THE STANDING SENATE COMMITTEE ON HUMAN RIGHTS
EVIDENCE
OTTAWA, Monday, April 20, 2026
The Standing Senate Committee on Human Rights met with videoconference this day at 4 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.
Senator Paulette Senior (Chair) in the chair.
[English]
The Chair: Good afternoon, honourable senators. I would like to begin by acknowledging that the land on which we gather is on the traditional, ancestral and unceded territory of the Anishinaabe Algonquin Nation. My name is Paulette Senior, senator from Ontario and chair of this committee.
I would like to invite my honourable colleagues to introduce themselves.
Senator Arnot: My name is David Arnot. I am a senator from Saskatchewan.
Senator McCallum: Mary Jane McCallum, Treaty 10 territory, Manitoba region.
Senator Robinson: Welcome. I am Mary Robinson, representing Prince Edward Island.
Senator Arnold: Good afternoon. Dawn Arnold from New Brunswick.
Senator K. Wells: Kristopher Wells, Alberta, Treaty 6 territory.
Senator Pate: Welcome. I am Kim Pate. I live here in the unceded, unsurrendered and unreturned territory of the Algonquin Anishinaabe Nation.
[Translation]
Senator Hébert: Good morning. I am Martine Hébert, from the Victoria division, in Quebec.
[English]
The Chair: Thank you, 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 three panels. In each panel, we will hear from the witnesses and then the senators around this table will have a question-and-answer session. I will now introduce our first witnesses, who have been asked to make five‑minute opening statements.
With us by video conference, we have Martin Kwan, Legal Scholar and Affiliate Member, Center for Information, Technology, and Public Life, University of North Carolina at Chapel Hill. Welcome. Joining us in person at the table, we have David Lie, Professor, Canada Research Chair and Director, Schwartz Reisman Institute for Technology and Society, University of Toronto.
I now invite Professor Kwan to make his presentation, followed by Professor Lie.
Over to you, Professor Kwan.
Martin Kwan, Legal Scholar and Affiliate Member, Center for Information, Technology, and Public Life, University of North Carolina at Chapel Hill, as an individual: Honourable senators, thank you so much for the opportunity to speak with you. Today, I am here as someone who is deeply concerned about AI’s impact on work, education and cyberscams.
The current Fourth Industrial Revolution forces us to race with the machines. AI is no longer just a powerful software mimicking our reasoning, but it is also takes the form of robotics and humanoids, so an incremental replacement of humans is inevitable. Many job losses will be permanent. We are already seeing a high unemployment rate for young people, as entry‑level jobs are vanishing. The speed of replacement is faster than people learning new skills. Having a stable career becomes a thing of the past, with more and more people having to embrace gig work and underemployment. Nobody can be spared from this nightmare, and it will hurt vulnerable groups more, and it will cause social problems and widen the divide.
But just because AI cannot be stopped, it does not mean we do nothing. Canadian laws, values and principles obligate policy actions to protect work and protect vulnerable groups. The Supreme Court of Canada has repeatedly emphasized in many judgments that “Work is one of the most fundamental aspects of a person’s life.” In addition, it is an “essential component of his or her sense of identity, self-worth and emotional well-being.”
As a signatory to the International Covenant on Economic, Social and Cultural Rights, the Canadian government is obliged to safeguard the international human right to work.
As a matter of law, the Supreme Court has held that “Canada’s international human rights obligations should inform the interpretation” of the Charter of Rights, and the court has specifically made reference to the international right to work.
Employment is the backbone to many Charter rights. It’s about our dignity, our family life and our ability to integrate into Canadian society. Canadian courts have affirmed the “right to earn a livelihood” as “an interest of fundamental importance” which is “not beyond the scope of constitutional protection” and “should not lightly be overridden.”
Work is clearly indispensable, but the AI crisis exactly goes against this. AI, of course, presents many opportunities, but the fruits of AI will not be automatically distributed fairly. Some groups will experience net harm du- this painful AI transition. They need policy-makers to safeguard their rights.
Now, ensuring AI literacy is the national top priority, and it will help Canadian people better capture the opportunities. But AI tools can be prohibitively costly to vulnerable groups, so the government has a responsibility to make access to AI tools as inclusive as possible.
I personally prefer to call AI literacy “AI awareness” because, realistically, AI literacy cannot be about requiring everyone to become master users. Instead, the real policy goal is to ensure a nationwide understanding of AI’s capability and how it is structurally reforming the job market landscape so that Canadians can adjust expectations about careers, plan more carefully and act early.
As a matter of principle, the international covenant does not simply require vocational training, but it explicitly also requires vocational guidance. The government has an obligation to guide Canadians through this unprecedented change.
Very importantly, AI literacy is not just about work. It actually serves broader functions for Canada’s economic security. AI has been abused to carry out cyberscams, which are highly sophisticated, with, for example, face and voice impersonations. It threatens economic security, as scam victims have lost their life savings and Canadian businesses have been targeted. Anyone can become vulnerable to AI-facilitated scams. That is why AI literacy and awareness matter to all of us, not just to vulnerable workers, as we all face heightened risks of cybersecurity.
That being said, it is also important to bear in mind the limited role of AI literacy, as it will not stop replacement and work reduction. We have already seen profit-making corporations nevertheless choose to replace people with AI and automation for cost and efficiency.
In the long run, the government will have to explore options and contingency plans other than AI literacy alone. For example, it is time to revisit the role of corporate social responsibility in terms of their contribution to employment for society. By analogy, in the realm of environmental, social and governance, or ESG, we question whether it’s right and proportionate for large companies to pursue profit maximization at the cost of sacrificing the environment and ethics. I think there is room to apply the same logic to companies causing a mass reduction of jobs, and also reinforce the ethical responsibility of AI providers who turn a blind eye to possible misuse of their AI products for crimes like scams and also human rights infringements.
I will stop here for now. Thank you so much.
The Chair: Thank you. I was about to stop you, so thank you for doing that on your own. We will now go to Professor Lie.
David Lie, Professor, Canada Research Chair and Schwartz Reisman Institute Director, University of Toronto, as an individual: Thank you. I truly appreciate the opportunity to speak with you. My name is David Lie, and I’m a professor of electrical and computer engineering and Canada Research Chair at the University of Toronto, where I also serve as the director of the Schwartz Reisman Institute for Technology and Society. My work focuses on how important technologies affect people in their daily lives. Over the years, I have found that among the most urgent of these concerns is how to ensure that digital systems protect the security and safety of their users.
For example, in 2021, while we were in the midst of the COVID-19 pandemic, Newfoundland and Labrador’s provincial health authority experienced one of the worst cyberattacks in Canadian history. A malicious ransomware attack crippled the province’s health care system, forcing hospitals to cancel critical medical procedures, including surgeries and chemotherapy treatments. The personal information of over 100,000 Canadians was compromised, and the attack cost the province at least $16 million. While it is fortunate that no one died as a direct result, such incidents serve as a reminder of how important the safety and security of our digital infrastructure is.
Today, we face a new challenge. Artificial intelligence is being woven into the digital systems on which we are so dependent. Yet AI is a technology that is advancing beyond our ability to control and scrutinize. While control remains a significant technical challenge, greater scrutiny, through increased transparency, is something we can address today. Earlier this month, Anthropic, a leading AI company, announced that its latest model, Claude Mythos, has discovered thousands of vulnerabilities in software that millions of people rely on. Alarmed, Anthropic deemed Claude Mythos too dangerous for public release, instead limiting it only to a small number of organizations. While this is responsible, in some ways, it also keeps critical information about its full capabilities out of view for everyone else. If we cannot assess the level of danger, how can we know whether such advances will make failures like the 2021 ransomware attack more or less frequent?
As Canadians, we must develop a coherent, national AI strategy to address the next time new capabilities such as these arise, which, surely, they will. In light of these circumstances, I will now outline three areas of focus where I believe we most urgently need to protect the rights of Canadians.
First, we must protect children. It is absolutely critical that children are protected from the potential harms that AI systems could pose. AI has the potential to negatively affect our children’s education and harm their mental health. We have already seen incidents of “deepfakes” involving minors and overreliance on AI by students. One possibility is restrictions, or even outright bans, on exposure to AI for our youngest children until we understand better the effects of AI on childhood development.
Second, humans must retain the agency to make decisions. As AI systems become more capable, there will be a temptation to believe they are more reliable than humans. We must resist this. While AI systems may demonstrate impressive speed, memory and resistance to fatigue, this should not be confused with superior intelligence or judgment. In all major decisions, particularly those involving moral or ethical considerations, humans must be the ultimate decision makers and bear responsibility. If Canadians lose this critical role, they risk losing agency and may eventually lose their ability to meaningfully participate in economic life.
Third, as I have mentioned, we need greater transparency. Regulators, independent experts and the public must have sufficient information to evaluate and anticipate risks and harms. AI systems will continue to develop at a breakneck pace, and visibility into their capabilities is essential. We all have a right to know how we will be affected. As was the case for social media companies, like Meta, for example, some actors will be tempted to withhold information that doesn’t serve their interests. We cannot accept this and should create conditions that encourage greater transparency.
In conclusion, I urge you to consider the three areas of focus I have outlined — protecting children, ensuring human agency and promoting transparency — as priorities in a charter of rights for the AI age. The technologies I have described here are already impacting millions of people. Yet Canada still lacks any comprehensive AI legislation, and we simply cannot wait for a tragic, large-scale incident, like the one in 2021, before we act. We need decisive action to equip Canadians with the tools and resources to exercise control over how AI affects us all. The time is now to develop an enforceable, rights-based AI regulatory framework that is consistent with Canada’s international commitments to human rights and that follows a coherent national AI strategy.
Thank you for your time.
The Chair: Thank you both for your presentations. We will now proceed to questions.
Senator Arnot: This question is for Professor Kwan. Sir, do you see the right to work evolving into a justiciable issue in the context of algorithm decision making? Secondly, if courts become the primary mechanism for addressing harm, are we effectively accepting that protection will come only after harm has occurred?
Mr. Kwan: Thank you, senator, for the questions.
Regarding the first, as the Supreme Court has said, the right to work will be used as a persuasive and powerful resource to interpret the Charter of Rights and Freedoms, so I think the right to work can be applied to the algorithm issue, and you can safeguard, for example, citizens’ privacy rights and many other rights that are found in the Charter.
Regarding the second question as to whether it will be a post‑event protection, I think it will not be the case because the right to work imposes an obligation, a pre-event obligation, on the Canadian government to take action to intervene to ensure that people receive training like AI literacy and, as I have said, vocational guidance as to how AI is changing the job market. The right to work functions both pre-event and post-event in the courts and as a policy mandate. I think it can be used as a strong and powerful policy basis to intervene. Thank you.
Senator Arnot: Thank you. This question is for Professor Lie. How explainable are current systems in practice? And if systems cannot be meaningfully explained, how can individuals exercise rights in relation to them?
Mr. Lie: The first question was about how explainable systems in practice are. There are many different systems that can be considered AI. For many of the leading frontier models that are the most complex and capable, currently, we do not have the technologies or methods to make them explainable. This problem is growing worse.
Could you repeat the second question?
Senator Arnot: If systems cannot be meaningfully explained, how can individuals exercise rights in relation to those systems?
Mr. Lie: I’m speculating, but there could be ways to constrain how systems that cannot be explained are used so that they do not impact rights, and I would advocate that is a better route than trying to explain them given the current trajectories of where our technologies are going.
Senator Arnot: Thank you.
