know where it will be,” the first laser ever made was in Bell Labs. Bell, the telephone company, did not bother to patent it because they did not see any possible application of a laser in the telecom industry. Right now, AI is that laser. We know we have a powerful tool. It is very cool and “science-y,” but we don’t yet know exactly what we will do with it or where its greatest potential lies.
Senator Arnold: Thank you.
Senator K. Wells: I have a clarifying question based on some of the opening remarks where we heard very divergent statistics from two of our panellists. C.D. Howe Institute and CUPE, could you either send us the statistics and the research behind those statistics — these opinion polls or the peer-reviewed research you are drawing from — or, if we have a few minutes, share some of that and clarify for us?
Ms. Wyonch: I might be able to answer that one.
All the numbers I quoted in my opening remarks are mainly from C.D. Howe publications, and one of them was taken from a Rotman School of Management PhD student who had a really good presentation. I’m happy to share the research papers.
The main difference is that you can see a large proportion of labour potentially being susceptible to automation, but “susceptible” does not necessarily mean it will actually happen. For my results, I’m drawing from my own research, which I also compare to others’ research on susceptibility. I come up with similar numbers to those that were quoted in terms of susceptibility or potential risk.
However, when you actually pair that with how you see the labour market changing and adapting to these technologies over time, you see that it is much more gradual. Just because an occupation can be automated does not mean that it will be. I will draw another historic example: We have had ATMs, automated banking machines, since 1970. Yes, there are probably fewer tellers in banks, but there are still people at the desks when you go into a bank. That’s something we have been able to automate for almost 50 years at this point.
The difference is susceptibility —
Senator K. Wells: I will stop you there so we have enough time.
The bank I go to often doesn’t have tellers. It’s the bane of my existence, if it is even open or if I can even reach a person by phone or an AI bot. So, we may be losing some of that human contact already.
Regarding your opening statistics, please send the research behind those if you footnoted or referenced them in your remarks. That would be very helpful.
Now over to CUPE. Thank you.
Ms. Ryan: Thank you for the question.
In some ways, we’re saying similar things in terms of the statistics. There hasn’t been wide-scale private sector adoption. We don’t actually have the data on public sector adoption, though. I actually think public sector adoption might be further down the path in some sectors. In health care, we certainly see a lot of algorithms being used for a lot of different aspects in hospitals.
However, we don’t have the data; we have a lack of data right now.
I do have references. Also, the ILO numbers I referenced are about the possibility of automating. It is hard at this point to know where this is going. When I started looking at this, you had tech CEOs talking about 50% of white-collar jobs being eliminated. I would look at that now and say to take it with a grain of salt; it’s a sales tactic to a certain extent. I think folks have seen the limitations of the technology. Chatbots give false information; there are false citations and case law being generated.
Sometimes, implementation doesn’t go the way businesses expect, and the outcome sometimes does not result in something better. We look at public services most because we are a public sector union. Sometimes, however, the outcome is not great, such as a chatbot giving misinformation to folks about municipal services. We don’t want to see that; I don’t think any municipality wants to see that. Folks are being very cautious regarding the limitations in the technology.
We are at a point where it is difficult to tell. There is a lot of hype. The data we have is a bit old now if it is a year old, and it is hard to know, even in terms of private sector adoption. We know there are some big corporations that have definitely been enacting mass implementation of AI. It is hard to know where it will go. We have our micro-level information at CUPE in terms of our members, and we know that certain job classifications have been lost, such as medical transcription in hospitals; a lot of those jobs have been eliminated. Folks who wrote closed captioning at TV stations — a lot of those jobs have been eliminated.
In some cases, we have collective agreements that protect against job loss, so the individual worker does not lose their job, but the actual job is no longer and is performed by technology.
Senator K. Wells: I get it. It is sort of the Wild West, and it is clear to me we need more investment in the research — what is actually happening with AI — rather than speculation on what is happening and how rapidly this technology is shifting and changing.
We are trying to see what the baseline data out there is. If it is not there or it is not credible, that is something that maybe the federal government can start looking into, to develop that baseline data, particularly for those occupations that might be under the most immediate threat of job loss to AI and what that might mean for our economy. I think all of you made a strong case for women and marginalized communities who may be traditionally in these sectors, in entry-level jobs, because maybe they have not been given the opportunity or been promoted because of systemic discrimination around those higher-level senior management jobs. There could be a disproportionate impact in Canadian society.
Thank you all for your very informative presentations.
Ms. Wyonch: I was going to add that the data is out of date. The macro stats that I’m pulling from, if they are different than CUPE, it is a year out of date and StatCan only has two data points. That survey, the last issue, will be in August of this year. The 2025 Budget did have money for the TechStat system, and so I would say, since we are losing the only macro data point we have about business adoption and it is limited as is, I would very much urge the government to use the money budgeted in 2025 to get that TechStat surveillance up ASAP.
