THE STANDING SENATE COMMITTEE ON SOCIAL AFFAIRS, SCIENCE AND TECHNOLOGY
EVIDENCE
OTTAWA, Wednesday, February 25, 2026
The Standing Senate Committee on Social Affairs, Science and Technology met with videoconference this day at 4:15 p.m. [ET] to examine and report on matters related to the impact of artificial intelligence in Canada.
Senator Rosemary Moodie (Chair) in the chair.
[English]
The Chair: Welcome to this meeting of the Standing Senate Committee on Social Affairs, Science and Technology. My name is Rosemary Moodie. I’m a senator from Ontario and the chair of this committee.
Before we begin, I’d like to do a round table and have senators introduce themselves, starting with the deputy chair, Senator Burey.
Senator Burey: Sharon Burey, senator for Ontario. Welcome.
[Translation]
Senator Boudreau: Good afternoon. Victor Boudreau from New Brunswick.
Senator Arnold: Dawn Arnold from New Brunswick.
Senator Petitclerc: Chantal Petitclerc from Quebec.
[English]
Senator Hay: Hello. Katherine Hay, Ontario.
Senator Greenwood: Good afternoon. Margo Greenwood from the beautiful province of British Columbia.
Senator Muggli: Tracy Muggli, Treaty 6 territory, Saskatchewan.
The Chair: Before we hear from our witnesses, I’d like to thank all the senators on this committee for your thoughtful consideration throughout our work on Bill C-12. As our committee has done in the past, we worked diligently and quickly, given the tight timelines we were provided by the government motion. The job we were given was a tough one, and I believe we did it very professionally. I’d like to share my sincere gratitude to all the support staff of this committee.
The work we do in this committee doesn’t just happen; there’s an entire team that supports us. We have interpreters, multimedia technicians and other support staff who need to prepare and support us throughout the busy schedule, so I want to thank you all for your flexibility.
I especially want to give a special thank you to our Library of Parliament analysts who had to draft the report within very short timelines. Even though you did not have a lot of time to compile all the information we had, the report that we tabled was exceptional and a true reflection of our professionalism. Of course, none of this work could have happened without our clerk, Ferda Simpson, who did an exceptional job in getting the witnesses to appear and to be flexible with the minister and her officials.
Colleagues, I’m proud to work with you, and I’m proud of how collaboratively we work as a committee. Thank you again.
Today, the committee begins its study of matters related to the impact of artificial intelligence in Canada. This study will examine issues including data governance and sovereignty, ethics, privacy and safety and the risks, benefits and social impact of artificial intelligence here in Canada.
Joining us today for our first panel, we welcome, from Innovation, Science and Economic Development Canada, Mark Schaan, Associate Deputy Minister, Department of Industry. And from the Treasury Board of Canada Secretariat, we welcome Dominic Rochon, Chief Information Officer of Canada; and Kara Beckles, Executive Director, Privacy and Responsible Data. Thank you for joining us today.
For your opening statements, you will have five minutes, followed by questions from committee members. Mr. Schaan, the floor is yours.
Mark Schaan, Associate Deputy Minister, Department of Industry, Innovation, Science and Economic Development Canada: Honourable chair and committee members, thank you for the opportunity to contribute to your study on the impact of artificial intelligence, or AI, in Canada.
AI presents both significant opportunities and complex challenges. Today, I will be sharing with you information on the current work of Innovation, Science and Economic Development Canada, or ISED, in this space, and I look forward to your questions.
ISED is responsible for the Pan-Canadian Artificial Intelligence Strategy launched in 2017. The first of its kind, Phase I of the strategy invested $125 million and focused on the foundation, establishing the three national AI institutes — Amii in Edmonton, the Vector Institute in Toronto and Mila in Montreal — to anchor the Canadian AI ecosystem.
[Translation]
Through Budget 2021, the government invested an additional $443 million in Phase 2, taking a three-pronged approach to increase the capacity of institutes to commercialize AI research, attract, retain and train AI talent and advance the development and adoption of AI standards.
[English]
Phase II includes a $275-million allocation to Canada’s Global Innovation Clusters to accelerate the commercialization and adoption of AI technologies. This has translated into 94 projects announced across the five clusters, with over 427 project partners and over $372 million co-invested with industry into the Canadian ecosystem as of December 31, 2025.
Phase II also provided an initial investment to the Digital Research Alliance of Canada to provide dedicated computing capacity for AI researchers.
The 2024 Fall Economic Statement provided incremental funding of $174 million to support the continuation of the AI commercialization activities under the Pan-Canadian Artificial Intelligence Strategy. It also announced the launch of a $500‑million program to the Business Development Bank of Canada to provide financing and expertise to help small- and medium-sized enterprises adopt digital technologies, with a priority focus on AI.
[Translation]
Other federal departments are also actively supporting the commercialization and adoption of AI. For example, the National Research Council of Canada administers the $100‑million Assistance IA program, which was launched in 2024 and helps small- and medium-sized businesses with challenges in developing and integrating generative AI solutions and deep learning. The program supports businesses in terms of awareness, planning and implementation so that the technologies are developed and deployed safely and ethically.
There is also the Regional Artificial Intelligence Initiative, with a budget of $200 million, implemented by the regional development agencies. The initiative aims to support the commercialization of new Canadian AI technologies and accelerate their adoption in strategic economic sectors, such as agriculture, health care, clean technology, manufacturing and other regionally important sectors across the country.
[English]
Also, in the past 10 years, the three federal granting agencies — the Natural Sciences and Engineering Research Council of Canada, the Social Sciences and Humanities Research Council and the Canadian Institutes of Health Research — have provided over $1 billion in funding to academic AI research projects.
While Canada’s AI research sector is recognized as one of the world’s best, Canada currently lags behind many of our G7 partners with respect to national high-performance computing infrastructure capacity. Hence, the Canadian Sovereign AI Compute Strategy was launched in 2024 to help address this gap, with investments in public supercomputing infrastructure and expanded access to advanced AI compute to ensure that Canada’s researchers and companies have the compute capacity they need to compete globally. Of this funding, $700 million has been allocated to grow private sector AI champions, $300 million to offset compute resource costs for small- and medium‑sized enterprises and $1 billion to help build public supercomputing infrastructure.
On January 15, the government invited qualifying Canadian firms and consortia to submit their proposals for the construction of large-scale AI data centres designed to serve a broad spectrum of Canadian clients.
[Translation]
Canada’s approach to AI aims to strike a balance between the economic growth and innovation enabled by the technology, with measures to ensure safe and responsible development, deployment and use, while providing support to Canadians as it is increasingly adopted across the economy and society.
Launched in 2024 with an initial investment of $50 million, the Canadian Artificial Intelligence Safety Institute is working with domestic and international partners from government, industry and the research community to advance scientific understanding of the risks associated with the most advanced AI systems. Its work focuses on risks such as deepfakes, cyber capabilities, malicious uses and AI alignment issues.
A range of existing laws already provide protections related to some aspects of AI development, deployment and use. These include the Criminal Code of Canada, the Personal Information Protection and Electronic Documents Act, or PIPEDA, as well as sectoral legislation applicable to specific industries, such as the health and financial sectors.
[English]
The members of this committee may also have noted the Minister of Artificial Intelligence and Digital Innovation’s statements regarding the government’s intent to bring forward modernization to our private sector privacy law, which is the Personal Information Protection and Electronic Documents Act, or PIPEDA, to strengthen privacy protections and ensure that privacy risks with respect to AI are appropriately addressed under Canadian law.
I will wrap up by saying that we are in an interesting moment as we continue our finalization of Canada’s next national AI strategy. We used a national sprint over the course of last year to be able to engage with 11,300 Canadians and 28 task force members. We look forward to bringing forward that strategy soon. We would be happy to discuss that during comments and questions.
The Chair: Thank you, Mr. Schaan. Mr. Rochon, you have the floor.
Dominic Rochon, Chief Information Officer of Canada, Treasury Board of Canada Secretariat: Thank you. It’s a pleasure to be here with my colleague Mark Schaan and a key member of my team, Kara Beckles, who is the Executive Director of Privacy and Responsible Data. Mark Schaan has given you a quick overview of some of the things his department is doing that are outward-facing as they pertain to AI. I’ll quickly touch on some things that are inward-facing in the government from my position as the Chief Information Officer of Canada at the Treasury Board of Canada Secretariat.
I would like to highlight some of the key programs and policies that are being implemented across the federal public service to promote effective and responsible adoption of artificial intelligence.
AI has the potential to open all sorts of new ways for the government to work more efficiently and better serve Canadians, but if it is not used responsibly, it could also have life-changing consequences for the people the government serves.
That’s why the government’s key policies focus on a human-centred approach. In other words, our policies will help us focus on the needs of those we serve and the public servants who serve them in deciding where we adopt AI and how we integrate it into our work.