Senator McPhedran: My first question is to Professor Lie. I want to pick up on the point you were making about defence with an emphasis on children. I think before you can answer my question, I probably have to ask you to give us a bit of a primer on the differences between abuse via the internet and abuse or risk via AI.
Mr. Lie: The primary question was the difference between abuse by AI versus abuse online — or the internet. While the primary delivery of systems may be over the internet today, AI systems can exist in many embodiments. They can exist in increasingly autonomous vehicles or systems that have a physical embodiment. They can also exist in an institution, like a financial institution, making decisions that do not directly impact the people that are being harmed by those decisions over the internet, so AI systems are being woven into many of the systems that we perceive as digital or technological, but not all those decisions we interact with via the internet or via an online method.
Senator McPhedran: The distinction that I asked you to help us understand — and thank you for that — is partly from one of the statements made about protection of children or defence for children being a top priority, which I think we would definitely share. However, in countries like Australia, where there has been a very sincere attempt to use the law to do exactly what you described, we are now seeing the accounts of how, in fact, well over 50% of the children in that target age range are finding ways to access it. The question really translates into this: What is your sense of the capacity of lawmakers to effectively address the very legitimate concerns that you are raising?
Mr. Lie: You raise a good point. I have been closely following the events in Australia and other jurisdictions with the social media bans. There are some opportunities to learn from those experiences. Just because they have not worked out of the gate does not necessarily mean they are a bad idea; further study and refinement are necessary. The idea of curtailing or constraining the effects of AI on children is something that we should pay a lot of attention to.
Senator McPhedran: What is the capacity of lawmakers to do so?
Mr. Lie: Regulation and laws are one mechanism to bring about those constraints. Yes.
Senator McPhedran: You have nodded somewhat to the capacity of lawmakers and said it is one way. What are you seeing as other ways? Are you seeing an integration of other approaches combining for better protection?
Mr. Lie: The other way is through better education and literacy so that there are market forces that will also reinforce and perhaps complement any regulations or laws that are brought in to protect children. Yes.
Senator K. Wells: I will start in the room here with Dr. Lie. Thank you for being here with us. From your work and research, what countries are leading on issues of AI literacy and putting in appropriate guardrails and protections that we are talking about as being necessary? Are there some countries that are ahead of Canada that we should be looking toward? We have certainly heard from Minister Solomon about the fact that Canada is close to announcing an AI strategy as well.
Mr. Lie: That is a good question. There are several countries that I’m aware of that have made interesting and beneficial moves that we can learn from. For example, Switzerland has taken it upon itself to train its own model, which is driven by the government and citizens there as opposed to a for-profit, private corporation. So there is, in some sense, more accountability.
Another example is Singapore, which has made it a goal to make AI broadly available to all its citizens and to reduce barriers to AI access. I don’t know the exact details of how they will do that, but it is a worthy goal.
Finally, the EU, in general, has stood up comprehensive AI legislation. None of this is perfect, but it is a move in the right direction.
Senator K. Wells: Thank you for those examples. This might be a question for both our panellists. When thinking about youth, in some ways, AI may not be all bad if used in the right way. Here I’m thinking of myself as an educator. Are there ways we can harness AI for good? For example, using AI for students with learning difficulties or challenges or those with exceptionalities, what we used to call gifted students. Perhaps they can get enrichment from an AI tutor or basic content mastery in cases where a teacher with a classroom of 40 to 45 students in K to 12 simply does not have the ability to get around to everyone. This is particularly so when we see so many learners of different abilities put into the same classroom. There have been pilots more so in the United States than in Canada. These are AI schools that are piloting these models. I am interested in your thoughts. Have you observed anything where AI can actually help with educational attainment for young people?
Mr. Lie: Thank you. I will make this brief so there is some time. One area where AI can help with education is fulfilling gaps when there are not sufficient human resources. This is not an area where AI can supplant the ability of humans to educate other people. At the same time, artificial intelligence has capabilities that complement that. They can retrieve information much faster and can also be around 24-7 when a teacher or tutor is not available.
Senator K. Wells: Thank you. Professor Kwan?
Mr. Kwan: Governments in Asia have been encouraging university students to use agentic AI to boost productivity so that they can engage in one-person entrepreneurship. This is a very good model for students to learn employable skills and, at the same time, put their creativity into realization in products by using AI’s efficiency so that they can lower the costs and barriers for entrepreneurship and products. That is an interesting example.
Senator Pate: Thank you to both of our witnesses for being here. My question follows up on the last two questions. We know that Canada has signed the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law. We also know that signing on to a convention is not legally binding on countries. Are there things from the convention and what other countries are doing — if there is anything beyond what you advised that we should be recommending to the government to be included in legislation, regulations and policies that the government could implement? That question is for both of you. Maybe we can start with Mr. Lie.
Mr. Lie: I don’t recall the details off the top of my head, but from the EU regulation, the parts that are helpful are the ones that require more oversight. Depending on how the AI is deployed, taking into account the harms that could arise from the deployment, there is graded oversight or requirements to evaluate the safety of the AI system before it is deployed based on the level of harm it could potentially cause.
Senator Pate: Is there anything else?
Mr. Kwan: I’m not an expert on EU regulations, but one pressing concern is how you implement more regulations like the EU model. It may attract U.S. retaliation in terms of trade and trade tensions. That is also one legitimate consideration in terms of how we can borrow from the EU model. Thank you.
Senator Pate: Okay. My question then is to you, Mr. Lie. The model in Switzerland, which involves the idea of a national strategy instead of allowing it to be overtaken, propelled or generated by private industry, sounds like something you would recommend. Is that something both of you would recommend? How would you see us doing that? Should the federal government expand the ministry now that Minister Solomon has to do that kind of work?
Mr. Lie: To achieve something like what Switzerland did, you would need two types of input. You need human expertise, which we have. In some sense, AI was invented in Canada, and we still have many of the world’s experts here in Canada. Then we need access to compute resources. As many of you have probably heard, the current generation and technologies around AI are fairly compute intensive. Many experts are working to reduce those needs, and we hope that they’re successful. At the current time, it is still a very compute- and resource-intensive area to work in, and that is something that I know myself and my colleagues face significant hurdles in.
Mr. Kwan: No comment on Mr. Lie’s approach. Thank you.
Senator Arnold: Thank you to our witnesses today. This question is for Professor Lie. Number two of your recommendations was about us retaining human agency, and this will become more and more difficult over time, as we trust it more and more.
I noted in some of your research you talked about a program that you have at your school called “embedded ethics and engineering initiatives,” and I wonder if you could expand on that a bit, because it has come up a lot. How do we teach ethics and morals to the people who are creating some of these systems?
Mr. Lie: Thank you very much. I should give credit where it is due. This is a program started by my colleague Dr. Sheila McIlraith, who is also part of the Schwartz Reisman Institute for Technology and Society, the institute that I am the Director of.
Embedded ethics is a new way of teaching ethics to — exactly — the engineers and computer scientists who will be designing digital technologies. The approach is to actually embed the concept of ethics and ethical analysis in the process of designing things. In many of these courses, students train and apply concepts that they learn by building toy systems. We can think of these as assignments. During those assignments, the traditional way would be that you would have some objectives and you would build your toy system to meet those objectives, and you learn through that process.
The embedded ethics approach adds ethical considerations to that approach so that it becomes an objective and something that they must take into account when they come up with a design and to understand that you don’t have to have maximized performance or profit. These are taken and balanced against ethical considerations.
Senator Arnold: Do you have any tangible examples of how that — what are the outcomes of that? Have you seen a difference as a result of this program?
Mr. Lie: I can give an example of how it was done. I don’t have statistics. That would be something that I can ask my colleague afterwards, if you’re interested, to see if we have collected any — by “tangible,” you mean quantitative statistics? Yes, I would have to check on that.
A qualitative example of how this could be done or how we did this is that students, for example — this was around COVID-19 — were asked to design an application for tracking COVID exposures. As you might recall, we actually did deploy one. The Canadian government did deploy one. In the design of that, there were various trade-offs. You wanted to have accurate and timely notification, but you also had to balance that against privacy and the possibility that someone may be incorrectly labelled as having COVID and the stigma that might be associated with that.
So this is an example, and students, I can say, qualitatively said that the experience was much more interesting, and they felt they learned a lot more than your standard assignment, but I don’t have any quantitative metrics off the top of my head.
Senator Arnold: Thank you.
Senator McCallum: How can Canada ensure its AI strategy aligns with societal values, protecting individuals from potential harms, such as data misuse and algorithmic bias, while fostering innovation that benefits humanity? Is that even possible?
You talk about embedding ethical analysis. Can that be removed even if someone embeds it? Is there an ability to remove it? And to what extent is data sovereignty necessary for AI safety in Canada?
Mr. Lie: Okay. So there are three questions there. I’ll start with the last one because I remember that one.
The data sovereignty question is important. Data is one of the ways that we improve AI systems, and so the ability to retain data and use it for the benefit of Canadians is going to be a critical thing that we need to keep track of. I think it’s a very nuanced issue, so it doesn’t mean that we cannot share data, but we need to recognize that data is very important, and it should be treated very carefully.
On the second question of embedding ethics into systems and whether those ethics can be removed — and I think this is related to the first question now, which was how can we ensure that the systems protect human rights and reinforce societal values. I believe, in both of those cases, that it’s important that the AI systems are transparent, both the construction, the intent, and their actual operation.
There will be a tendency to — especially if it’s a profit-driven company, but, in many cases, to not show all the cards or to show one side, a beneficial side, without necessarily talking about the harms or the riskier sides. Even with systems that were designed with ethics in mind, over time, the use of those systems may drift or the systems may continue to be developed, and some of the ethical parts may not retain their strength or they may not function in the way that was originally envisioned.
I bring it back to my third point, which is that we need transparency so that, collectively, we can understand the impact that the AI systems will have. We cannot just take the word of AI system creators to say that, “These are the benefits, and you will want these benefits, and never mind potential harms.”
Mr. Kwan: I have a few very brief comments. To protect Canadians, it’s important to use human rights as the policy mandate. I think the standing and power of human rights and its relevance are increasing in the AI context. For example, there is a recent Canadian case involving a company called Clearview AI. It is a U.S. company, and it collects online photos of people so that they can use AI to compare the photos with law enforcement databases as a commercial model. The court says these infringed human rights, and I think we can use human rights and privacy rights as the basis, as the courts have been willing to stand for human rights even in the AI context.
Thank you.
Senator McCallum: Thank you.
The Chair: Thank you. Before we go to the second round, I just want to insert a question or two myself.
Professor Kwan, you talked about the importance of AI literacy in terms of building awareness, and this is something I feel very strongly about as well. I’m wondering if there are any regions that you can point to where AI literacy is working in terms of awareness? You mentioned it’s also for economic security. Could you share any knowledge you have about that?
Mr. Kwan: I will focus again with examples from Asia. For example, in Hong Kong and in Asia, such as Singapore and Thailand, we are facing cyberscam issues. The cyberscam from Southeast Asia utilizes AI to make the scams very realistic, and people can’t differentiate between it. But with government promotion of AI awareness and people starting to understand that AI literacy is so important, that they know that the photos they receive or the messages or the forged messages that they receive can easily be fabricated. Therefore, AI literacy is about how we educate the public that AI has a lot of capabilities and it’s prone to be used for crimes and scams.
I think the public education aspect is very important. It’s not just about K-to-12 education and asking students to learn AI tools, but it’s about a broader, nationwide approach for everyone. I hope that answers your question.
The Chair: It does. I’m also wondering about effectiveness. Do you have any evidence in terms of the effectiveness of that in Asia or elsewhere?