That is my little rant. You are already on it, but the sooner, the better.
Senator K. Wells: That is very helpful, because we will be looking for recommendations from our hearings that we can hopefully put forward to the government for response, if not adoption.
The Chair: Thank you so much. I think I have a couple quick questions. I’m curious about the algorithmic management systems. Ms. Ryan, is there anything else that you would like to say about the challengeability of that legally?
Mr. Killorn, you said something about AI as a multiplier, not a replacement, but then later talked about fewer workers being needed. Could you connect those dots for me?
Ms. Wyonch, the stats or surveys that you talked about, are they from the CEOs of companies specifically? The reason I ask is because we heard in a previous panel, I believe, that some CEOs have no idea what is coming but the people doing the work do. Could you just clarify that?
I will start with Ms. Ryan.
Ms. Ryan: Yes, that’s a very good question and a huge concern that we have with algorithmic management systems. Often, the actual algorithm is proprietary. You’ll hear references to the black box algorithm, and when that algorithm is making significant decisions about workers — the same applies to public service users. Do you get flagged for a fraud investigation? Is your application for a certain benefit automatically processed by an algorithm? How do you actually challenge that?
We have been talking with our lawyers at CUPE. Say, for example, something like that ends up going to an arbitrator, and arbitrators are saying, “We don’t really know how to handle this either because this is a proprietary algorithm.” It is an off-the-shelf tool that maybe an employer purchased. How did the algorithm make the decision that is affecting a worker’s rights?
There is a fundamental problem there in terms of transparency, accountability and human rights. How do we know it is discriminatory? How do you prove that? It is a really high bar, and that’s where we’re calling for real guardrails around the use of this technology in decisions that affect workers and public service users, because we also call for that in terms of public service automated decision making.
Mr. Killorn: It is a good question. In our experience working with it, trying to put our finger on what it does for us, that is where I get that idea that it is a multiplier, not a replacement. It feels like the Iron Man suit. Someone has to be controlling it — someone who has a sense of what they want to accomplish and how to do it. There was this burst of coding that came out of the emergence of Claude Code because English was now a coding language for the first time.
But at the federation, we’re not “vibe coding” tools for farmers. We have a software engineer who is a PhD-level software engineer and a network architect who is a PhD-level network architect, and they are experienced in the private sector and know what to look for. This thing does not replace expertise. It is not going to replace the farmers’ expertise any more than the tractor or the plow did. You still need that expertise to execute with these tools. As exciting as it might be to have an idea and put it out on the App Store the next day, that’s not a product that is going to be secure and able to withstand real commercial use.
The Chair: The labour reduction would be the people — who would they be?
Mr. Killorn: That’s a great question. We have sort of thrown a bit of dispersion on that idea that 40% of the white-collar workforce is going to be disrupted by this.
It takes time for this technology to diffuse through big organizations because their data is not ready for it. But there is no question that there are sectors of the economy — coding, legal and accounting — where if your sector is codified and your expertise is in knowing and interpreting that code, you are going to see a reduction in your head count in your department. Org charts are going to become flat. The hierarchical org chart that we have come to know is going to be replaced by a flat org chart that interacts with the source of truth. A human can manage maybe eight people beneath them. The idea is that hierarchy is going away, and a person, the force multiplier, the CEO of the company, will be able to manage the whole company because they are going to create this single source of truth and AI is going to help them with their metrics and report back to the CEO — data collection, management and analysis.
If a robot can do your job, we’re getting to the point where a robot is going to do your job, and I think the big question is around this loss of jobs. We need to understand — and I don’t have the answer — if they will also be replaced with more and better jobs, as we have seen in advanced economies previously.
The Chair: Thank you.
Ms. Wyonch: Yes. In terms of the data sources that we’re drawing from, the high proportion of businesses that say that they have no plans to adopt AI or investment in it is irrelevant to their business is drawn directly from the Canadian Survey on Business Conditions. The target population is all live establishments on the business register that have employees. That said, I don’t know which of those employees or the CEO would actually respond to the survey. It would be whoever StatCan gets on the phone or is delegated to fill it out. For the other research on susceptibility versus whether you’re actually going to be automated and risk in the labour market, that was a much more quantitatively intensive exercise using the monthly Labour Force Survey data, the census and the O*NET skills occupation data set from the States, crosswalked using a Statistics Canada-developed crosswalk. I also used a Bayesian Gaussian support machine, which is a rudimentary type of the old school of AI, to do the statistical analysis. There is a methodology attached to the reference I will send. I sent it along with my remarks, and it is pretty data intensive, but in terms of where businesses are at, it is based on the two data points we have from Statistics Canada. That’s all we really have. So how certain they are and how much they know — there is the same noise as with any other survey.
The Chair: Thank you.
Senator Robinson: I’ll ask my two questions and maybe get some written responses.