Take, for instance, our Directive on Automated Decision-Making which we put in place in 2019 and has already been updated four times since then. It requires departments and agencies to undergo rigorous evaluations. This ensures accuracy, fairness and respect for human rights when considering the use of automated systems to support decisions that affect individuals, their rights and their interactions with government. Its goal is to identify risks early and avoid or mitigate unintended impacts or harms.
Another example is the Algorithmic Impact Assessment tool. This helps departments navigate the directive’s requirements. It’s built around a risk-based approach: The higher the risk, the more detailed and comprehensive the safeguards need to be. The guide on the use of generative artificial intelligence gives employees practical tips on how to use AI tools safely, effectively and responsibly in their day-to-day work.
[Translation]
I also want to highlight the recently launched public AI registry, which helps build confidence in the government’s use of AI by showing Canadians how and where the technologies are being used across government. The registry is designed to ensure that AI is used in a way that is accountable, transparent and reliable.
However, central to all this is the federal government’s Artificial Intelligence Strategy. It defines how the government leverages AI to improve productivity, strengthen its capacity for scientific research and improve services to Canadians.
Adopting AI across the federal government also requires great caution and responsibility. That is why the strategy emphasizes the need to deploy AI systems safely and ethically to prevent bias or unintended impacts on people or communities. It also reinforces the need for clear communication about the areas in which AI is being used and the impact of its use on services.
Our goal is to update the strategy every two years to ensure it remains relevant and responsive. The AI strategy is part of a broader modernization ecosystem that involves digital service design, cybersecurity and responsible data management. This includes the 2023-26 data strategy for the federal public service, which outlines measures taken within the public service to ensure that data is a reliable, well-organized, harm-free input and ensures compliance with the measures, which is necessary for AI to operate.
This brings me to the issue of data sovereignty. When government data is stored in the public cloud, it is hosted by third-party providers that may be subject to foreign laws. There is a risk that another country could legally require access to the data. It is important to note, however, that Canada and the United States have long-standing information-sharing agreements and mutual legal assistance treaties that govern the exchange of information for legitimate legal and security purposes. These frameworks provide structure, predictability and oversight, reducing the likelihood that foreign legal mechanisms will be used in a way that undermines Canada’s interests or public trust.
[English]
To further protect sensitive information, the Government of Canada applies a range of safeguards, such as restricting the types of data stored in third party environments, encrypting data to prevent unauthorized access and using standardized contractual clauses that spell out strict security and privacy requirements for third party providers. Together, these measures help ensure that even when data is hosted by a third party, it remains protected, controlled and managed in a way that maintains the confidence of Canadians and this committee.
In conclusion, Madam Chair, the government is embracing AI to improve services to Canadians and support public servants, but with strong safeguards in place. We are ensuring that sensitive information remains protected and that Canadians can have confidence in how their data is managed.
The Chair: We will now proceed to questions from committee members. For this panel, senators will have four minutes for the question, which includes the answer.
Senator Burey: Thank you for your testimony and for appearing at our committee today. I think you know that as senators, we are kind of faced with answering the public to say that we’re hearing about AI, we’re hearing about “deepfakes” and we’re hearing about issues with data sovereignty due to malware and all the things that are happening. You gave some of the investment that the government has done, but what would you say are the greatest risks and opportunities facing Canada regarding the increased adoption of integrating AI throughout society?
Mr. Schaan: Thank you for the question. I’m happy to start.
I think the risks are kind of in two buckets for me. One of the risks is that we fail to provide reliable and secure opportunities for artificial intelligence deployment to citizens. The minister likes to say, “Technology moves at the speed of innovation, but adoption moves at the speed of trust,” and I think we have a real risk that we will not have built a foundation of trust that will actually enable the AI opportunity to deploy. That means we need to ensure that we’ve combatted the real considerations that citizens have about the loss of control over their data, the loss of their jobs, the loss of the potential sovereignty they face and potentially — in the worst kinds of outcomes for AI — the loss of our world. I think the biggest risk is to build that foundation of trust and that responsible and reliable path forward to allow for AI.
The second kind of risk is actually that we fail to take up this opportunity. Canada has an extraordinary history in this particular technology space, like we do in many others. Three of the most cited AI academics in the world for the last five years are all Canadian. Many of the godlike people of AI are Canadian. But we again run the risk that we’ll have deployed and invented a technology that we failed to reap the rewards of, and that’s the risk I think we’re seized with.
The opportunity is: When deployed responsibly and reliably and with effective controls, AI has the capacity to be able to fundamentally increase our productivity, our efficiency and our consumer choice, as well as a legitimate step change in services, including in spaces like health, agriculture, transportation and logistics.
Mr. Rochon: I agree wholeheartedly with Mark Schaan’s response. Maybe I’ll add the following: At the end of the day, AI is important, but we need to understand that we have to rethink how we actually work. Rethink how everything is done because these tools are going to allow us to be more productive, more efficient and work faster.
But in so doing, we can’t forget the fact that we have to have guardrails in place as we take on these new tools because, as you pointed out, they come with similar opportunities for malicious actors out there who will use them for nefarious reasons. In Mr. Schaan’s world of Innovation, Science and Economic Development, there needs to be guardrails put in place for greater society so that as this is adopted, it’s adopted in a very responsible way to guard against some of those dangers and risks. Inside government, we’ve already put in place many of the controls, but the risk here is one of increasing awareness, training and then putting in place the appropriate guardrails in order to make sure that it’s adopted appropriately.
Senator Burey: Thank you.
Senator Hay: In my former life, I was the president and CEO of Kids Help Phone and acceleratorKHP, which launched nationally at the Elevate Festival and internationally at Davos in 2025. Kids Help Phone, or KHP, began using AI/ML in 2017, and I joked with the founding chair, Ed Clark, and the founding funder, Heather Reisman, both of the Vector Institute, about how KHP, a not-for-profit, was using AI before Vector actually launched. I picked up on something you said about “at the speed of” because that not-for-profit had to move at the speed of youth, and AI helped them scale by 250%. We were in uncharted territory, as was everyone, and I believe we still are as it relates to AI, including data governance, AI sovereignty, AI cloud tech stacks and so on. So thank you for being here.
Part of this study is to dive into data governance guardrails and the ethical use of AI. I think there’s a common belief in Canada — right or wrong, perceived or real — that Canada has less or little governance or guardrails in place specific to AI use, such as data storage, data travel, data sovereignty and so on. You touched on this, so I want to have you reinforce it. The majority of major data and AI companies are U.S. companies, particularly in the storage and travel of data. Anthropic, OpenAI, AWS and Microsoft are examples, and they’re all governed by U.S. law even when on Canadian soil, making it impossible I think — or perhaps not — to defend our digital sovereignty and digital borders.
If an executive order from President Trump came out and could potentially require AWS to release its data — let’s say it’s KHP data that is stored at AWS in Montreal after it boomerangs through the U.S. with Twilio — what would we do? We’re also seeing a situation unfold this week between the U.S. defense secretary and Anthropic, where the government is demanding the company drop guardrails, but the company is refusing because of the potential use in weaponry and surveillance of its citizens.
The questions I have are two: As Canada builds its sovereign cloud, how do you see that unfolding to protect Canadian data and people, building guardrails to protect while not handcuffing innovation?
I think you’re going to answer that.
And also, please answer how to protect against U.S. interference and infiltration.
The Chair: Before you start, I’m going to give you an additional three minutes to answer these questions.
Senator Hay: I spoke the entire time, but I had it on one page. How is that possible?
Mr. Schaan: I’ll try to be super quick and leave some time. I think it’s important that we ensure the committee and Canadians are aware that oftentimes we refer to AI as the great ungoverned or as an unregulated space, and I want to be clear at the outset that artificial intelligence is subject to many domestic laws that actually have really important fundamental protections in place for Canadians. Particularly, I would point to our private sector privacy law which — while the minister seeks to modernize it to ensure an effective enforcement regime — does control the use of personal information by private sector actors and has meaningful mechanisms in place to ensure that consent, access and control rest with citizens and that there are effective guardrails in place. Does it need to be modernized? Absolutely, and that’s why the minister is pursuing it, but that also joins sector-specific obligations, including in the financial sector where the financial sector regulator has specific guidance on the use of artificial intelligence. That’s accompanied by our transport regulator, and that’s accompanied by our health regulator for things like AI in medical devices.
While we don’t yet have an AI-specific general application statute in Canada, we do have a number of effective controls that are playing really meaningful roles in protection.
Coming to the second bucket, it doesn’t leave us invulnerable, though, particularly to the possibilities of foreign coercion and data misuse. That’s one of the reasons why we’re building sovereign capabilities in this country, and it’s why our infrastructure investments are so important. I think the Prime Minister’s speech at Davos highlighted that we need to find a third way between the hegemons and hyperscalers that actually allows us to have some strategic autonomy in this space. It’s why our $2-billion investment on the infrastructure side is really key.