Mr. Kwan: In terms of statistics, I don’t have any right now, but we can see the scam figures are constantly dropping in Southeast Asia and the broader Asia region because people are starting to build up the AI awareness that I talked about. Technically, you cannot say that it is literacy because it’s not about using AI tools, such as generative AI, but about how generative AI can be abused and how other AI tools can be used for fraudulent schemes, so I think that awareness is crucial.
The Chair: Thank you, professor.
Professor Lie, I wanted to ask you about the embedded ethics that you mentioned and if the engineering students had any reaction to that in terms of — you don’t always think of engineering students meshed with ethics, unfortunately.
Mr. Lie: In defence of the profession, I should say that to be certified as engineering, you have to take an ethics course. At least, the profession does care about ethics, but to be honest, a lot of times, people grumbled at having to take the ethics course because it was put at the very end or off to the side.
Again, anecdotally, what I heard from students was they would much rather have it embedded with the other technical material that they were taking. My understanding is that this made it more relevant, so as a result, it was more interesting and easier to learn and digest. They also felt that, overall, maybe tied to that, they got a better education because they could see how ethical conversations were important in the exercise of engineering design.
The Chair: I have just a quick question for each of you. Were you or your institution, Professor Lie, consulted for the upcoming strategy that the government will release?
Mr. Lie: We got the same invitation for responses and input that everybody got, but we didn’t get any particular consultation, no.
The Chair: Professor Kwan, were you consulted, albeit by the U.S.?
Mr. Kwan: I was not consulted by the U.S.
[Translation]
Senator Hébert: I’d like to hear your thoughts on the economic side of AI.
We know Canada’s facing significant challenges, especially in terms of demographics, and that they are causing significant labour shortages in several regions.
We also know Canadians expect governments to be more efficient in providing government services. Meanwhile, departments and agencies are going through an attrition process. I would add that when it comes to productivity, we are laggards and we’re facing significant challenges.
AI clearly represents an alternative solution when you consider these three phenomena, a tool that governments and businesses must take into consideration to face these very real challenges that undermine our ability to develop as an economy.
I’d like to hear your thoughts on how to reconcile that with an approach that would set guidelines or bulwarks on the ethical issues you raise, for example, since we need this tool.
What can we do about that, and how do we reconcile everything?
[English]
Mr. Lie: You rightly point out that there is a tension between increasing productivity and filling gaps in our expertise and workforce, and also protecting workers from being displaced by artificial intelligence. Artificial intelligence seems to be able to solve the former, but we don’t want the latter to happen. That’s exactly the challenge that we need to meet.
The ongoing dogma we hear in the media and even from business leaders is the idea that we can have AI intelligence agents replace workers and maintain the same level of productivity with fewer workers. I believe that is short-sighted and just not very creative.
What we should be asking is: If we keep the same number of workers and make them more productive with AI, what can we achieve? I have not heard many people talk about this yet, but I believe we can promote the idea that people should look at the prompts that they have in a different way. We can ask how to complement the workers we have or make them more effective with these tools rather than thinking about how to replace them.
[Translation]
Senator Hébert: In many regions across the country, it’s not about replacing workers. It’s about continuing to provide services, precisely because there are no workers and we’re having a hard time finding some.
I’m trying to see how to set guidelines or parameters to avoid the abuses we’re exposed to with AI, but also to help us deal with these very real economic challenges in many regions across the country.
[English]
Mr. Lie: I don’t know how to set up those guardrails correctly. I do know that they are needed. I think it is important, as you state, in cases where there is a real need and a lack of service, that we fill those services with the tools available to us, and one of those tools could be AI.
In the three areas that I emphasized in my statement, I would say one thing that is important — and you can maybe think of this as a guardrail — is if the AI is being used to provide a service where there are still moral consequences or moral judgment involved, it cannot be completely done with an AI. We should think of ways that a person doesn’t have to be physically there or where maybe they can use AI to deliver these services remotely. If it’s a medical service, maybe it could be combined with some sort of telemedicine or robotics if the technology becomes available so we can augment the services of people to provide services where they’re needed.
The Chair: Thank you. We’ll now move on to second round. Three minutes each.
Senator K. Wells: I think this might be a question for Professor Kwan. We’ve seen that AI systems are increasingly being used in employment-related decisions, yet accountability often remains very diffuse. Just today, I was browsing on LinkedIn — and you know how ads pop up that you might be interested in? I saw a statement at the bottom of an employment ad that said, “AI is not used to filter applications.” It jarred me a little bit to see this new language that has emerged to make hiring practices explicit.
How do either of you think Canada should assign legal responsibility when AI systems result in discriminatory or harmful outcomes in hiring or employment-based decisions? This is particularly important because, as we heard, you may not even know the models or how these systems are arriving at this filtered conclusion. We also know that often the models on which they are based have their own particular racial biases behind them. Maybe we’ll start with Professor Kwan and move to Professor Lie.
Mr. Kwan: It is important to put everything into the context of human rights. The courts are very willing to apply human rights to those contexts to safeguard any human rights infringements, but the difficulty for vulnerable groups is that they may not know that their applications have faced bias. That is why it is important to not just focus on legal measures. We also need to emphasize corporate social responsibilities and also the ethics involved so that it should be a scheme where companies see it as a social responsibility to adopt fair and transparent algorithms and AI systems, as the courts may not always be there to help them, and vulnerable groups may not always be able to rely on courts.
It’s not just a cause; we need a corporate culture that emphasizes that human rights are to be complied with, and those rights are to be observed.
I hope that suggestion helps.
Senator K. Wells: Thank you.
Mr. Lie: In response to the ad that you encountered on LinkedIn, which is a good way to start, I think we should remember that it’s not at all clear that just because it’s reviewed by a human that the decision is necessarily free of bias.
This is where I think my point about transparency is going to be very important. An AI system can improve the fairness of such decisions when augmented by people because AI systems can be more transparent than people. What I mean by that is that I can make an AI system review the application at different times of the day after it was just restarted. If the application was on yellow paper or red paper, or if I changed this word to that word, what is the outcome? I might not be able to ask a person to do this. They may not like me if I ask them to do all this work for just a single application.
The AI systems also represent an opportunity to reduce bias and make decisions more fairly when deployed in a transparent manner and augmented with human judgment.
Senator K. Wells: Just to clarify, if the algorithms are listening, I’m not looking for a new job.
Senator Arnot: This is a question for both witnesses. What is the single most important legal reform Canada’s legislators should implement within the next two years on these issues?
Mr. Lie: I don’t know about a single legal reform, but I believe Bill C-27, in terms of privacy and the AI legislation, was an important step, and unfortunately, I know it was not successful. I believe parts of that bill should be revived and revisited.
Senator Arnot: Professor Kwan, do you have any comments on that same question?
Mr. Kwan: We definitely need to implement indirect bans on AI and disruptive technologies. The social media ban is actually a good move because it’s not just about protecting children, but it sends a strong signal to companies that they need to be accountable for their tools and products. If it’s harmful, it’s the companies’ responsibility to sort it out.
Legislation that is an indirect ban or social media ban seems to be a niche area, but it actually has broader functions for the whole nation and AI development because it sends a signal to companies. It’s important to have some form of bans. An outright ban would not be possible because of geopolitical tensions with the U.S., but some form of indirect actions and emphasizing the corporate social responsibility role. Thank you.
Senator Arnot: Thank you for that answer, sir.
Senator McCallum: How would you define AI sovereignty? How important is that definition, since sovereignty exists on the spectrum with different thresholds? We look at national security, defence, critical systems and the requirement that AI be developed and deployed entirely within Canadian borders. Then you have lower thresholds maintaining that systems developed by foreign entities meet Canadian standards. How would you define AI, and how would it be used in different contexts?
Mr. Lie: If I go back to my statement, there are two things that we lack, currently, with artificial intelligence systems: control and the ability to scrutinize them. Those form a basis for having sovereignty over AI systems in our country.
Certainly, scrutiny and transparency are going to be easier to enforce if the systems are within our control and legal frameworks.
When it comes to control, I mentioned that control is difficult because it’s still an evolving technology, and some of those abilities are yet to be developed. A related control is the ability to constrain how it’s used and what it can do, which is maybe easier. You can think of control as doing something to the AI system to have it behave in a way that’s beneficial. The constraints are putting locks or a box around it so that it can’t cause harm.
Both of those will be easier to enforce if those systems are within our legal frameworks. That does not necessarily mean that the systems have to be developed entirely within Canada, but at the same time, that is one avenue to ensure that it is entirely governed within our frameworks.
Senator McCallum: Thank you.
Senator McPhedran: This is a question to both of our witnesses. To what extent do you know of research initiatives that are specific to guardrails? If you do know of such specific research, in your estimation, is there adequate funding for this research? Is there an emphasis on coordination among research bodies as opposed to creating competing silos?
Mr. Lie: I can speak to this. I have researched directly in the area of guardrails, and I lead a project that is funded by NSERC. I’m relatively fortunate that I have resources and that I’m able to share those resources with my colleagues in this area. Nonetheless, those resources, in some ways, can be inadequate to compete with some of the resources that companies have. There are also instances where students and researchers feel drawn to leave the area of pure, socially oriented research to work at a for‑profit company, and that’s something that we need to compete with.
Maybe it’s not a clean answer. I have some resources, and I’m very thankful to the Canadian government for providing those. At the same time, it is a competitive race with many players, and many times we feel outgunned.
Senator McPhedran: Thank you. Do you have anything further on coordination among research initiatives? The question is also to Professor Kwan if you want to respond.
Mr. Kwan: I have no comment on this. Thank you.
Mr. Lie: There is some coordination. The particular project I have has a group of 19 professors and 40 students who meet, and there is some coordination there.
I would say that we are, in some ways, left to our own devices in coordinating this. What could be more beneficial is a more structured framework to bring various people — I’m only working with them because we applied for this grant. We were funded. Thank you very much. We worked together on this project. For those on other projects, we don’t have a framework for coordinating with them.
Senator McPhedran: Thank you.
Senator Pate: My question is to both of you and started with Mr. Kwan’s response as to why we need a ban. My question is: In a context where you are talking about market-based work that is being done and where enterprises are corporate and capitalist based, I think I heard you say — and please correct me if I misunderstand, which is completely likely, as I’m techno-twit — how do you achieve those kinds of guardrails? If you leave it to the company or if you ban companies with the idea, they will then develop some ethical, moral or legal structures that will confine and not allow them to do the very things they wish to do. Why they are influencing young people is because corporations are making money from doing that, and they want to continue to do that. I’m at a bit of a loss about how the ban alone, without other mechanisms, would actually achieve the objective of some kind of better ethical, moral compass to the AI process. Please help me understand what you meant, Professor Kwan, and then, Professor Lie, if you have something to add, I would be happy to hear it.
Mr. Kwan: I based my proposal on the framework of environmental, social and governance, or ESG, which is about how a company achieves profit maximization while not sacrificing the environment or ethics, like the use of child labour, as a means. ESG is a market mechanism that not only prompts the company to comply with its own initiative but also sends signals to consumers and the public as to whether the company is ethical or not. The goal of my proposal on corporate social responsibility and some sort of indirect ban is to influence not just the company but also the public to understand and differentiate which AI is ethical and what irresponsible use or proficient use of AI is, so that we use market forces or consumer forces — or, in the case of children, allows teachers, parents and children themselves to understand which one is good and which is unethical.
Senator Pate: Thank you.