Mr. Killorn, you said in your remarks that AI is a national security risk. I’m wondering about connecting that to food security because, as we know, a nation that cannot feed itself is not a secure nation. I’m wondering if you could give us an answer as to how AI integration will impact food pricing. Will it decrease the cost of food?
My other question, or statement, is this: If I say that AI integration in agriculture increases predictability for farmers — as I think you’ve heard — and that should, in itself, lead to improved profitability, which can then lead to youth retainment by making farming a viable career path for youth, I wonder if you would agree with that statement and if you might expand upon it. I’m sure we’re out of time.
The Chair: We started a bit late, so let’s see how far we get.
Mr. Killorn: Thank you so much. Those are all great questions.
This thing is going to be a boon for productivity. It’s going to generate a tremendous amount of wealth. The leaders of these technology movements will assure us that wealth is not a zero-sum game, but it also means that inequality shouldn’t be the worst it’s ever been.
Egalitarianism reached a peak after World War II. We established a middle class, and now we’re losing it. We can’t afford food, and we can’t afford shelter.
I really want to leave the committee with the challenge of understanding where the productivity boom is going to happen and ensure that that wealth doesn’t flow out of Canada and back to the frontier models or whoever is designing the technology. We have to do that. We have to capture a portion of that wealth for Canada, and agriculture is a fantastic example.
We can’t let our farmers give up the power of their data to third parties. We have to ensure they govern that. There is the potential for efficiency gains and placing agriculture and technology in agriculture in the full capital stack — from funding, insuring, planning and planting — the land becomes an investment vehicle. Farmers become capital managers — investment managers — and this future is incredibly appealing to youth.
This generation wants sustainability. They want authenticity. They want the real world, and this is still the real world, albeit with a digital intelligence layer on top of it.
The PEI Science Fair was filled with ag projects this year, and our sustainable engineering school at the University of Prince Edward Island, or UPEI, is a great example. The whole country needs to be thinking about how to implement this technology. These are the beginning days, and we have to design capital flows that — in AI, we talk so much about reinforcement learning, and that’s how we get to self-learning AI. But we need to design circular capital flows as well that feed the benefits back into our social systems and our education and results in even greater thinking.
We invented this technology in Canada, and it could stand alongside smartphones and office software as technology that we failed to commercialize. We can’t have that happen.
It’s a very exciting time. I wish you the best in your study. It will lower food prices. Inflation is so sticky. We’re producing so much new money. We have monetary policy questions that are making it difficult to bring prices down, but there is no question that, if we get the governance piece right, implement this technology on-farm and think about food as a matter of national security, we can bring a more affordable breadbasket to Canadians.
Ms. Ryan: I want to point out, though, that it depends on how AI is implemented. You can also have algorithmic pricing of food, which could affect affordability. That’s where an algorithm targets — for example, Instacart was found to be doing algorithmic pricing: Based on the data it had about individual consumers, it gave individualized prices for people who were purchasing food.
This is where the regulatory guardrails can come in and say, “This is not an appropriate use of AI that benefits society.”
We want AI and technology that benefits society, benefits people and benefits affordability. When corporations are trying to monetize it and harm people through these practices, that wouldn’t be a benefit.
Mr. Killorn: Thank you. I’m not supposed to bring up blockchain, and I try not to with farmers, but for those who are interested, AI is part of a complex of technologies that we call Web3. Web1 is your basic initial internet. Web2 is social media. Web3 is AI and blockchain, and they’ve emerged simultaneously. They’re like capitalism and democracy, where one is perfectly suited to manage the other.
What blockchain will deliver — and what the government needs to learn how to harness — is decentralized ledger technology. With respect to what the government and the private sector are doing, we now have technology where that becomes fully transparent and trustworthy.
There’s a sister technology to AI — cryptography — that allows us to, again, put a harness on capitalism and ensure that it does what we want it to do.
The Chair: I did say that was the final word, but I’ve been convinced it’s not.
Senator Robinson: Ms. Ryan, to your point about food and algorithmic pricing, in Canada, we rolled out the Grocery Code of Conduct recently. I think your comment dovetails nicely into what we should be considering for version 2.0 of that. It’s one thing to talk about the production of food, but once that food leaves the farmer’s warehouse or bins — whatever it is — that’s when all the real value seems to accumulate.
I really appreciated your comment. Thank you.
The Chair: Ms. Wyonch, do you have something you feel you must say?
Ms. Wyonch: Since we’re out of time, I can follow up offline. I won’t take the last word.
I’ll just say thank you for your time, and good luck with your challenge of sorting through all this.
The Chair: On behalf of the committee, I’d like to sincerely thank you for taking the time to appear before us today. Your testimonies will be very helpful in our deliberations, and you’ve left us with some questions as well.
Honourable colleagues and guests, that concludes the public portion of our meeting. We resume in camera to discuss a draft agenda.
(The committee continued in camera.)