But it accompanies some interesting approaches that are coming, legally and technologically, to be able to help in some of these things, like foreign coercion. One of those is whether you can actually transfer over the control panel responsible for things like the cloud to a Canadian operator that is actually the de facto operator. That’s essentially a technological solution that hands over cloud control, for instance, to a domestic operator.
The second thing is whether you can use things like encryption such that even if that executive order were passed, that company is not in full possession of the data control panel such that consent is required for both encryption keys — the one held by the cloud operator and the one held by the data subject — and they can never access both without the consent of the individual.
There are technological solutions that can accompany some of these legal solutions, but I think your broader point is: How do we path a third way that allows for Canada to have strategic autonomy in this space? That’s where we’re trying to scale Canadian champions, build the infrastructure and make sure our laws are actually keeping up with the notion of what it means to be sovereign and have effective controls and capacities in this new modern era.
Senator Hay: I’m going to pull the thread a little bit, knowing how these U.S. companies reacted to Donald Trump saying he could demand data. Microsoft, at first, pushed back and said “no” and then capitulated recently.
Are you confident that if there were an executive order that said this to AWS — “A” being Amazon, owned by Jeff Bezos — then the data on Canadian soil would be safe until we build our own infrastructure?
Mr. Schaan: This is where these technological solutions are coming into their own. I would say it is an important consideration for us for the government data that is housed and operated by our government agencies, but also these technological solutions of two-key encryption and some of these access to control panels are actually growing in utilization.
I can’t give you assuredness today with the answer “all Canadian data,” but I think in the current operations of the CLOUD Act, which requires a warrant from an entity to request it, there are some guardrails or hedges in place that are allowing us to maintain meaningful control, even in some of those abnormal situations.
Senator McPhedran: Thank you for making the time to be with us in person.
As I’m sure you well know, OpenAI and other AI chatbot makers are facing multiple lawsuits in the United States for alleged harms that have been done to young people, which have led them to self-harm and/or to go so far as to commit suicide. We know that OpenAI executives were recently summoned to Ottawa. You may have been involved in that conversation regarding their failure to warn officials of the activity on the accounts of the Tumbler Ridge school shooter.
You speak about trust, and you link that to guardrails. Can you tell us, please, what legally binding instruments protect Canadians now, and what legally binding instruments do Canadians need for ongoing protection?
Mr. Schaan: Thanks so much.
I’ll build on the answer I made to your honourable colleague. Things like our Personal Information Protection and Electronic Documents Act do have fundamental controls over the use of personal information in a number of these situations. It’s why the proposed modernization of that law includes a number of new features. Under our current law, consent is required for the utilization of that personal information, and a number of factors need to be put in place by companies when they are transmitting, consenting or utilizing that information, even in the context of an algorithm.
Where we see a real need for modernization in that space is in a number of areas. Specific to chatbots, it’s important that it allows for things like knowing when you are interacting and engaging with an automated decision-making system. When I am providing personal information to what I believe to be a third party service, do I get to know that it’s actually an automated service before I provide my personal information to it? When it makes a decision about me, how do I get to know what personal information was actually utilized in the making of that decision? That is another zone we think is important.
The third is around protections for the full digital identity and personal information we have. I think about this in the context of likeness in the “deepfakes” case, where I should be able to have the capacity to ensure that likenesses of me are within my control and capacity for consent in the same way that an analog or physical manifestation of my identity is.
These are gaps that I think we seek to fill.
As it relates to online platforms and chatbots in the engagement space, Canadians expect online platforms to have robust safety protocols and escalation practices in place to protect users in the public. We are working alongside our colleagues at the Department of Justice Canada, the Department of Canadian Heritage and Public Safety Canada, all of whom were present at the discussion earlier this week that you referenced, to review existing measures to ensure the safety of Canadians online, particularly children, is safeguarded, including their privacy but also their interactions.
We have tools. They include the Criminal Code and a number of duty-of-care obligations for informing as it relates to imminent harm, but that’s not to suggest that we don’t have remaining gaps. I think those are what the government is currently contemplating, particularly in the wake of recent actions.
The Chair: Sir, just to follow up on Senator McPhedran’s question, as I think about harms and potential harms, one of the areas where we are using a lot of AI is in health care. When people provide consent at the outset of the use of a personal device — for example, to control their chronic illness — do we have written into the process of the use of that device that as new iterations of that technology quickly and furiously come along, consent must be asked for repeatedly so that the person is aware of what they’re dealing with as things change?
Mr. Schaan: In the context of PIPEDA, the consent of personal information is meant to be both meaningful and also relevant to the tasks or functions that it was presented with at the time of consent.
The balance we’re trying to strike in the modernization is how we ensure that consent remains meaningful. You and I have better things to do with our day than spending all of our time hitting “accept” on the notion that our information is being shared or utilized. What we are looking at in the balance is to provide new mechanisms for individuals to be able to provide active consent when it comes to new uses or potential new iterations while allowing for something that the Europeans have done incredible work with, which is a legitimate business interest: If I order a package from an online distributor that has a courier service that they automatically need to use to ship me that package, the transfer of my address, name, general information and phone number should be assumed. That’s a legitimate business interest extension of the information that’s been provided.
To your point, though, when it’s being manifested in a totally different context, such as when a retailer has my information and now wants to provide it to a third party seller of a totally different service, then I should have to provide active consent because that’s not a related function. That’s the balance we’re trying to strike.
Senator Greenwood: My question has to do with Tumbler Ridge. I’m from British Columbia, so that’s big on my mind, of course. I’m just going to repeat a few questions in case there is anything else you would like to add, and then I have some other ones I would like to ask.
What work is the government doing to create and enforce a legal framework that could prevent tragedies — like Tumbler Ridge and OpenAI not responding — from happening again in the future? What are you doing to protect people from these situations? I imagine those include governance structures and legal frameworks around companies.
What is the government doing to hold companies accountable if they withhold information such as this or information that leads to great tragedy?
If you want to add some more to your responses to the previous questions, that would be great. If you think you’ve answered it, then I will take that answer too.
Mr. Schaan: First of all, I should have acknowledged at the very outset the extraordinary tragedy that was Tumbler Ridge and the significant loss of life and the absolute tragedy that it was to see young people suffer at the hands of an assailant. Obviously, it’s a tragedy that we need to ensure is never repeated.
As I noted, I think Canadians expect online platforms to have robust safety protocols and escalation practices in place. We are looking at our existing tool kit alongside all of our partners, like the Department of Justice Canada, the Department of Canadian Heritage and Public Safety Canada. We all have tools in our tool kit, whether it’s private sector privacy laws, forthcoming online harms legislation or aspects of the Criminal Code. We are ensuring that we are bringing all of those to bear. Specifically, I would say the government has met with the platform in question regarding its safety protocols to press the importance of engaging law enforcement when there is a credible threat of harm.
We are going to continue to engage with the platform and online service providers on this issue, and we are actively contemplating the necessary actions to prevent future harms, looking again across that panoply of potential tools to make sure we bring the right ones to bear. I think there is some carriage or weight that needs to be borne by a number of these. There is a role for private sector privacy law when you are handing over your personal information and how that engagement happens. I think there is some work for the Criminal Code when it comes to the risk of harm and the imminent threat of violence. There is potential for the role that online harms legislation can play in terms of content and those considerations. We’re examining all of those opportunities.
Senator Greenwood: We don’t have a lot of time. We can’t have this happen again. What are the accountability measures that you will hold companies accountable to? How will you ensure they have those things and the legal frameworks that back those up?
I have one other question. I always listen with interest when I hear the words “data sovereignty.” I am nehiyaw, and Indigenous data sovereignty is an area that we continue to explore and determine what that looks like. This adds a whole other dimension to ownership, control, possession and access, particularly for First Nations, Inuit and Métis peoples in this country. Talk to me about how you are considering that in the activities that you’re doing.
Mr. Schaan: Indigenous consultation and engagement, including the role that nations play in asserting their digital sovereignty and having the capacity to assert their digital sovereignty, were important considerations in the development of the forthcoming national strategy.
I think your point about an added layer of complexity is absolutely right. I would also say, though, with that added layer of complexity comes some unique opportunities. There are some interesting opportunities, for instance, in the digital infrastructure build-out around the relationships that can be played out between Indigenous nations and the opportunity to build out net new digital infrastructure of a more sovereign nature to actually provide that domestic capacity to have a much greater assertion of control, access and possession in a way that’s different from what the current alternatives might be.
That said, I do think we’ve seen the necessity of ensuring that we take a distinctions-based approach because this looks different in different contexts and in different communities. The provision of digital information and sovereignty in a health care space looks very different depending on where one is, so we’re trying to ensure that we are thoughtful and engaged as we carry out the next steps on that.