Mr. Lie: I would like to clarify. In my statement, I preferred the possibility of bans or constraints, not necessarily as a mechanism to induce certain corporate behaviour but as a mechanism to better protect children until we understand better the effect of AI on them. This is more of a pause on the deployment and rollout of AI in our schools or organically through the internet, so we have time to understand and study the effects. To encourage or create the conditions for better transparency, I think there are many mechanisms, and bans may or may not be the best way, but I think it is important that we encourage transparency and ask, “Have you studied the effects of AI on vulnerable populations?” before they deploy. If they have, wonderful. We’ll encourage deployment or make it easier. If they haven’t, maybe they should go back and do that and show us all the data, not just the data that best suits your case.
The Chair: Great. Thank you so much. You certainly filled our roster with a lot of questions today. I want to thank each of you for agreeing to participate in this meeting. Your assistance with our study is greatly appreciated.
I will now introduce our second panel. Our witnesses have been asked to make an opening statement of five minutes each. This will be followed by questions from the senators.
With us at the table in person today is Teresa Scassa, Canada Research Chair in Information Law and Policy, University of Ottawa, just around the corner. Joining by video conference, please join me in welcoming Jennifer Pybus, Canada Research Chair in Data, Democracy and AI, Department of Politics, York University.
I now invite Professor Scassa to make her presentation, followed by Professor Pybus.
Teresa Scassa, Canada Research Chair in Information Law and Policy, University of Ottawa, as an individual: Thank you very much. I really appreciate the invitation to be here today.
I work in the areas of privacy and AI governance. We are now seeing AI technologies having wide-ranging impacts across society, and the area of work is no exception. I will outline some impacts linked to human rights and work before briefly identifying some areas for policy attention.
AI will undoubtedly impact employment in Canada. Because things have moved so quickly in such a short time and because we are facing multiple overlapping impacts — for example, the pandemic, AI technologies and trade challenges with the United States — it is difficult to predict with certainty what the full nature and extent of those impacts will be.
Further, as AI capabilities change, so too do the implications for work. The shift from more conventional forms of automation to the use of generative AI has been transformational. We are on the precipice of yet another transformational shift with agentic AI, and so these things make it very hard to predict what is coming.
The impacts on the labour market come at a bad time for Canadian youth. Students in some programs of study are seeing significant decreases in employment opportunities in their fields. There is uncertainty about the future of work and where entry‑level jobs will be. Our youth today face other substantial challenges driven at least in part by AI technologies, including social media addiction, social alienation, polarization, mis- and disinformation and toxic deepfake harassment. We know that youth mental health is in serious decline. Although some of these concerns may seem to stray from the labour context, youth are our future, and young people are struggling.
Artificial intelligence systems will have impacts on more than just what jobs will be available and what those will look like. AI algorithms are often at the entry point for employment, with these tools being used to screen and assess job applicants. Once in the workplace, AI technologies are increasingly used to manage and evaluate employees. AI technologies can assess engagement, monitor remote workers, track off-site employees, evaluate the tone of communications, rate the productivity of the employee and could even be used to assess emotional well‑being. Agentic AI systems can take things to a new level, automating human resource responses to AI monitoring.
Harms from the use of workplace AI systems range from adverse impacts on employee morale to unfair automated treatment, as where an agentic AI system issues a reprimand for underperformance without considering personal circumstances such as a sick child.
Privacy rights may be infringed by over collection or misuse of personal data. Discrimination is also a significant risk where the data on which algorithms are trained is biased, or where algorithmic design fails to accommodate diversity or incorporates stereotypes.
Some categories of AI-related work pose their own human rights considerations. Some employment contexts, such as call centres, may be highly surveilled, increasing stress and anxiety for workers. Some work will involve AI piecework. There are already platforms that offer low pay for repetitive tasks, such as labelling data. Content moderation is often performed by humans, who are required to review and assess large volumes of distressing and even horrifying content, with little or no mental health support. In some cases, this work is outsourced to countries in the Global South, which pushes the consequences of some AI technologies onto the most economically desperate.
These are all points of intersection between AI technologies, work and human rights. What are potential policy options?
The provinces will clearly play a role, as some issues will need to be addressed in employment standards legislation. Issues of discrimination in employment often fall to human rights commissions. At both federal and provincial levels, attention will need to be paid to whether human rights legislation requires amendments to address algorithmic discrimination, which will be complex to investigate and establish. Apart from issues of whether laws should be changed, there is a need for adequate resources for human rights commissions to engage in technologically complex algorithm-based human rights investigations.
Privacy laws have already been carrying a significant load when it comes to AI regulation in Canada, but our private sector privacy laws offer uneven coverage in Canada. I hope that we will soon see a federal bill to reform the Personal Information Protection and Electronic Documents Act and that these reforms will offer significant enhancements to privacy protections.
We do not have ex ante risk regulation for AI technologies in Canada. The artificial intelligence and data act that was part of Bill C-27 died on the Order Paper, and the government has signalled that there are no current plans to revive it. While it is clear that AI regulation may be a significant trade irritant, the lack of appropriate regulation creates vulnerabilities for Canadians.
I hope that your study will surface many key issues in this area and explore potential solutions. Thank you for your attention, and I look forward to your questions.
The Chair: Thank you, Professor Scassa. Over to you, Professor Pybus.
Jennifer Pybus, Canada Research Chair in Data, Democracy and AI, Department of Politics, York University, as an individual: Thank you, honourable senators, for having me here today.
Today, I will present some evidence from a use case my team at York University has been working on that examines AI gender harms, with particular attention to the governance of social health data. By this, I mean non-clinical data that individuals voluntarily share about their bodies through digital platforms, apps, smart devices and generative AI. This challenge is further compounded by the fact that such health data can be reused by AI systems in both upstream contexts — where end-user data is accessed — and downstream contexts — where inferences and predictions are mobilized at scale.
Given the committee’s focus on work, I want to situate this challenge within a growing reality: AI systems are increasingly performing forms of health and emotional labour that have traditionally been carried out by regulated professionals. In doing so, they are poised to displace regulated labour, with largely unregulated systems.
For the committee, this is a critical human rights challenge that needs urgent attending to.
My use case builds on my research in female health technologies, including two privacy audits of mobile applications, such as menopause apps and baby-tracking apps, and my current SSHRC-funded research on intersectional information harms.
As part of this work, we have surveyed over 300 Canadians experiencing any stage of menopause to better understand how and where women access information about their symptoms and experiences. Our findings correspond with the most recent Abacus Data survey, wherein Canadians are increasingly turning to digital and AI-driven tools to fill gaps in their health care. Nearly 89% of Canadians go online to access information about their health, and those who use generative AI for their health needs are five times more likely to experience harm compared to those who do not. While research shows millennials and younger people are more likely to use AI for their health, our survey’s preliminary findings show that women are using ChatGPT, among other large language models, to get information about their symptoms and to manage the mental health challenges brought on by this life transition.
The turn toward agentic AI can be understood in the larger context of gaps within formal health care access. For example, 6.5 million Canadians do not have access to a family doctor, and many who do are not being seen in a timely way. Similarly, the Menopause Foundation of Canada equally notes that 4 in 10 women will see a GP about their symptoms, and of those who do, 70% will find the information unhelpful.
Since women’s health has long been underserved, it is not surprising that agentic AI is increasingly poised to become a de facto health intermediary, already filling a labour gap that our health care system is not adequately addressing. While some forms of agentic AI are being developed responsibly in clinical contexts, the reality is that many AI systems that women are turning to in lieu of their general practitioners, or GPs, are operating outside of meaningful standards, oversight or accountability.
This shift introduces a range of significant and compounding harms, five of which I would like to emphasize based on my research. One, misinformation and inaccuracy: Emerging research, including one study published in Nature, suggests that over 20% of health information generated by large language models may be inaccurate or misleading.
Two, profiling and data sovereignty: Health-related data can be repurposed for advertising, profiling and monetization. This comes with digital sovereignty risks, as many apps in our audits were directly processing data in the U.S., not Canada.
Three, inference-based risks: AI systems can generate predictions about users’ bodies’ health conditions and behaviours, often without their knowledge or consent.
Four, intersectional harms: These systems may produce biased, culturally inappropriate or incorrect information, disproportionately impacting marginalized groups across gender, race, age and socioeconomic status.
Five, potentially unequal privacy harms: There are potential privacy risks being disproportionately experienced by those with limited access to health care, leaving these individuals more exposed to the collection, use and circulation of their sensitive health data.
Taken together, this leaves a growing segment of Canadians, particularly women, turning to AI systems to perform essential forms of health, emotional and informational labour even as these systems operate without the safeguards and protections that would normally govern such work.
Thank you.
The Chair: Thank you for your statements.
Senator Arnot: This question is for Professor Scassa. Is PIPEDA structurally capable of governing AI systems, particularly those based on inference rather than data collection? You have answered that question already in some senses, but I want to know where you see the significant enforcement gaps that need to be addressed and that this committee should be aware of. What would a minimally sufficient federal framework look like in legal terms?
Ms. Scassa: In terms of governing inferences, the privacy commissioners across Canada have been engaging in some really productive collaboration to harmonize approaches to their respective statutes to the extent possible, including with respect to AI, and the emerging position across Canada is that inferential data may be a collection of new personal data. So it is captured by the law, and it would be subject to the law, which is important and useful, and it suggests that there is plenty of flex left in principles-based legislation.
However, where the legislation falls short is where you have identified: in terms of enforcement. What we are seeing, particularly with large platform-based companies from the United States, is that we have investigations, recommendations and then silence. There are no responses or almost insulting responses to those recommendations. So we need, and we have needed for a long time now, much better enforcement of data protection laws. That will be one thing for privacy reform.
The other thing I’m concerned about with privacy reform is that we don’t give up too much. There is a risk there. We have flexible, principles-based legislation. There is a risk that in the desire to free up more data for use for innovation — because we are seeing this tension between innovation and governance — that privacy rights will be undermined, and that we may see a narrowing of the principles in the legislation. That’s one of my concerns in the reform process.
Senator Arnot: Thank you.
Senator McPhedran: Thank you to both of our witnesses. I’m particularly grateful for the references to youth. I want to frame my question in terms of guiding principles because we often develop laws starting out with some guiding principles. To the best of my knowledge, in Canada, we don’t have — including in the legislation that has so far failed — guiding principles that create a very clear focus on human rights and that articulate that as a key guiding principle. Do you think that is needed? If you do, can you tell us a little bit, in practical terms, about what that would look like if one were actually drafting something to add to the law?
Ms. Scassa: When we had a battle with Bill C-27, there was a lot of discussion about whether the bill should recognize privacy as a fundamental human right, and there was a lot of resistance to it on the part of the government. A big part of that resistance was constitutional phobia: the fear that a bill that was based on a trade and commerce power that started talking about human rights would be seen as not falling within a federal head of jurisdiction and encroaching on the provinces.
To some extent, toward the end, the government was backing down from that position. It will be a barrier to seeing human rights being framed front and centre in the legislation with the same kind of orientation, for example, that you see in the European Union’s General Data Protection Regulation, or GDPR. That is because this legislation is going to be grounded in trade and commerce power, and there have been questions in the past about its constitutionality. That’s unfortunate. That said, privacy is a fundamental human right. The courts have called it quasi-constitutional legislation. In many of their decisions interpreting this legislation, the courts are becoming much more forthright in their articulation of the human rights dimension of the rights protected by these laws.
We need to really ground these kinds of principles and approaches in needed legislation. For example, when it comes to children, we have the concept of privacy by design and the concept of privacy by default. The U.K.’s age-appropriate design code makes privacy by design and, by default, essential for applications developed for children. We could do the same thing. We could do it within the context of this legislation. We could do it constitutionally. We just have to do it. Part of the frustration is that we are going to have to stand up to some of these industries and platforms and say that, when it comes to children, these are the new rules. That code, at least, is gaining some momentum internationally, which is helpful for Canada.
Senator McPhedran: Thank you very much.
Ms. Pybus: One of the studies we did is that we audited baby-tracking applications. When it comes to whatever application you are looking at, in all of the audits we have looked at, there is a huge gap between what people think they are consenting to — if they are actually reading the privacy policy, which is very long and would take up to three days of someone’s life to actually read through them — and what is going on with the app. One of the biggest challenges is that the idea of consent is extremely onerous and challenging for whoever has to undertake that.
Secondly, when it comes to what happens to that data, what are they consenting to? When you think about what AI systems are doing behind the scenes and when you look at how this profiling is going, how this data is being used, how this is feeding into an AI system and how these inferences are being created, what we started with has been completely transformed. Nobody is agreeing to the data that they actually think they are putting into the system. This is a huge challenge.
From the perspective of the platform, with baby-tracking applications, for example, we found that AI was looking at access to when women are breastfeeding, when they have had miscarriages, when they need to change the diaper, medication and all kinds of different things that are being accessed and shared with third parties that were advertising identifiers in each of the apps we looked at.
Supposedly, there is a family policy on Google that says that if children’s data is involved, it should be protected. I think this is where Canada has a role to play. From their perspective, children have to be actually playing the app, and then we will protect their data. If the parent is actively inputting data about their child into this app, then it’s the parents’ responsibility, and we can access this data, so we have data going from the moment of conception all the way to when the child is growing. I think this is a very important aspect that needs to be considered for future legislation.
Senator McCallum: There is a dangerous trend where each new technological advancement creates additional barriers for already struggling communities. This concern is echoed by Natiea Vinson, Chief Executive Officer of the First Nations Technology Council, who identifies how AI systems built on Western frameworks often fail to serve communities’ needs.
Voices from Canada’s social sector paint a picture of a country where AI advancements that could be beneficial in addressing inequality instead threaten to create a two-tiered society that risks exacerbating inequities at scale, with one tier equipped with the latest AI tools and capabilities, while another is increasingly left behind by systems they cannot access, influence or benefit from.
What recommendations do you have to address this, keeping in mind that each new technological advancement builds upon previous disparities, which, in some cases, make it increasingly difficult for disadvantaged groups to catch up without intentional intervention?
Ms. Scassa: That’s a terrific question, and I don’t think there is an easy answer. This is a tough burden for governance in general, but for communities where there are already structural inequalities, limited resources and strains on capacity, it’s overwhelming. The issues are different, too, so the ways in which dominant cultures think about the challenges posed by AI may be important in that context, but they don’t necessarily respond to all of the concerns.
I think that it just becomes an overwhelming challenge. If you think about the tremendous work that’s gone into Indigenous data sovereignty and that still remains to be done, that’s just about data, and it has required tremendous work and capacity. It’s very complex, and AI is rolling along behind that and raises all sorts of new issues and requires new capacity.
In terms of policy solutions, I think we need to build capacity, and that’s a tremendously important part of it. We also need to build literacy and capacity within communities to support research and to give people the tools that they need to understand, think about and develop their own perspectives and frameworks for these tissues. I think that’s a big piece of it.
Beyond that, there are some synergies in the work that’s being done on bias and discrimination and the thinking that’s being done about data and data inputs into AI and so on where work can be leveraged. Yes, I think this is an absolutely enormous challenge. Sorry, I really wish that I could suggest an easy solution, but capacity is going to be an enormous piece of this.
Senator McCallum: Ms. Pybus, did you want to comment?
Ms. Pybus: My honest answer is that I don’t have an answer to your question, but I think it’s incredibly important that it’s being raised.
One thing that struck me about the women we’ve talked to in the research, the workshops and the surveys that I’ve done is that one of the reasons that a lot of women have turned to AI when it comes to both managing their symptoms and using it as a diary to vent is to feel sane. It’s actually a sign of empowerment for them. They feel “I can get a handle on this” and “I can understand my symptoms,” which is not necessarily a bad thing. The question is: Where are they turning, and how do we ensure that this is safe, that their data is going to be safe and that this is not going to come back to them in an unexpected way? For example, it might impact their employment, as some data has shown, or it might impact different areas of their lives or come back with misinformation.
One key thing to think about is why vulnerable communities or people might go toward these technologies, what they’re gaining from them and then how we safeguard that or ensure that there are guardrails to make sure that this doesn’t create more harm in the long term.
Senator K. Wells: I want to pick up on some of the conversations we’ve been having here in this session. When we see AI systems continuously scrape, infer and recirculate personal data, I’m wondering, from both of you, how do we understand the right to disappear or the right to be forgotten as a fundamental human right?
Silence means it’s a good question. We’ll start in the room, please.
Ms. Pybus: My answer is it’s not possible.
I think that if you look at an application, for example, you can see and unpick all the problems that are there. You have a developer, an earnest person who wants to earn some money and to create something, but in order to create an application, there is no such thing as writing the code from scratch. You need to use all these different companies to provide different developmental and monetization tools et cetera, and we start to look at this back-end infrastructure, which predominantly belongs to all the platforms. Google is in 99% of the apps in everybody’s phone. Meta is in about 66% of these applications, and they’re providing these things, and the developer is asking for certain services. Those third parties provide those services, and this exists in this data-for-service economy.
The challenge is that you have something like Google Analytics, which is a good example, providing different ways of monitoring how people use the apps. Google, for example, has 500 “app events” that exist so that every micromovement inside the app can be monitored and put into their own AI system, which ends up in Vertex AI platform. That’s where they create different features about the user, which then get stored in Google’s Feature Store and can be used whenever they need. The challenge that this brings is that once this data has entered into Google’s universe, you can’t get it back. Once they’ve created a feature, that’s part of their own infrastructure, and that’s why they’re worth so much. Because the more features they have, the more they have at their disposal to create these kinds of different behaviour profiles, which is why everybody goes to them in the first place.
How you deal with this is, I think, a big one in terms of understanding the role that these companies are playing, so finding ways to ring fence the type of access that they have and what they do with that data.
The short answer is it’s not possible to fully disappear. The question is then: How do they use these different assets that they create from everyone’s personal data, and how do we limit that?
Ms. Scassa: There are certainly privacy advocates who suggest that the focus on consent and control over data is now becoming entirely misplaced, and the real focus needs to be on addressing inappropriate uses of data and controlling exploitative and inappropriate uses.
There is plenty of work to do there, that’s for sure, simply because the control part has become virtually impossible. What I would add to that mix and to those challenges is the fact that we are so omnipresent in this digital world and that it’s very hard to control or limit the collection of our data by private sector companies because of the way this operates. Add to that the fact that if you are not part of the data, you are excluded from some opportunities and benefits, and that’s a challenge as well.
Then add to that the fact that, as the private sector collects all of this data and organizes this data and profiles this data, we have the growing problem of the interrelationship between the government and the private sector when it comes to that data. So the question of state surveillance, access to that data, the characterization under the Privacy Act of some of that data as publicly available data and therefore data in which you have no expectation of privacy — these are huge challenges. So just to make it worse, there are those dimensions as well.
Senator K. Wells: It seems like, say, a child born today and lives their life and lives a long life — almost every moment of their existence could now be tracked by algorithms, and all that data is just going to be widely available, right?
Ms. Scassa: Well, some of it will be proprietary, some of it will be shared with the government and some of it will end up on the dark web.
Senator K. Wells: Pay to play.
Ms. Scassa: Yes.
Ms. Pybus: Can I add one point? It is really important to — I’m not a lawyer, so I should preface that. That’s Teresa Scassa’s job. But when we think in terms of how these systems operate, we’re used to thinking about privacy as this one-for-one. Jennifer did this, that company knows this about Jennifer, and then it will target Jennifer. Actually, with artificial intelligence, it’s not at all how it works. Yes, they will know what I’m doing based on my activity within that application, but as soon as they start to make features about me, then they’re going to decide, “Well, how do we understand who Jennifer is? We’re going to combine her feature with everyone else’s feature that is a little bit similar, and then, based on this, we’re going to come up with a new insight about her, and then we’ll use this to profile her.”
So, really, privacy, instead of a one-to-one, is actually a one‑to-many and a collective problem because the more data that these companies collect, the better they can understand who I am, because the targeting is knowing me through knowing everyone else.
It’s kind of a philosophical problem, but it’s actually a very real problem in terms of privacy law.
Senator K. Wells: We’re seeing that expression right now, and it’s been brought up in Parliament, which is algorithmic pricing. So we’re facing the same product, but because of the profile they have of us, that now a different price is shown to that individual without seeing a false choice, not knowing anymore what the actual real price is.
The Chair: That’s time. So if there is time, we can comment on that in the second round.
I’m now going to me. Okay.
Professor Pybus, your comments, particularly around AI and health care advice and the 20% that you mentioned, make me think about issues around accountability. We know, in our current medical paradigm that we live in, doctors actually pay a bunch of money for health insurance and insurance to protect them around issues of liability, whether that’s through the hospital or the doctors themselves or both. But if AI gives you misinformation or wrong information that could lead to health implications, where do issues around liability stand with that? Is that even a question, and is that something to consider?
Ms. Pybus: This is a really important question to put on the table and to think about.
You can take an example like AMIE. “AMIE” — so it is a friend, I guess — is being developed by Google. It is a diagnostic tool that they’re working through right now, and it will be used with telehealth, and a patient will come and it will diagnose the patient. Apparently, according to the talk that I went to, by the Google representative in Toronto, AMIE was doing the best at diagnosing patients; better than doctors.
But the good thing, at least, we can say with AMIE is that it’s being developed with doctors and it is being clinically tested in hospitals, not left in the wild, but there are a lot of physicians who are part of this discussion.
On the other hand, we can turn to Oura Ring, which actually many of the women in our survey are using. The Oura Ring just announced that they now have a brand-new women’s health platform that will be fuelled by an AI agent. When I looked into it, there are no health care professionals that I can see that have helped to create this. They’re more venture capitalists. They used to be Finnish. Now they’re an American company that has put this out.
Similarly, if you go and look at the custom versions of ChatGPT that exist, there are already a few GPTs on women’s health, promising, again, specialized women’s health services and medical advice.
I think this is where this question of liability really comes into play, because what happens when you have this increasing number of Canadians who go into these digital spaces? They feel safe and protected because they have the Oura Ring. This seems like a reputable company, but then where are the guardrails to assure that its users are not getting misinformation and that this data is not being misused. From the PIPEDA perspective, it’s social health data, and because they’ve agentically decided to go and find out about their body, it’s not considered as clinical data but considered as any other data. So it can be used for advertising, it can be shared and it can be processed anywhere, and all of those kinds of protections that would be afforded if it was in a clinical setting disappear.
There are a couple of problems or policy issues there. One is should social health data be considered like any other data? I would argue no, that it’s very intimate, and people are giving information that they would only tell their doctors. How do we get a hold of all of these different third-party private entities that are putting out all of this advice? How do we know if they’re actually safe and well tested compared to the kind of frenemy Google AMIE perspective, where at least there are some doctors in the room who are helping to ensure it is safer.
The Chair: Thanks for that response. Professor Scassa, do you have anything to add?
Ms. Scassa: Liability is going to be a really important issue in this space, and it’s a very complicated issue because, even in the regulated health care context with the use of AI, it can be complicated to think in terms of who should be held liable if something goes wrong. Is it the hospital for their implementation of the tool? Is it the doctor for their use of the tool? Is it the AI company for their development? Was it something in their data? Then, of course, you get into issues of transparency and the ability to get access to that kind of information. So there are going to be some really interesting liability questions.