Mr. Rochon: Mr. Schaan has been speaking, rightfully so, about the Personal Information Protection and Electronic Documents Act, or PIPEDA, which is the outward-facing privacy rules that apply to companies. Inward to the Government of Canada, we have the Privacy Act and the Access to Information Act, which are two acts written back in 1983. The Privacy Act has never been modernized. You can imagine the sorts of rules that are in place. You are no doubt familiar with the United Nations Declaration on the Rights of Indigenous Peoples, which requires us to update these acts. We’re in the throes of looking to update both the Privacy Act and the Access to Information Act. Last year, we already started engaging in Indigenous consultation on the Access to Information Act. We’re taking that very seriously, and we’re lining ourselves up to have parallel tracks for both of these pieces of legislation in terms of our engagement with Indigenous communities.
Senator Senior: I am almost at the edge of my seat listening to you, so thank you. I want to direct this first question to Mr. Rochon and, perhaps, Ms. Beckles about the inward-facing work that you’re doing with the public service.
We know that there are cuts coming within the public service, if they’re not already happening. Is there a pilot or something similar that you have been using in relation to AI? Is there any evidence you see coming forward with respect to how effective that can be? Are you anticipating that some of this AI use will replace staff?
Mr. Rochon: Thank you for the question. I will frame it in the following way, and Ms. Beckles may want to add to the way that I’ll explain the adoption of AI within the federal public service. This is not a new phenomenon. It’s not that we’re waiting for a starting gun to go off and for AI to start being adopted. It has been used more often than not in scientific areas for over a decade now, but of course, with the advent of agentic AI and the evolution of AI, it’s more accessible to all public servants.
I mentioned in my opening remarks that we created the AI Register, and the register is still a minimum viable product. It is a beta version of what the AI Register will ultimately look like, but we wanted to get it out quickly because we wanted to be transparent and demonstrate to Canadians where AI is actually being adopted. That register already has more than 400 projects identified, so departments and agencies are adopting AI at a rapid pace across all types of services, such as in back office services for public servants to do translation or to do transcription for a committee such as this one. Some of them are also starting to look at outward-facing services to Canadians.
As all of that AI is being adopted, we’re concerned about the following questions: Are people trained? Do public servants actually understand how to prompt these tools as they’re being introduced? Is there duplication happening? The AI Register is going to be a powerful tool to see where these thousand flowers are blooming and to make sure that things are being adopted in a responsible way. We’re now consulting on this beta version of the register to ask what Canadians want to see related to where this technology is being adopted.
What is still not so much uncertain but unclear is the return on investment and the value proposition. Intuitively, it’s going to be more efficient and productive, but it’s not a one-to-one trade-off to say, “I’m introducing AI, and therefore I can get rid of these three public servants.” It’s a bit like asking me what the value was of introducing the internet, and did the internet replace public servants? No, it’s a tool. It’s a valuable tool that allows public servants to work differently. Similarly, AI is a very powerful tool that will allow public servants to think differently and deliver services differently. How do they leverage that? I don’t think AI is going to replace a public servant, but a public servant who understands and leverages the use of AI will replace a public servant who ignores and doesn’t want to use AI. That is how I would look at it.
Right now, in the context of the Comprehensive Expenditure Review, that is leading to affected employees. There will be a reduction in the workforce, but that is a separate conversation from the introduction of AI. AI is happening; it’s coming to play. It won’t replace the people who are now being affected. It will complement the future of the federal public service.
Senator Senior: Thank you for that, and I appreciate the distinction, but I can’t help but converge the two in some way. If considerable cuts are being made, it’s possible that the staff being replaced because they’re not adapting to the tool may not need to be replaced by new staff, right?
Mr. Rochon: The logic is sound. There is no doubt about it that, intuitively, you may think if you have to cut back, then maybe the introduction of some of these new tools will allow you to compensate for the loss of that staff, but that’s not the calculation that’s being made.
Senator Senior: I understand that, and I think it’s still early days.
Mr. Rochon: Indeed.
Kara Beckles, Executive Director, Privacy and Responsible Data, Treasury Board of Canada Secretariat: In the AI Register, the earliest use case of AI goes back to 1994, so it’s not something new that we’ve been grappling with. Many of those older use cases weren’t a matter of “Let’s replace something.” It was, “We have this problem. AI is a tool; it helps us solve a particular problem,” and that’s how we’re looking at it today. We have issues where we want to provide better services to Canadians or we want to analyze data or use data in a better way, and AI happens to be the best tool in order to do that.
There will be transformation impacts on the public service. When we went from using typewriters to using computers, steno pools disappeared, but we needed more technicians who knew the software and the hardware and helped us get set up. It was a transition in the types of jobs.
There were the same concerns then about the impact on the public service that we see today, and it’s about a transformation and how we help our employees through that transformation.
[Translation]
Senator Boudreau: You both alluded to my question in your opening remarks. Based on the little research and reading I did to prepare for this committee, it seems to me that there are two divergent camps in civil society: those who want to quickly move forward with AI at all costs to benefit more from it, and those who want to slow down so that regulations are in place to ensure more controlled growth. You alluded to that in your remarks.
My first question would be: Where are we in the range of other jurisdictions? Are we in first place? Last place? I feel like we’re rushing a little too fast and not putting enough emphasis on regulation. I would like to know where we stand globally as a society and as a government.
Mr. Schaan: Thank you for the question. The development of AI governance and its growth in an international economic zone is very exciting. [Technical difficulties] of the situation in the United States right now, our path was almost identical to that of the European Union, the United States and Australia, which seemed to want to build AI governance responsibly, cautiously and with security considerations for their citizens. That’s why Canada built the Global Partnership on Artificial Intelligence in a previous G7 and coordinated with organizations like the OECD.
With a new President of the United States in office and the Artificial Intelligence Action Summit in Paris last year, the U.S. government has now indicated that regulations and security considerations are an obstacle, a burden on U.S. organizations and companies, and it has stated that it wants to take part in a strategy to fast-track the use of AI.
Canada’s focus is more on balance. The European Union has continued its efforts with a regulatory zone. We have a bill for building capacity, but it’s not much. We’re just beginning to regulate. The U.S. is now really obsessed with unregulated AI. Canada is in the middle of the pack. We have an incredible research capacity. It’s not really clear that Canada has the capacity to commercialize and adopt AI at the same rate as the United States. However, with the national strategy, Canada has an opportunity to develop a real capacity to fast-track its AI function, but with a responsibility and caution that is lacking from the actions of others. That’s Canada’s place in the global AI zone right now.
Senator Boudreau: My second question may be more for Mr. Rochon. All government departments and agencies will want to use artificial intelligence at some point. We tend to create a number of silos. The federal government is built vertically, not necessarily horizontally. When we try to carry out horizontal projects, it doesn’t always work very well. I’ll use Phoenix as an example. What is the government going to do to ensure that artificial intelligence within the government is developed in a systematic, controlled and consistent way across the various departments rather than having 30 or 40 different silos?
Mr. Rochon: Thank you for the question. It’s very astute, and I agree with the way you described the environment. To answer your question, you have to understand that AI is based on a whole digital ecosystem. It’s not just about AI; it’s also about how we manage digital projects. You mentioned Phoenix, for example.
Unfortunately, as the Chief Information Officer for the government, I see that the responsibility is spread too widely. There is Shared Services Canada and there is the Canadian Digital Service in another department. There are too many departments. There isn’t really a core group that sets a horizontal vision for all digital projects. Our technical debt is quite severe. It will take an approach where we can prioritize resources — not just from a money perspective but also from a capacity perspective — to transform and modernize our systems and then adopt AI in a responsible way.
On my end, what we are doing for the Treasury Board of Canada Secretariat is putting a strategy in place. We asked that there be a hub for artificial intelligence in the strategy precisely so that we could make those decisions. In addition, the registry is a tool that will enable us to follow up on this.
Senator Petitclerc: Thank you for being with us. It’s fascinating. It’s very interesting. We obviously see all the positive opportunities that are there. I want to come back to the safeguards, the risks and the government’s responsibility. My question is along the same lines as Senator Boudreau’s and Senator McPhedran’s. It focuses on more vulnerable populations, such as children and young people. I’m very interested in young people and children. There is a UNICEF report that talks about opportunities for young people in health and also about risks. Is the government’s approach to apply a precautionary principle or is it a reactive approach?
I will draw a parallel that is probably very flawed. Given young people’s experience with screens and social media, I think we can say that we are in reactive mode and that the precautionary principle has not been applied. We look at the countries that say no — Australia has just legislated on this — and try to play catch-up because we have not properly assessed human risks and we just react. That’s what I worry about with AI: Do we have the ability and the will? Do we choose to say that we will not take any risks? We’re really going to go with a precautionary principle, especially when it comes to vulnerable groups.
Mr. Rochon: This is more of a question for Mr. Schaan’s department than for mine. To answer your colleague’s question a minute ago, there’s an economic issue that shouldn’t be forgotten either. Going slowly will hinder economic opportunities. I don’t want to say that economic issues have to be the priority, but we have to take everything into consideration and assess the risks.