But liability can be a very blunt tool for regulating AI, but we are starting to see its effect in the context of social media platforms, where we have failed to regulate and we’ve struggled to regulate, and now the big lawsuits are coming along with respect to the harms that have been done, particularly to children. With those lawsuits comes a change of heart, a change of practice and platforms changing their behaviour.
It’s a terribly slow and backwards way to protect people and to solve problems, but we have a history of seeing liability issues step in and make a difference to companies’ bottom lines, and therefore drive change in policies and practices. It’s going to be a complex piece, but it’s going to be an important piece as well.
[Translation]
Senator Hébert: I have a question for Ms. Pybus.
You talked about the health sector. That raises a whole series of issues, because it’s quite representative of what’s happening in the economy — I mean, in a number of communities, access to health care is difficult and there isn’t always a doctor available. Take Quebec, for example. The health care system is particularly problematic and not very accessible.
Several years ago, I went through menopause. I might’ve liked to have had access to ChatGPT at the time, although I understand the associated risks you pointed out. You’re absolutely right that we need to be careful about the parameters.
Some witnesses suggested earlier that relying on a system similar to the ESG standards, used in other sectors of the economy, would be a way to solve the problems posed by AI. The health sector is affected, but it could apply to other sectors.
Actually, these standards take into account the environment, the social aspect and the governance system. In this case, for example, we could look at the environmental footprint.
Regarding the social aspect, we could look at the impact on humans, the fact that access to unbiased care is provided, or the existence of a form of personal data protection.
As for governance, the system might also apply to several sectors. You said that in the health field, for example, development at the clinical level with doctors could be a requirement, with human supervision.
Thinking of a system to regulate AI, would ESG standards, adjusted depending on the sectors, be an interesting avenue to explore?
[English]
Ms. Pybus: I will do my best to answer your question. Thank you. I agree with many things that you are saying, and I think this is actually why artificial intelligence, on the one hand, is quite appealing to a lot of people because if you don’t have a doctor and you’re situated in a more rural location, it’s quite challenging to get access to good health care. Here is the amazing solution for you. It’s quite interesting. Some of the women make comments saying, actually, I like talking to the chat model because it’s quite polite. It’s very understanding. There is a kind of emotional, affective connection they have with it. That also corresponds with the Google study that found that with Amie, when they interviewed people about their experiences with it, they said that Amie was much more polite than all the doctors they talked to.
So there are some things to learn there, but what should be taken as important is that there are aspects of empowerment that people experience in these kinds of solutions. I don’t want to categorize it as bad or that we shouldn’t pursue it, but if we do, then we need to be quite careful.
To your point as well, this may actually be a role for the state: to have their own AI or built into health care, wherein there are, at least, those safety and guardrails that are built in from the start, where there is at least some oversight and something where people can stand up and say that there is some oversight to ensure that this is less harmful to Canadians versus going to a venture-capitalist-created AI agent and having that conversation there.
There is so much in what you said, and I was trying to capture it all, but needless to say, there are some opportunities there, but it needs to be carefully looked at.
Senator Hébert: What about an environmental, social and corporate governance system in terms of an analogy?
Ms. Scassa: There is merit there. One of the challenges when we talk about AI is that we talk about such a broad range of different things — from the wild west of whatever the platform companies are doing and the products and services that they are pushing, to things that are being very thoughtfully and carefully developed in the context of the health care system for health AI.
There is this very broad spectrum, and one of the challenges is that the more thoughtful development of systems — because we have the capacity; we’re capable of doing this and applying those principles and developing AI, and there are people working on these systems, typically, within the health care system, but although we’re capable of doing that, the other things just roll along, and the for-profit AI health care business rolls along and often steamrolls over those types of principles.
We’re in a very challenging point right now in history, where we have this incredibly powerful technology, and we have many people who know that we can do great and important things with it and are trying to work on those sorts of things quietly and carefully and are paying attention to these governance issues. At the same time, we have innovation going full steam ahead with very few restraints, pushing to dominate the marketplace. One of the challenges that we’re facing is what’s going to come out on top, and that’s where all of us have some very real concerns.
Senator Hébert: Are you suggesting then that we develop something different for the health care system?
Ms. Scassa: This is already work going on within the health system around appropriate health data governance and about the development of AI. A lot of really encouraging and interesting innovation is happening in the health care context, but there are also very significant challenges in scaling it beyond a single hospital or a single research study. This is something that some governments — provincial governments in particular — are turning their attention to: How do you take this innovation for good and get it to scale? That’s going to be one of our challenges.
Senator McPhedran: Thank you very much. One of the comments made earlier — I’m pretty sure it came from you, Professor Scassa, but I have a sense that Professor Pybus would probably agree with it as well — was about the importance of building literacy and competency in communities. I want to take this very laudable goal a little further by looking at the practicalities for individuals and communities to actually enforce their rights should we ever get around to defining them clearly in law, in relation to AI, because — and this goes to liability as well — the kinds of resources you have to have in order to claim your rights are essentially through litigation. We’ve seen this in the human rights world for a long time. If you don’t have state‑funded, effective mechanisms that actually allow for enforcement on the claim of rights, they are on paper or on the screen.
Are there any thoughts on the practicalities of building in ways and means for this kind of litigation strategy to occur?
Ms. Scassa: That’s a really good question. I mentioned earlier that privacy commissioners across Canada have been doing really interesting work around AI and AI governance through their privacy regimes and their privacy legislation, and they’re going to see more of that put on their shoulders through amendments to legislation.
One of the concerns with things like privacy commissions and human rights commissions as well is that these are government-funded bodies that do a lot of very important work in the trenches around important human rights issues, but their budgets do not keep up with their mandates and with the things there are being imposed on them. Certainly, in the context of the burden being carried by privacy commissioner’s offices around AI, this is going to be an area where they will need more resources, but for human rights commissions, as a result of the complexity of determining bias or discrimination in algorithms, they will really seriously need more resources as well as legislative amendments to carry that burden.
There is the publicly funded ombudsperson privacy human rights mechanisms, where we need more investment, but I’m not sure we’re going to see it, and that’s going to be critically important.
Then there is the litigation side. Class action lawsuits and privacy have really taken off in ways in which some people — well, it is what it is, but it has certainly attracted corporate attention. Privacy class actions are now big business that motivate companies to be more careful with our data to the extent possible.
Rather than human rights-based litigation, we may see these class action lawsuits being brought against corporations, where even if it doesn’t go to court, just the threat of a lawsuit, the cost of defending and the settlement costs are going to get the attention of companies and make them more interested in better AI governance. It is not ideal. It is clumsy. It is expensive. It happens after the harm has been caused. It’s not the best way to do things, but that may be all that we are left with.
Senator McPhedran: And, of course, a class action does not preclude strong human rights arguments as part of that.
Ms. Scassa: No, it does not. But that really does have to be supported and funded because it will be extremely challenging and expensive, and there will be all sorts of barriers to overcome around trade secrets, confidential information and causation, which will be very difficult to establish. It’s going to be very challenging litigation.
Senator McPhedran: Any thoughts on that, Professor Pybus?
Ms. Pybus: Not specifically in terms of litigation but in terms of the literacy point that you raised before, the only thing I would add is that there is a challenge as well. We have done a lot of futures cone workshops, so basically getting people to imagine different futures based on profiles they have created with some of the data we have accessed in our audits. I think one of the biggest points that has come out of those is how grateful people are to just be able to have a space where they can imagine how tracking works and how profiling works, so they can actually imagine what their futures might actually look like and unfold.
I think one of the key sites for literacy is how to help people imagine how these spaces are actually working so that they can start to come up with and engage them in a more kind of agentic capacity.
This is slightly on the adjacent side, but just based on our own research talking to a number of people, we don’t really know how any of this works; I don’t understand it; I press the buttons; I know I’m supposed to put in prompts, but I really have no clue how any of these systems actually work.
I think there is a desire by people — certainly the ones we have talked to — to have a clearer understanding so they can understand the stakes and their own positionality within these systems.
Senator McPhedran: Thank you. Thirty seconds left, and written answers are welcome. Can you foresee the possibility that we would end up with state-established endorsements for safety, like a Better Business Bureau kind of rating, to help people actually make decisions about what they should risk giving their information to?
Ms. Scassa: One of the things that, for example, in privacy law reform I think the government has shown some interest in — and I showed it in the artificial intelligence and data act as well — is this idea of setting standards and having companies certify themselves against those standards. That’s a more private version of that in the sense that the government provides the legislative infrastructure for the adoption of standards and then for the consequences that follow from certification, but that that becomes a private thing where companies conform to standards and then seek certification.
So I think we are more likely to see something like that with the certification bringing the trust factor so people can choose companies that have been certified as compliant with certain good governance standards and so on. So we may end up moving in that direction. Of course, the artificial intelligence and data act is off the table now, so who knows, but there may still be standards and conformity assessments supported in part by the EU Artificial Intelligence Act and the desire of Canadian companies to conform to those norms to have access to those markets as well.
The Chair: Thank you both so much for your very compelling presentations and responses. It will certainly go a long way in supporting the work we are doing in terms of our deliberations and when we get to report writing. Thank you very much.
Our final panel of witnesses have been asked to make an opening statement of five minutes each. This will be followed by questions from the senators.
With us by video conference, we have Mark Daley, Chief AI Officer & Professor, Western University. Welcome. Joining us by video conference, please welcome Evelyn Forget, Distinguished Professor, University of Manitoba.
Welcome to you both. I invite Professor Daley to make his presentation, followed by Professor Forget.
Mark Daley, Chief AI Officer & Professor, Western University, as an individual: Chair, honourable senators, thank you for the invitation to appear.
For nearly all of human history, the only source we have had of high intelligence has been other human beings. Intelligence was embodied: It arrived in a person, with a history, a culture, a family, a salary, a need for rest, and rights. We could educate it, hire it, credential it, move it across borders and, all too often, waste it. But we could not manufacture it at will. And that is what has changed. In creating machines that think, we have commodified intelligence.
I want to choose my words carefully here. I do not mean to say that we have commodified wisdom, conscience, responsibility or judgment. We have not. But we are beginning to commodify a narrower thing that is nonetheless economically profound: cognitive capacity. The ability to draft, code, search, summarize, plan, test, reason across documents and carry a task forward, that capability can now be rented, metered and replicated at scale. This is a new fact about our world.
There is a reason some of us feel the ground shifting, and it is not merely that the demonstrations are more polished this year. Over the last decade, an empirical pattern in AI has been remarkably consistent: When we scale computation, systems acquire new capabilities.
The breadcrumbs are no longer small. In mathematics, for example, my own discipline, AI systems have recently produced solutions to long open problems associated with Paul Erdős, one of the 20th century’s great mathematical problem posers.
This same threshold is visible in more ordinary work. AI agents are different from the familiar chatbot model. They do not simply answer and stop. They can inspect files, run software, edit code, test their own work, use tools and return to you with results.
Why does this matter for economic security and human rights? Because labour markets do not pay for abstract intelligence. They pay for bundles of tasks. When parts of those bundles become available as a cheap, scalable, rented input, institutions will reorganize around that fact.
In plainer language: once intelligence becomes a commodity input, the market will try to arbitrage the expensive human parts of knowledge work.