There are principles that need to be introduced, there is no doubt about it. However, your social media example is very apt, I think. We have to find a solution where we can move forward and adopt this new technology. Otherwise we will have problems from a competitive standpoint with our global partners. We absolutely have to come up with a solution for creating an ecosystem in Canada and adopt the technology very responsibly. That is what is at stake. All the tools and laws that need to be modernized affect the issue, and you have to be able to establish principles and rules without necessarily knowing that. . . .
Unfortunately, the technology is evolving very quickly. We’re talking about AI. In two years, we’ll be talking about a new quantum computer that will completely change the stakes once again. It’s very complicated and, unfortunately, if you take too thoughtful and too slow an approach, you will miss the boat on a lot of economic issues. My colleague Mr. Schaan may want to add some information.
Mr. Schaan: I would like to highlight a few aspects of your question. First, it is important to note that our aspiration is to have a regime with the capacity to respond to ex ante technology and that is not linked only to ex post efforts. It’s very difficult, when an existing technology has injured or caused harm to people, to start from scratch and launch a new technology. It’s difficult. As my colleague said, the technology is advancing very quickly.
What capacity does the government have to develop strategies? There are a few interesting and more important aspects. In our bill, we have adopted functions or ways of doing things that are guided by principles. It’s not that unique a piece of legislation or bill. It is really in the area of principles that are able to evolve with technologies. In modernizing privacy legislation, there is an opportunity to include a design code for children to protect them from increased risks to their private information. That’s really at the beginning of the concept of the technology.
Second, we are also making progress in the research area. From the outset, it is important to be able to assess the potential harms of the technology. It also means working with industries and businesses and using the right tools and technologies. This includes our new AI security centre, which conducts research into AI assessment zones. What are the capabilities and tools to properly assess in advance the possibilities of technologies and algorithms that cause harm to citizens?
Third, it is about cooperation with companies and co‑designing standards and protocols so that we are all moving at the pace of technology and not just being tied to specific legislation that has more specific rules or functions.
[English]
The Chair: Thank you. Dead on time. I’ll just insert a quick point here.
In the research world, we call it the precautionary principle. Technologies today and innovators are not applying that principle. As we move forward, some of what you’re talking about is quite a temporal discrepancy. There’s absolutely no way we can keep up.
What are we doing to try to shift that mentality into more of a precautionary principle approach?
Mr. Schaan: There are a couple of zones that are very useful. The first is advancing evaluation science. Some of this is about having the capacity to be able to run metrics and run algorithms through tools to understand and anticipate potential harms. That science involves the evaluation, for instance, of chemical or biological risks as it relates to algorithms or the creation, utilization or propagation of synthetic content, and these are evaluative tools that we actually need to help design and develop.
The second, as I noted, is the embedding of principles-based approaches within our legislative frameworks, including the consideration and the concept of a duty of care and having liability standards that apply to the application of that duty of care. It transcends specific use cases and specific technological instances, but instead it invokes the principle saying that you will be liable for cases when you don’t invoke or carry forward your duty of care to the end users of the systems that you are designing.
Those are the ways in which we can hopefully get to a space where we’re not going to be able to say, “I don’t know what the next chatbot is; I don’t know what the next iteration of agentic capabilities is,” where we are going to want to say, “Do we have a law on the books that forbids this?” But instead, can I say, “No matter what the next thing is, here are the principles you need to apply to ensure you’ve actually extended and thought through the potential harms and you have actually put in place the appropriate guidance to the user when they deploy and utilize that technology.”
Senator Muggli: If we’re going to do a word cloud at the end of this study, probably the biggest word is going to be “guardrails,” and then it will be “tools, rules, principles and PIPEDA.” That’s what it’s going to look like.
I want to take us back a bit because we’re just starting this study. You are, in fact, the first witnesses.
We want to ensure we’re focused on solutions in our study. I’m looking for feedback from you on where you think we should spend our time in this study. Should we work on guardrails? Should we have new legally binding instruments? What guidance should our study offer to government to address the largest and most important gaps in this issue?
Mr. Rochon: You want to be looking at guardrails, but at the same time and in balance with that, you have to look at how we can ensure Canadians are actually adopting this technology. In government, we started putting guardrails in place as of 2019. In 2025, we realized we needed to change the tone of what we were doing. We were putting out rules, Algorithmic Impact Assessments and warning signs, and we recognized that most people in government were afraid to actually use anything because of all these rules, and it set the wrong tone.
Therefore, although the adoption of AI within government was happening, it wasn’t happening at a pace that was rendering our workforce competitive. There needs to be a balance struck.
Senator Muggli: Do guardrails equal principles?
Mr. Rochon: Among other things, absolutely.
Senator Muggli: Please tell us your definition of guardrails.
Mr. Rochon: There will be some confusion in terms of guardrails as far as the federal public service is concerned and guardrails as far as the rest of Canada is concerned. I will turn it over to the expert when it comes to our rules and transfer that question over to Ms. Beckles.
Ms. Beckles: The current guardrails we have in place for the Government of Canada come out of the Directive on Automated Decision-Making, which basically says that any time we’re using AI in a capacity that is going to affect somebody’s access to services from government or impact their life in some way or even impact the decision that a person is making about whether someone gets access to a service, then there’s this set of rules that kicks in which says, “Okay, now you have to do an Algorithmic Impact Assessment. You have to get a peer review done on your technology.” We need to understand it can’t be a black box. You have to be transparent and let people know that what they’re interacting with is AI or that AI was involved in this process. There has to be a recourse mechanism. These are all the guardrails we have in place when adopting AI within government.
Senator Muggli: Do guardrails equal legislated mechanisms?
Ms. Beckles: Within government, because the Treasury Board has a policy suite that is mandatory within government, they’re not legislated. But they’re all mandatory within government.
Senator Muggli: Should they be? Should we be working on legislation to have legislated guardrails?
Mr. Rochon: I would go so far as to say yes, but I’ll let Mr. Schaan answer about the outward-facing aspect, which may be a bit more complicated.
Mr. Schaan: I’d start where Mr. Rochon started in one point, which is to say that I don’t think the choice or luxury of this committee is to decide whether or not AI will happen in Canada. AI is here. It is happening, and in many cases, including in the Kids Help Phone example, it’s doing remarkable things to improve citizen service and effective outcomes for Canadians. The interesting question for this committee is: Knowing what we know about what is holding Canadians back from actually maximizing this opportunity, what needs to be put in place to be able to turn the opportunity into reality? Some of that will be around trust and confidence. And to that question, guardrails factor in.
I won’t weigh in as to whether you guys want to speak to this interesting question: Do we regulate AI specifically, or do we regulate AI vertically, as in do we need a stand-alone piece of AI legislation that covers all algorithms and all use cases, or do we regulate in health and then transport and then financial services and then the use of personal information while using these various tools to get at it? I won’t push you in one direction or another, but fundamentally, it would be useful to have recommendations out of this that say, “These are the things, no matter which way you go, that you need to ensure you address.”
I would capture the big barriers for Canadians in these four buckets: They’re worried about losing their data, losing their jobs, losing their sovereignty and losing our world. In what meaningful ways can we help allay those concerns for Canadians? On losing data, we must update our privacy laws and think about the conditions and the principles that need to be put in place and how we can start to create a more positive approach toward data sovereignty, which isn’t actually ex post but is about creating the conditions for data sovereignty. How do we allow people to have greater leverage and mechanisms for control?
On jobs, how do we think about skilling and ensuring we’re ready? Without horning in on previous answers to previous questions, part of what will happen as it relates to the displacement of labour as a function of AI will be that the benefits of AI first accrue to the individual before they accrue to the enterprise, and we have to think about how to translate those individual benefits into the next generation of highly valued work.
On losing sovereignty, what are the preconditions for what we think we need to have as sovereign capabilities in this space? On losing our world, what is our bet on things like artificial general intelligence, or AGI, or misalignment of AI systems? That’s a very wide terrain that I’ve just laid out.
Senator Muggli: Thank you for providing four guardrails for us.
Senator Arnold: I think my brain is exploding, but I appreciate you being here today. You have said it a number of times: the loss of our world. I would say that is a real fear of Canadians, but they don’t even really know what it is.
We talk about AI and the tools. Your dermatologist is going to scan you and find your melanoma; yay, that’s great.
I want to talk about generative AI. Are we doing that in Canada right now? Do we have one of those enormous centres that is doing that work right now?
Mr. Schaan: We have generative capabilities embedded in a number of industrial capabilities in the country. We’re one of only four countries in the world that has a large language model company. There are many in the United States, there are many in China, there’s one in France and there’s one in Canada. Our large language model company, Cohere, is an enterprise provider of large language model solutions where its model embeds itself into corporate cultures to be able to agentically and from a generative perspective provide solutions, whether that be RBC or Bell Canada, which are just two of their clients in the Canadian context.