The first human parts under pressure will not be the most senior ones. The vulnerable layer is the apprenticeship layer, such as the junior analyst, the entry-level developer, the policy officer or the administrative coordinator. These are the roles in which young Canadians learn by doing the routine parts of a profession under supervision. If those tasks disappear, the profession may remain, but the pathway into the profession narrows. The first sign may not be dramatic unemployment spikes. It may be a quieter absence: the job that is never posted, the junior cohort that is never hired.
The right to work is not a promise that every task will remain untouched by technology. History has never offered us that bargain. But our social contract demands a meaningful pathway into decent work, skill formation, economic security and social participation.
The burden of this will not be evenly distributed. The International Labour Organization, or ILO, estimates that in high-income countries, jobs at the highest risk of automation from generative AI account for 9.6% of women’s employment, compared with 3.5% of men’s employment. This is not a random disruption. It has a social shape.
So the question before Canada is not whether we should adopt AI. We will, and we should. The question is: On what terms?
The commodification of intelligence is not merely a technological event. It is a labour-market event, an educational event and a human rights event. If we wait until the damage is obvious in the aggregate statistics, the ladders into knowledge work may already have been quietly withdrawn, leaving Canada’s youth stranded at the bottom.
Thank you.
The Chair: Thank you, professor. We will now go to Professor Forget.
Evelyn Forget, Distinguished Professor, University of Manitoba, as an individual: Thank you very much. Honourable senators, thank you for the invitation to appear. I would like to make three central points about artificial intelligence, human rights and economic security in Canada.
First of all, I do not think that AI is likely to lead to a reduction in overall employment, but it will affect different industries differently. Total employment will very probably increase, but the kinds of jobs on offer will change significantly, and the transitions will be both protracted and difficult. The issue is not mass unemployment but shifting skill requirements, more job transitions and greater income volatility. We will very likely see large-scale structural unemployment alongside significant shortages of skilled labour in some fields. All this, of course, exacerbates inequality.
Some workers will navigate this successfully, and others will struggle, particularly mid-career workers who face the greatest barriers to retraining. Many have limited literacy or numeracy skills, while others are deeply embedded in sectors that are being restructured.
We can already see this in administrative and clerical work and in fields like computer coding. Routine tasks are being automated, while those that remain require higher degrees of digital fluency. Firms hire new workers with these skills while shedding entry-level jobs and experienced mid-career workers who are more difficult to retrain.
My second point is that neither the harms nor the benefits of the transition will be evenly distributed. Those people who are already most vulnerable, including some people with disabilities and those with low literacy and numeracy skills, will face the greatest disruption. In many ways, AI will exacerbate already existing inequalities in the labour market.
AI will affect these groups in two very different ways. The first is through the labour market, where they are more likely to be in roles that are restructured and made more contingent. But the second is through access to the state itself. As individuals lose or cycle through employment, they become more reliant on public systems for income and support. However, public services are also increasingly mediated by automated systems, including eligibility screening, compliance monitoring and digital-first delivery. For individuals with limited literacy, cognitive challenges or unstable living conditions, these systems can become additional barriers to overcome.
This creates a form of administrative exclusion. People are not denied support explicitly; they are unable to navigate the systems that provide it. And, once again, these demands fall most heavily on those people least able to manage them. Those without the time, resources or knowledge to appeal decisions are most vulnerable. The systems that are supposed to help in such circumstances may actually reinforce existing patterns of disadvantage.
Third, our current social supports are not well designed for the transition we are facing, let alone those just on the horizon.
Employment Insurance assumes stable attachment to the labour market and clear transitions between employment and unemployment. It does much less well when we’re talking about contingent work. Social assistance is inadequate, conditional, slow and complex. Retraining programs are often difficult to access and assume that individuals can absorb the financial and cognitive costs of participation.
To my mind, we need structural reform rather than incremental adjustments to ensure a more equitable sharing of the benefits and harms of AI.
One possibility is a guaranteed livable basic income, which should be understood not only as an anti-poverty measure but as economic stabilization that provides continuity in the face of disruption. It would allow individuals to navigate transitions without falling into crisis.
AI means that workers may need to re-skill multiple times over their working lives, and a guaranteed livable basic income could provide income support during these transitions. Whatever system we put in place, we need to reduce administrative complexity through automatic enrolment and simplified rules.
One benefit of AI is it could replace some of the special programs that various levels of government like to create for very specific groups of workers whenever a crisis emerges in a particular industry. These narrowly targeted programs are difficult to predict and quickly become obsolete and almost always leave out some of the people most in need.
As AI transforms the economy, the central question we face is whether our social architecture will evolve alongside it to recognize and respond to need in a world where both work and access to the state are increasingly mediated by machines.
Thank you, and I look forward to your questions.
The Chair: Thank you both for your presentations. We will now proceed to questions from senators. Senators, you have five minutes for your question, and that includes the answer.
Senator Arnot: I would like both the witnesses to address this issue. Is AI more likely to increase or decrease economic security for vulnerable populations? I think Professor Forget has already answered this question, but are existing social supports sufficient to absorb AI-driven disruption? Is universal basic livable income the only policy intervention that will most effectively deal with economic security, or are there other components that you might think are applicable?
Ms. Forget: I am very much a supporter of guaranteed basic livable income because it doesn’t require a great deal of forecasting to address rapidly changing economic conditions. It provides support for everyone who needs it when they need it. If appropriately designed and implemented, it can serve as a useful backstop.
Is it sufficient? Absolutely not. We need to spend a great deal more time and energy thinking about retraining programs and providing those opportunities for people. Our education system, in general, has done a very poor job of preparing new graduates for the kinds of transitions they will be facing.
Professor Kwan talked earlier about this being the Fourth Industrial Revolution. If you think back to some of the previous industrial revolutions, we are facing similar kinds of issues. One can go as far back as the 14th century looking at the agricultural revolution. Technological change has displaced huge numbers of people and brought about massive changes in society. Similar kinds of things happened in the 19th century. My sense is that simply making incremental changes to programs that were built for a very stable industrial economy is not sufficient.
Basic income is a challenge for many people to think about, but there are many ways we need to think about how to bring progressivity back into our tax and transfer system. If we look at OECD countries, for example, inequality in Canada is not particularly bad compared to other countries before the tax and transfer system, but many other countries, many European countries in particular, do a much better job of redistributing income, so after taxes and transfers we are among the worst in the OECD. These are the kinds of things we need to look at: how we deal with the kind of inequality that will emerge, particularly if AI exacerbates existing challenges and increasing inequality.
Mr. Daley: I agree entirely with my colleague on all points. I would just add that part of the right to work is right to purpose. For better or worse, we find purpose in our work in Western society. Universal basic income I see as potentially necessary as an intervention but perhaps not sufficient.
Senator Arnot: Thank you.
Senator McPhedran: One of the challenges that we face as lawmakers at the national level is exactly that: we are at the national level. Much of what we are hearing from you and other experts involves remedies or protections that require a high degree of cooperation, and the division of powers needs to be not so divided.
Professor Forget already knows I’m a strong supporter of her work and also Senator Pate’s bill on livable income, but I’m also very caught up with, Professor Daley, your comment about commodified intelligence.
You both have raised such powerful issues that I’m still thinking as I frame this. I guess you can leave it to us to try to figure out what we can and cannot do at the federal level, but how do we as a society go forward with the knowledge that if we don’t commodify the intelligence of many people in Canada, we will probably suffer for that? We talk a lot about guardrails. I’m just wondering if either one of you felt like venturing into that territory, both in terms of the general population but I’m also really interested in youth and the learning of youth and the capacity to work, the opportunities to work and how we try to facilitate that.
Mr. Daley: I’ll start by addressing youth in the context of education because this technology is incredibly powerful for education but also could be incredibly deleterious. The French philosopher, Jacques Derrida, wrote about pharmakon, which is a Greek word that means both poison and remedy. He was writing about Plato’s writing on writing and what writing does for memory. If I write stuff down, then I can remember more stuff. But if I write it down, I’m also likely to forget it, but you can’t have writing without the good and the bad.
We have technology now with high intelligence on tap for our youth, and that gives them the opportunity to have a one-on-one tutor on any topic they want, which was unavailable to all but the wealthiest five years ago. However, they also have the opportunity to say, “Could you just write my Philosophy 101 essay,” and so the actual change in the education system isn’t just about adopting technology. It’s about talking about values. If you want to use this technology to cheat yourself out of an education or livelihood, you can, but you can also use this technology to tutor you, to teach you things that otherwise you wouldn’t have access to.
It comes back to fundamental humanistic inquiry. Who do you want to be? Are you using this technology to help you become who you want to be? That’s true in the education setting and is true beyond in the labour market.
Senator McPhedran: To build on that for second round would be, again, the guardrails question. Do we want to be regulating the exact choices you have just summarized for youth?
Ms. Forget: I think we’re all struggling as educators with exactly these problems right now. I was very intrigued by the commentary earlier about embedded ethics and the engineering program. I think we’re seeing similar kinds of things in health care. I teach in a college of community and global health, and many of my colleagues are very focused on ensuring that health care is available, particularly in remote and rural communities. So there is very much a focus on how AI can be used both within the classroom setting and in terms of the delivery of health care to actually bring about positive changes.
Now, we’re dealing primarily with graduate students and people well along in the educational process, and I’m not sure how I can take that back further and make comments about what happens at lower levels of education and how we bring that about, but these are things that we’re all struggling with, and I certainly don’t have answers to it.
I have a lot of curiosity, and I have a lot of curious students.
[Translation]
Senator Hébert: Mr. Daley, you talked about the risks to the labour market, particularly for young Canadians or women holding entry-level jobs, more precarious jobs, or jobs that require less training.
Canada has a significant labour problem. We know there are labour shortages in many regions and industries. We’re also trying to increase the productivity of our businesses, because we’re lagging far behind. The various levels of government are facing the same issue, where citizens are asking for more productivity while maintaining the level of service. AI can be used for that. How do we reconcile all that? What you said is scary. My goodness, where are we heading? How do we reconcile all this?
Mr. Daley: Thank you for your question.
[English]
I don’t mean to be scary, but I mean to be honest. We do have machines that think, and they don’t replace humans entirely. There is much more we do than think. But it is one of the fundamental conceits of the Western world that humans are smart, and that’s what makes us who we are. Bringing that into a society where we’re so conditioned on that is a challenging endeavour. It’s by no means impossible. It doesn’t have to be scary. It’s opportunity. What the opportunity requires is humility from us.
I’m a theoretical computer scientist, and GPT-5 Pro is better at theory than I am. That was a hard moment for me, but once I get past that, I say, okay, great, I can outsource; I can use it to teach me more. We need to see this as an empowering technology — it’s not replacing me as a human, it’s replacing some of the things that I used to have to do solo. It’s augmenting me. You can see this across the labour market.
When you talk about productivity, having a cognitive assistant, if you’re using it well, if you’ve been educated and trained, and are thoughtful in how you use it, it cannot help but increase your productivity in a way that having a team of 10 employees used to be the only mechanism for increasing your productivity in that kind of way. You can scale that across government services and across the private sector, and we’re seeing that sort of adoption happen right now, and we’re just really in the early days of it.
[Translation]
Senator Hébert: We heard from experts on a system similar to ESG standards. Relatively speaking, that may be in keeping with the philosophy you just articulated. What are your thoughts on that? When you think about it, it makes sense. AI has environmental and social impacts, as well as governance impacts. Is this an idea you find interesting? Is it an avenue to explore?
Mr. Daley: Yes, I think it’s a good idea.
[English]
It’s about creating sets of rubrics and shared values that reflect what Canadians and Canadian society want. So, in the case of ESG, it’s up to each individual corporation and entity to determine how do we reflect our values and the decisions we make? AI is a complicated enough technology that it makes sense to take this multifactorial approach. I think this is incredibly well advised.