Senator Arnold: It’s integrated already?
Mr. Schaan: That’s right.
Senator Arnold: It’s happening with just regular tools using it?
Mr. Schaan: That’s right.
Senator Arnold: Thank you.
The Chair: We have five minutes left to ask questions. Please take note of the questions, and we ask for a response in writing.
Senator Hay: Senator Arnold, KHP is building GenAI right now with Cohere, so it’s happening.
My question may be controversial. I want to pick up on the social media that was talked about earlier and the algorithms that are being built, as well as the safety councils that big companies invite organizations to be on — I’ll call that performative, not authentic — and I’m going to be very cynical here.
How will we hold these multi-trillion market cap companies to account as they’re building algorithms that drive young people to services like KHP? If they can build the algorithms to drive what they’re driving right now, for sure they can build algorithms for safety. For sure they can, and they’re not.
I guess that’s my question. How do we hold them to account?
Senator McPhedran: My question comes from looking at the European Union Artificial Intelligence Act of 2024, which has a tiered approach and is very definitely geared to the protection of humans. Where in the federal departments is the work being done on legislation to protect Canadians? Not policy, not principles, not guardrails, but legislation.
Senator Burey: I think there is a pattern here. We are talking about accountability, and that was my question. I wanted you to get more specific. You talked about the Criminal Code and the duty of care and really the cost of not doing what’s right, and I think we want to have some suggestions from you.
My last one is to expand on the return on investment that you spoke about, Mr. Rochon. Give us some more information about that and where you are.
Senator Senior: I really appreciate your openness and the sharing of information. Thank you for that. Related to the last two questions, I’m more relating it to the strategy. I don’t know if you can speak to this because I know it’s supposed to be out in the first quarter of this year. Will the strategy be inclusive of accountability? Possibly legislation? I would like to get a sense of what the strategy will cover and then, more specifically, is the strategy inclusive of equity-seeking groups and organizations as well as gender?
The Chair: Senators, this brings us to the end of our first panel. I would really like to thank Mr. Schaan, Mr. Rochon and Ms. Beckles for your testimony today. You got us going in a really exciting way. Thank you.
Senators, joining us today for our second panel, we welcome by video conference, from the Canadian Institute for Advanced Research, Elissa Strome, Executive Director, Pan-Canadian Artificial Intelligence Strategy. Thank you so much, Ms. Strome, for joining us today. And thank you for putting up with our delay. For your opening statement, you will have five minutes, followed by questions from committee members. Ms. Strome, the floor is yours.
[Translation]
Elissa Strome, Executive Director, Pan-Canadian AI Strategy, Canadian Institute for Advanced Research: Thank you, Madam Chair. Good afternoon, senators. Thank you for the opportunity to speak with you today.
[English]
I am Elissa Strome, the Executive Director of the Pan‑Canadian Artificial Intelligence Strategy at the Canadian Institute for Advanced Research, or CIFAR. CIFAR is a Canadian-based global research organization that has been operating since 1982, and for over 40 years now, CIFAR has played a pivotal role in mobilizing the world’s top researchers to address society’s most pressing challenges, including those in artificial intelligence.
In 2017, the Government of Canada appointed CIFAR to establish the world’s first national AI strategy, with a goal of attracting and retaining top AI talent to advance Canada’s leadership in AI research, training and innovation and to enable Canada’s long-term success. To date, the Pan-Canadian Artificial Intelligence Strategy has attracted more than 130 leading AI researchers through the Canada CIFAR AI Chairs program. This is now the third-highest impact AI research cluster in the world.
This incredible research and talent base enables our national ambitions of deeper AI adoption, commercialization and global leadership in AI safety. Since 2017, Canada has made significant progress in building our AI ecosystem. We have a vibrant AI start-up community that is attracting significant venture capital investment. We see significant growth in technology jobs in our major centres, and overall Canada stands eighth in the world in the Global AI Index. With the launch of the Canadian AI Safety Institute in late 2024, the appointment more recently of our first Minister of Artificial Intelligence and Digital Innovation and a much-anticipated new national AI strategy expected this spring, Canada is poised to advance our goal of realizing the real social and economic benefits of AI. But it will take a coordinated, focused, all-of-ecosystem approach to achieve Canada’s vision for AI. At CIFAR, we have always believed that understanding the societal implications of AI is just as important as advancing the technical capabilities, and we very much welcome the committee’s work on social risks and the societal implications of AI. Let me focus a little bit on your three major areas of investigation.
The first is on AI governance and sovereignty. AI sovereignty means building Canada’s capacity to leverage AI for long-term social and economic benefit. It requires building strong domestic capability in research, talent, infrastructure and governance, while strategically collaborating with trusted international partners. Canada cannot succeed by attempting to do AI alone. The scale and complexity of AI mean that international collaboration is a necessary condition for sovereignty, rather than a threat. Canada’s work as part of the International Network of AI Safety Institutes is a great example of how we are advancing our sovereignty in partnership with trusted allies like the U.K., Australia, Singapore and Korea. And Canada can learn a lot from our international partners and allies. Other jurisdictions, like the European Union, have already developed and implemented AI legislation and are actively learning what approaches are most effective. Canada has an opportunity to consider those insights as we develop our own approach to AI regulation.
The second is on AI ethics, privacy and safety. Effective AI ethics, privacy protection and safety must be grounded in research. Research is not an optional complement to governance. It is the foundation that makes governance credible, adaptive and enforceable. Through the Canadian AI Safety Institute Research Program at CIFAR, we are mobilizing cross-disciplinary expertise to advance AI safety and address the socio-technical challenges of frontier systems. This work translates abstract concerns into evidence-based insights that inform policy.
Finally, the third is on AI risks, benefits and social impact. AI has the potential to transform industries, strengthen economies and significantly improve the delivery of public services. But the social impact of AI is fundamentally a matter of public trust. But trust is not automatic; we’ve heard about that already today.
It requires foundational understanding, citizen engagement and inclusive decision making. And Canada has a long way to go to build public trust in AI.
In closing, AI will underpin future prosperity, and Canada must continue to advance it while remaining clear-eyed about its risks and trade-offs. Research enables that balance by informing policy, guiding responsible deployment and maximizing public benefit. As the federal government refreshes the national AI strategy, grounded in the principle of AI for all, it must avoid false dichotomies between research and commercialization. We, in fact, need both. Strong research and talent remain foundational to ensuring all Canadians can benefit from AI. Thank you, and I look forward to your questions.
The Chair: Thank you, Ms. Strome.
We will now proceed to questions from committee members. For this panel, senators will have five minutes for the question, which includes the answer.
Senator Hay: Thank you, Ms. Strome, for being here.
Canada was an early founder of AI and AI research. One might think we’re laggards now; others have moved ahead of us. We face intense international competition and significant investments and regulatory clarity emerging in the U.S., Europe and Asia. You talked about that.
I also note that in Canada, companies and start-ups build amazing things here and then exit Canada because there are greater opportunities and access to funding and growth in other countries, specifically in the U.S. many times. Shout-out to Cohere, which is still here in Canada.
In your view, what should Canada’s “play to win” be to remain globally competitive and translate research excellence into durable economic advantage?
Ms. Strome: Thank you very much for the question, Senator Hay. It’s a really important question for us to discuss and explore.
We have a fairly robust and vibrant AI start-up ecosystem in Canada. There are over 400 high-potential AI start-ups in Canada, with “high potential” meaning they’ve already generated investment of over $1 million and are attracting additional attention. Our AI start-up ecosystem in Canada is also attracting some venture capital funding. In fact, about 30% of all venture capital funding in Canada is directed to AI start-ups specifically, so they’re taking the market share of venture capital investment.
Your question is so critical, though: How do we support those companies to be successful on the global stage? They can’t be successful just in Canada; our market is too small. We need to help those companies grow and scale in Canada, and in order to do that, there are a few things that are essential. One is providing them with the expertise they need to move from that start-up phase into a much more sophisticated, dynamic company. We don’t have a lot of that expertise in Canada because we haven’t done that already. We don’t have a lot of good examples of Canadian companies scaling up to a larger level.
Can we train some of that expertise? Can we repatriate some of that expertise that has moved to other jurisdictions, like the U.S.? Can we provide services, training and support to those companies? Those are some examples.
Although venture capital investment is significant in Canada, there is a lot of room for more. There was a study published earlier this week looking at how Canadian AI start-up companies fare compared to other jurisdictions, and we are definitely behind in what is possible. We need more seed funding for our early‑stage companies and also Phase A funding for them as well.
Finally, these companies need customers. A lot of start-up companies tell me they have a hard time accessing customers in Canada, whether they be federal, provincial or municipal governments or Canadian businesses. We need some policies, opportunities and incentives for Canadian organizations to be buying Canadian AI. Also, they need access to global markets, too. They need help getting out there and getting on the ground — soft landings — in other jurisdictions where they can sell their products and services.
Senator Hay: Around governance and guardrails, what do you worry about in terms of legislation and guardrails that could perhaps sit in the realm of AI doing good for the people of Canada and protecting us but then potentially hindering them from their innovation? How would we balance that?
Ms. Strome: Yes, it will be essential. I spoke about that in my opening remarks.
It is not a trade-off between data governance and innovation. In fact, the two have to work together. Data sovereignty is something that this committee is looking at closely, and that has already been talked about today. Data is the essential fuel for AI. We need to actually make sure that we have access to very large, robust, diverse data sets that are representative of the communities that the solutions are meant to be applied to.
Guardrails that unnecessarily restrict access to data could possibly be a hindrance to innovation, so we have to make sure that data is protected — safe and secure — and that we’re managing all of the privacy and personal considerations, but we have to make sure it is available for innovators.
Senator Burey: Thank you so much for being here.
You talked about the importance of research into the social impact of this technology. I don’t even think “this technology” is the right term. You also said that we have a long way to go in building trust. I think in answer to Senator Hay’s question about having start-ups have access to companies, that is probably one of them.
What is actually happening in the research area in terms of the social impact? What gaps do you see that should be researched? Just give us a little bit about what’s happening in this area and what the gaps are.
Ms. Strome: Great. Thank you very much for the question, senator.
One of the key areas of research that we are supporting right now at CIFAR — in collaboration with ISED through the Canadian AI Safety Institute — is we lead the academic research program. We have been doing a lot of work to think about the societal impact of advanced AI systems. We just recently published a year in review, looking at some of those considerations. We are thinking about, again, the impact on the information that people are receiving and the impact of advanced AI systems, for instance, in terms of misinformation and disinformation. Some of our researchers are looking at the impact of advanced AI systems on democracy and how AI systems can be used effectively or maliciously in elections, for instance.
We’re looking at the societal impact of AI within the health care system. There is an example of a technological approach to AI that has the potential to really benefit society. Likewise, there are the impacts of AI on the environment, climate change and biodiversity. Those are some of the positive impacts of AI that will ultimately benefit society.
Some of the gaps are still in our research areas. Since the technology is moving quickly and policy is trying to keep up with AI, there is probably a need for greater innovation and more research on how we can develop innovative approaches to managing AI, AI policy and AI regulation as we move forward.
We have a dedicated issue of the Canadian Public Policy journal coming out this spring that looks at some of those questions and how we can be innovative in our approach to AI policy as the field is moving very quickly. I would highlight that area as a need for more work.
Senator Burey: Do you think enough funding is going toward looking at these societal impacts?
Ms. Strome: I think there is room for more growth. Historically, in Canada, we have focused our funding very much on the technical development of AI, which is a lot of the work we support at CIFAR. We are now starting to think about investing and we are starting to invest in the societal implications. There is a whole generation of students, investigators and experts whom we need to train on these topics, too.
While we’re very strong on the technical side and we have some great expertise on the social side, there is room for more growth in that space.
Senator Burey: Would that be a recommendation to our study?
Ms. Strome: I think so. That is something that would be very well received. There is a lot of ambition across the country to address these questions, and there is room to do more.
The Chair: Ms. Strome, would you be able to share the references to the two articles you spoke about in the last few minutes?
Ms. Strome: Yes, I will.
Senator McPhedran: Thank you very much for being with us, Dr. Strome. I’m sitting here listening to you and watching you and I’m thinking of Dr. Fraser Mustard and Dr. Reva Gerstein.
Ms. Strome: Wonderful.
Senator McPhedran: I happened to be with them in the formative days of the institute, so it’s amazing to be sitting here these many years later.
My question is quite simple. Within your work and your teams, have you looked at establishing a guiding principle of using a human rights lens in any of the work you’re doing?
Ms. Strome: That is a wonderful question. That’s a superlatively Canadian approach to thinking about responsible AI. Canada has a very long and strong tradition of responsible AI that is both in our technical research and in our societal impact research on AI. It’s a culture that permeates our research community and has translated to many of our business enterprises, and it is also in the Government of Canada. Canada is a leader at the federal level in thinking about AI policy, regulation, legislation, et cetera.
Responsible AI is a key characteristic of Canada’s AI ecosystem. I sometimes call it Canada’s brand. That is very much resting on our strong tradition and commitment to human rights. We have researchers. My colleague Professor Catherine Régis at the Université de Montréal is a legal scholar. She takes a human rights lens to her work in AI policy internationally, particularly in thinking about how AI is implemented in the field of health care.
Another area that we can think about adding a human rights lens is in thinking about the guardrails as we move forward. What would future legislation or regulations look like? Again, in the upcoming issue of Canadian Public Policy, one of the recommendations from a researcher whom we work with at Osgoode Hall Law School is recommending a harm reduction approach to AI legislation and thinking about taking some of those principles that have worked well in public health and other areas of policy and legislation and seeing how we could apply them to AI. It’s very much a human rights-centred approach.
Senator McPhedran: Thank you.
Senator Senior: Thank you so much. I really appreciate your response to the last question. Just for my understanding, do you fund research?
Ms. Strome: We do, yes.
Senator Senior: Okay, thank you. Understanding that context, have you funded any research that deals with topics somewhat related to what you just talked about? One of the concerns is the equity lens and the issue relating to inequitable data and AI. I’m interested in that, and I’m also interested in the role of public education and whether there has been any research in terms of recommending how the public can become acquainted with or educated about AI.
Ms. Strome: Great. Thank you very much, senator. Excellent questions. We do fund research at the Canadian Institute for Advanced Research, or CIFAR. One of the hallmark programs that I lead is called the Canada CIFAR AI Chairs program. This has allowed us to build a very vibrant world-class community of AI researchers across the country. They’re affiliated with the three national AI institutes, and they’re working primarily on the technical areas of AI but also thinking about how it’s adopted in ways that will benefit our society and our economy. There is a small subset of researchers working on the socio-technical and social impact issues.
When we think about the equity lens, it is something that is first and foremost on my mind and something that is very much embedded in this culture of responsible AI that I described earlier, and we absolutely are pursuing research. We’ve done a lot of training with our researchers about what that means. What do equity, diversity and inclusion mean? What does it mean in the context of their teams and the people whom they’re recruiting and training? What does it mean in the context of the AI that they’re developing? We try to work with our researchers and make sure that they understand these principles and concepts and that they’re embedding them into their research. It’s actually something we evaluate them on as they’re submitting their proposals.
Second, on the role of public education, this is also something that we focus on at CIFAR. We have a course that’s available to all Canadians in French and English. It’s an introduction to AI called Destination AI. It’s just a basic introduction, but it’s a way for people to grasp the foundational concepts and to start to understand and think critically about AI systems.
It gives them a little bit of the technology background but also the social impact. It talks about generative AI and gets them thinking about how AI is impacting their own lives. That is something we’ve made available to everybody, but there is so much more that needs to be done, and it’s not easy to reach everybody. It’s something for which we need a much deeper investment and a national approach.
Senator Senior: Thank you. Just a quick follow-up to that, I appreciate that it’s available to everyone, but what’s the level of access to that information? How is it being promoted?
Ms. Strome: Yes, it has been quite challenging to reach Canadians. This is something that we launched about three years ago now at the start of the generative AI movement. We’re really trying to make it available to people. There are other AI literacy programs across the country, not just CIFAR. It’s not our primary mandate.
My colleagues at Amii, the national AI institute in Edmonton, have really embraced a mandate around AI literacy, and they’re really focusing on post-secondary education. Every student at the University of Alberta is required to take an introduction to AI course. Amii is working with over 40 post-secondary institutions across the country — universities, colleges and CEGEPs — to introduce an AI literacy course to all those students. It is really a team effort around AI literacy right now in Canada, but we are only just scratching the surface because we haven’t been effectively funded, and there is no clear organization that has a mandate to deliver this training to Canadians.
Senator Senior: What is your recommendation?
Ms. Strome: We need a national approach to AI literacy. There are many different activities and programs happening. They are not coordinated. We are a ragtag group of people who find each other and share our ideas and what we’re doing, but there is no coordination, and there is no funding for this work either. It really needs to be a matrix approach so that we’re providing AI literacy, helping build trust in AI and engaging Canadians in the conversation around AI. That’s one of the key missing factors in why Canadians don’t trust AI. We’re actually not including them in these conversations very effectively. So we need a robust strategy.
The Chair: Thank you. Just to remind our members, Professor Catherine Régis will be here for a future session.
Senator Boudreau: All the words you just said in the last two sentences were words I had written down in front of me to formulate my next question. The earlier guest said that we need to focus on the adoption of AI by Canadians. When you spoke earlier, you talked about the social impact of AI and the need for public trust. Canada has an aging population that’s afraid of technology. Convincing them to adopt AI is going to be difficult. There is already a lot of concern around social media, and they don’t know what they can believe anymore. There is a whole generation that probably still doesn’t do anything online in terms of banking or booking vacations or doing anything like that. There’s a fear, I think, of the technology, and so long as that fear is there, building up that public trust is going to be even more difficult.