The Chair: I have a follow-up on that, Professor Daley, with respect to the ESG question, which really is a voluntary tool. In my experience utilizing it in my previous life, it was a way that corporations could say that they’re good corporate citizens.
Mr. Daley: Yes.
The Chair: But also driven internally by employees choosing companies to work for, based on what their ESG commitments were.
That’s kind of on a more even playing field without the impact of AI, but with the impact of AI, does this just create a whole new ball game?
Mr. Daley: I think you’re going to see that in terms of, especially Gen Z, and if you look at sentiments toward AI, it’s fallen markedly in the last two years. They’re going to be looking for employers who share their values. Customers are going to be looking for companies that share their values. So you’re right, ESG is not regulated. It’s voluntary, but we have a market, a labour market and a consumer market that holds companies to accounts, and I think we can accomplish the same with AI.
The Chair: Thank you.
Senator McCallum: This is a difficult topic for me. I don’t really actually know much about AI, but I do read a lot.
The plan to reshape the U.S. government and implement far‑reaching policy changes across multiple sectors with an emphasis on deregulation could dramatically reshape the tech sector. From the rescinding of Biden’s AI Executive Order 14110, which imposed safety requirements and oversight on AI development to the reversal of Obama-era net neutrality regulations to the antitrust investigations that reduced scrutiny on big-tech companies, to the executive order that could jeopardize the EU-U.S. data privacy framework that currently enables transatlantic data flows in compliance with the EU’s general data protection and regulations, collectively these changes favour U.S. tech giants at the expense of societal and environmental concerns in the U.S. and Canada and around the world.
The relaxation of AI regulations in the U.S. could put Canadian AI and tech start-ups at a competitive disadvantage.
While Canada maintains some minimal oversight mechanisms to ensure ethical AI development, U.S. companies may accelerate innovation with fewer safeguards, which could make it harder for responsible AI companies to compete on a global scale.
What are the various ways this will impact and influence Canada?
Mr. Daley: That’s an incredibly insightful question, senator. This is a geopolitical question. There is a race between the United States and China right now in terms of developing capacity. We can see that, right now, the U.S. is openly using frontier AI models in the prosecution of war, and so this is more than just a consumer issue. This is truly an international issue.
You’re right. The removal of any kind of speed brake or regulation has profound implications for that race between those two superpowers and Canada as a middle power. Our Prime Minister has written and spoken eloquently on variable geometry, where we work with like-minded nations who have like values as middle powers, and collectively, we have the ability to bargain in ways that we don’t individually.
I think the best remedy for Canada lies at the global level, at the level of international relations, and finding those allies. We’re building those allies with European nations and others who share our values and whose combined economic power is enough to stand up to superpowers. What we can’t do is modify the lack of regulation in other jurisdictions. We need to figure out how to deal with that, and there is no easy answer.
Senator McCallum: Ms. Forget?
Ms. Forget: I’m not sure I have anything to add to that. I find that a terrifying situation, and I agree absolutely with my colleague.
Senator Arnold: Thank you both for being here today. I can’t say it’s terribly uplifting, but we appreciate your input very much.
I’m assuming you work daily with grad students or students of some sort, and our last witness made reference to students having a hard time right now. I’m wondering: What do you say to your students? What kind of advice do you give them on their career selections right now?
Ms. Forget: I work primarily with graduate students where I am, and they are focused on applied questions within the Canadian health care sector. I think many of them take AI quite seriously in their doctoral work and in their graduate work, generally. In most cases, they’re further ahead of me in terms of what it is they understand and where they see it going.
I’m not sure my advice is particularly useful to them at this point, but I see a lot of engagement. I see a lot of engagement with these students and a lot of recognition that there are real potential benefits to AI down the road despite the challenges that are undoubtedly part of it.
As was mentioned earlier, there are parts of the country that are facing great shortages in the labour market, and AI is a way of addressing some of those issues. There are certainly areas of the country that are facing inadequate health care and inadequate social services, and AI is a way of bringing about those changes.
We do require governance, no question about it, but I think, for the most part, my students see this as potentially beneficial moving forward. They’re reasonably optimistic. It’s not something that terrifies them. It’s part of the world that they’re preparing for.
Mr. Daley: I agree entirely with my colleague about students at the graduate level, so I’ll speak about the undergraduate level. I would be skeptical of anyone who tells you what the world will look like in terms of the best jobs 5 or 10 years from. I certainly have no idea.
The advice I give high school students and undergraduates who ask “What do I do?” is do something you love because then at least you’re going to be happy. I can’t tell you what the next top 10 jobs are going to be. No one really knows, so study something that matters to you. Study something that you’re passionate about. Learn how to learn. Learn how to adapt. Learn how to think. That’s honestly the best advice I can give right now.
The Chair: Professor Forget, in your opening remarks, you mentioned the importance of the fact that current social supports are not aligned with what’s coming, and you also talked about the need for structural reform. Would you like to expand a bit more on what structural reform could look like?
Ms. Forget: My own personal interest would be in a much more far-reaching tax reform. We need to rethink the way that we tax income and wealth in this country. I think we need to think about how we deliver social programs, and that’s all I mean at this point.
I certainly have my own preferences for how it would unfold and what would happen, but what I think we really need is a broad conversation. I think that as a country, we have shied away from those big questions for a very long time. I think we’re recognizing that what we have does not meet current needs. However it comes out in the wash, so to speak, I think we need to talk about it much more thoroughly.
Mr. Daley: I agree with everything that my colleague has just said.
The Chair: We should have such agreement in both the houses.
Ms. Forget: Just suggest a tax reform commission and watch it not happen. The agreement, I mean.
Senator McPhedran: I want to go back to some of what we discussed before, but I don’t think any of us have asked you if there are any specific protections. I understand the uncertainties, but perhaps, in drawing from other experiences, you could discuss the sorts of mechanisms you have seen work in helping to protect people who have a strong need for information and a strong need for a future, for options, and not a lot of resources to winnow what is safe and productive for them.
Ms. Forget: Whom are you asking?
Senator McPhedran: Whoever wishes to answer is most welcome to start.
Mr. Daley: It’s a difficult question, so we’re looking at each other to say I don’t want to take that one.
I don’t know. I’m a computer scientist, and you’re looking for structural social interventions at a scale that we haven’t seen in a couple hundred years. And as my colleague pointed out, we didn’t handle the last Industrial Revolution as well as we might have if we wanted to put human flourishing, in the Aristotelian sense, at the centre, but for me, when I look at my duty to our students here, that’s the core.
I want students to come here and flourish. I want this to be a place where they can explore, experiment and learn, and I want technology to aid in that, but I want it to fundamentally be human-centric. So I can’t give you the exact structural policy interventions. You need an economist for that. But I can give you my value-based take on it is if we centre the human being and human flourishing, I think we can work backwards from that to mechanism design, but that’s probably a job for many people.
Senator McPhedran: Forgive me if this is a bit too personal, but do you have thoughts or have you taken a position on the whole concept of a basic livable income to allow people to, as you pointed out, pursue what they love?
Mr. Daley: I am in favour of that type of economic intervention, but I think we have to go beyond that. People need purpose. We have a fundamental duty to our citizens to ensure that they can live and that they can flourish, but they need more than just income, and that’s the tricky part.
Ms. Forget: There have been many economists who have asked that question over the years. You can think of John Maynard Keynes and John Stuart Mill, who were asking precisely that question: As the world advances and as technology changes, how do we educate human beings to find meaning in their lives?
Some very smart people in my life told me that there are two kinds of people in the world: There are people who work to live and people who live to work. I think that all of us sitting in this room, all of us participating in this meeting, have been fortunate because we have the kinds of jobs that give meaning to our lives. These are things that are really important to us. I don’t want to denigrate other kinds of jobs, but I have certainly done some kinds of jobs in my life that did not create great meaning in my life, but they provided me with an income that let me live my life and find meaning elsewhere.
I think part of what we’re doing as a society is recognizing that we’re not just people who need to work; we are people who need to live, and we need to go far beyond the jobs we do to find that kind of meaning in our lives.
When I focus on a basic income, it’s not about giving people meaning; it’s about giving people enough stability to think about how they can live their lives. I think that’s part and parcel of what we should be doing within the educational system, to take people beyond that and to let people find and create meaning in their own lives. It doesn’t have to happen in a nine-to-five job or in jobs that aren’t fortunate enough to be nine-to-five jobs.
Senator McPhedran: Super quick last question, which is: Do either one of you believe that in Canada we have a legally defined right to work?
Ms. Forget: I don’t believe we do.
Mr. Daley: I don’t know.
Senator McPhedran: We don’t, because we have no social and economic rights whatsoever in the Canadian Charter of Rights and Freedoms. But it has been interesting how many witnesses have referenced this right to work. Thank you.
The Chair: Professor Daley: some of us around the table are on other committees looking at AI from different angles and have heard much scarier testimony than what you have given us, because you talk so much about values and meaning. I recall at least one or two witnesses in their scary testimony talking about how the advances in AI are really being driven by not necessarily putting humans at the centre but putting money at the centre. I think this is where it becomes more scary because — contrary to what someone said at this table. I think it was you, Professor Forget, who mentioned that you don’t believe that AI will necessarily replace workers. But what we’ve heard in other settings is that that’s the intent if money is at the centre.
I wonder if you had any comments on that.
Ms. Forget: What I meant was that the total number of jobs after the process of AI begins to unfold will increase very dramatically. I certainly didn’t mean to say that people wouldn’t be displaced. I think that’s part and parcel. People are being driven out of jobs that currently exist, but new jobs are being created, and that’s why we have simultaneously or will have simultaneously large-scale structural unemployment and labour shortages.
The challenge is getting these people over here retrained for these kinds of jobs. It’s certainly much easier to train some people than it is others — to train young people, for example, rather than 50-year-olds or 45-year-olds to take on new kinds of work.
But I think the total amount of employment, the total amount of output and the total amount of everything will increase with AI. I think it’s going to be a tremendously productive outcome. So I don’t think we’ll see massive unemployment.
The Chair: Thank you. Professor Daley?
Mr. Daley: Madam Chair, I think I can imagine which of my colleagues you are speaking of, and I can hear the testimony in my head. To a first approximation, they’re not wrong. They know their business. This is their world. But I don’t feel the same sense of doom. What I feel is a sense of agency.
This is a critical moment right now. If we don’t make the right decisions, then it’s possible we end up in the world they’re describing. If we just sort of step back and allow the economic incentives, as you have said, if capital alone drives this — and we’ve set up a system that makes that easy, especially in Silicon Valley. Everything runs on venture capital. If you allow that system to run to its conclusion, you may get something that looks like a cyberpunk dystopia. And I love cyberpunk fiction — William Gibson is great — I read it as a child. But I read it as a cautionary example, not a how-to manual.
There are other ways of looking at the world and other ways of governing this technology. I think Canada is forming coalitions with like-minded nations and forcing this on the world stage, having conversations around standards because both the U.S. and China want to sell to the rest of the world. There are places where we can exert our power with our allies to push toward a world that is human-centric rather than one that is purely capital‑centric. But those are choices and policy decisions that you have to make as our policy-makers.
The Chair: That’s a great way of punting it back to us. Thank you. Again, thank you for your presentations and the very engaging conversation that we’ve had. On behalf of the committee, I want to sincerely thank you for taking the time to appear before us today. Your testimonies will be very helpful as we put together our deliberations.
(The committee adjourned.)