Realistically, what can the federal government do to address that? We have an aging population, and convincing them that this is the way to go in the future, I think, is going to be a very hard sell.
Ms. Strome: That’s a very fair point and an important observation. Thank you, senator.
I think older Canadians have a right to be distrustful. They are vulnerable for sure. They’ve been taken advantage of by malicious actors and by scammers. We’re hearing all the time about how older people are being taken advantage of, so it’s very important that they are skeptical and that they are careful and manage their trust carefully.
How do we engage them? They absolutely need to be part of this conversation because they bring such an important perspective — even that distrustful perspective is an important perspective — into the conversation.
Realistically, I don’t think that older Canadians are the primary demographic for adoption of artificial intelligence, or AI, but we want them to be comfortable with it as a tool and a resource that is embedded in our society.
A couple of ways to engage older Canadians, I would say, are certainly courses that are designed and tailored to their specific questions and their specific applications, showing them where they should be careful and critical and maybe distrustful in approaches and also where they can start to build and understand the technologies and start to build trust, so there should be literacy programs that are tailored to older Canadians.
Also, I think they bring an important voice to the conversation. One of the things that we’ve been doing with partners at Mila and across the country is thinking about citizens’ assemblies around AI. Can we bring Canadians together, perhaps, across different demographics? There’s one happening in Montreal that we’re a part of — I think it’s next month — for youth and really bringing young people together from across the country with different perspectives to provide their perspectives on AI and then share those perspectives directly with policy-makers. That’s what we’ll be doing with the results.
Could we also do citizens’ assemblies with older Canadians? I think that’s an important opportunity going forward. Again, it’s the way to include more people in the conversation of AI and have those voices be heard.
The Chair: We welcome you to that event that I will be hosting next month.
Senator Muggli: Thank you for being with us today. Hopefully, you didn’t answer this question already because I went down a bit of a rabbit hole on your website while you were talking. I just noticed one of your Canada CIFAR AI Chairs, Reihaneh Rabbany, along with a colleague, created “. . . a user‑facing app that will be able to verify the authenticity of online information.”
It says the app will be available to the public with a prototype expected in March 2025.
Where is that at?
Ms. Strome: March 2025? It might be March 2026, actually. There could be a typo there.
It’s a great project, and Professor Rabbany is one of the world’s experts in online misinformation and in developing tools that identify truthful information using AI. AI can be a problem, but it can also be the solution, and she’s working toward that end.
The Veracity app, as far as I understand it, is still being pilot tested. We’ve recently provided some funding to Professor Rabbany’s collaborators in their public engagement and in their rollout of the app and in engaging with different demographics on the use of it.
It’s still a research project, I would say, but I think the beta version is meant to be out next month.
Senator Muggli: It sounds like an excellent tool to be able to support people. Maybe if you’re not super strong on critical thinking, you have an app to help you out.
That’s all I have. Thanks.
Senator Greenwood: I have just a couple of quick questions. I was jotting down some words as you were speaking.
I wondered how the United Nations Declaration on the Rights of Indigenous Peoples interfaces — because that’s federal and provincial legislation in this country — with some of the research. I know you were talking about policy and legislation earlier.
I know that all of these things need data. I was wondering about data sovereignty as it relates to Indigenous Peoples in this country because in the data world, we’ve never solved that one, I don’t think.
Ms. Strome: Indeed.
Senator Greenwood: I asked the previous panel this as well because I think it adds a whole other dimension of complexity when you’re working with this particular population, particularly as it relates in the health world.
Ms. Strome: Agreed. Thank you, senator. Great questions.
With respect to the United Nations Declaration on the Rights of Indigenous Peoples, I’m very proud of some of the work we’ve done at the Canadian Institute for Advanced Research, or CIFAR, to think strategically and hand in hand with Indigenous leaders and communities about the impact and opportunities around AI. It’s something that we’ve been thinking a lot about.
Of course, our mandate is research, education and training, so we focused on that space. Over the years, we have supported training programs for Indigenous high school students on AI. We currently provide a scholarship for all Indigenous students who are accepted into our training programs every summer, and we support their participation in that programming.
I think the biggest activity that we’ve done at CIFAR around engaging and integrating Indigenous perspectives in AI is a training program that we’ve launched — again, it’s free to all Canadians and available in French and English — called Indigenous Perspectives in AI, and it is meant for people who don’t already have that experience and those perspectives.
Every summer, for every student who is part of our national training programs on AI — somewhere between 250 and 300 students every summer, which is a lot of Canadian students but also includes international students as well — we require them to take a special curriculum module on Indigenous Perspectives in AI. It’s not just work that they do at home on their computers. We facilitate a sharing circle and a workshop on those perspectives as we move forward.
The data sovereignty question is really an important one, and the principles of ownership, control, access and possession, or OCAP, really need to be front and centre as we’re thinking about AI applications with, for and by Indigenous communities and leveraging Indigenous data.
I think it’s this principle of “nothing about us without us.” Where Indigenous data is being leveraged for AI tools, it should be at the request of and the desire of Indigenous communities and bringing in those principles of OCAP first and foremost. Indigenous data should be leveraged for the benefit of those communities, and that’s where both the social and the economic benefits should accrue.
We’re seeing more Indigenous start-up companies in AI and more organizations that are working in this area, and I’m pleased to continue partnering and working with those groups and supporting what they’re doing.
The Chair: Thank you.
Senator Petitclerc: Thank you, Ms. Strome, for being here. It’s quite fascinating and a little scary.
I have a question. It’s a straightforward question. I’m trying to get a sense of the whole environment of AI in Canada and the world, I suppose. One thing that you’ve talked about is the disinformation. I want to address what we read about bias when it comes to AI-generated content and sexism, racism and ableism. I have three or four questions.
The first question is: How major of a concern is that, and is there a solution to that? Is there a way to get around that, and who is responsible? Who has leadership? Who does it? What is the role of Canada?
Also, I’m thinking about exposure to such bias, especially for youth, so my last question will be: Are there any studies on the impact of such biased content on youth, such as psychological and social impacts?
It’s a multi-question question.
Ms. Strome: Thank you very much, Senator Petitclerc. Bias is, unfortunately, a critical concern when it comes to AI applications, and the bias problem predates generative AI. We talk about generative AI approaches as the modern AI approaches, but even traditional AI approaches were fraught with bias. Unfortunately, the bias creeps in the data sets that are being used to train these AI models. Bias is often a result of not having a representative enough data set. It’s not large enough, it doesn’t have enough samples or it doesn’t have enough diversity in it. Sometimes the bias creeps in because of the way human developers think about the approach and the tool that they’re developing.
One of the ways that we’ve tried to mitigate AI bias in the past is by teaching AI developers, our students and the community and setting strong principles around how to build robust data sets because you actually get better AI tools if you have better data sets. So we focus on the quality of the data set as the foundation for developing good AI tools, and one of the ways that we think about mitigating bias in AI is diversifying the teams that are involved in developing these tools. If you have people with different lived experiences helping to develop, test and think about what the user needs as these tools are being developed, that has been shown to be absolutely essential to mitigating bias in AI systems. We do that at the Canadian Institute for Advanced Research by diversifying the talent within the AI ecosystem, offering scholarships to students who identify as members of under-represented groups, thinking about gender and diversity and having policies in place to advance equity, diversity and inclusion in AI so that our ecosystem is more diverse and the tools that we develop are developed by more diverse people. That’s one of the solutions.
Who is responsible? This is a great question. If you think about commercial AI products, how do we mitigate bias in those commercial products where we don’t have this culture of academic research driving it forward?
The AI tools that have demonstrated their effectiveness, their approach to responsible AI and their efforts to mitigate bias are actually the tools that people want to buy. They’re better AI tools, first of all. They’re more effective in doing the thing that they’re designed to do, but they also appeal more to consumers. Companies that have been able to demonstrate that they follow trusted principles around AI and show how their tool has been improved because of the approaches that they’ve built in are the ones that are more successful.
Do I have time for your last question on the exposure to youth and the impact on youth?
This one is harder. I can’t think of any specific studies. I might have to get back to you on that, but we know that youth are impacted by their engagement with AI tools. We’ve seen that. We’ve got the Kids Help Phone example, and we’ve got other examples of research in this space. Of course, we know youth are impacted by their exposure to media in general and social media in particular. I will have to get back to you on the question about if there are any studies on the impact of AI bias on youth.
Senator Petitclerc: Thank you.
The Chair: Thank you. This brings us to the end of this meeting. I would like to thank you, Dr. Strome, for your time and for being with us today. Any information that you’d want to forward to Senator Petitclerc could come through our clerk.
(The committee adjourned.)