top of page
Radio show microphones

Wonks and War Rooms

Apple Podcasts
Spotify
RSS Feed

Season 8 Episode 1: Provincial, National and International AI Approaches

3 days ago
20 min read

In this episode Elizabeth chats with Christelle Tessono who is a doctoral student at the University of Toronto and a researcher at The Dais, to unpack what AI governance really means. They chat about AI as a socio-technical system which includes the technology but also the labor, land, water and infrastructure behind it. The conversation traces Canada’s Federal AI policy landscape and provincial response like Quebec’s Law 25 and Onatrio’s Bill 194.  Christelle argues initiatives about AI governance need to come from collaboration between people  across a wide range of departments and agencies as well as beyond government. 


Additional Resources:


Contributors

Host: Elizabeth Dubois

Research Lead: Nathan Kazmir Poklar

Transcript: Areej El-Sharif

Audio Editing: Sarah Pledge Dickson

Translation: Julianne Bernier Robert

Publishing and Promotion: Aaron Brisson

Producer: Sarah Pledge Dickson

Episode Transcript: Season 8 Episode 1: Provincial, National and International AI Approaches

Read the transcript below or download a copy in the language of your choice:



Elizabeth Dubois: [00:00:05] Welcome to Wonks and War Rooms, where political communication theory meets on the ground strategy. I'm your host, Elizabeth Dubois. I'm an associate professor and university research chair in politics, communication technology at the University of Ottawa. My pronouns are she/her. This season, we're talking about social and policy issues surrounding uses and governance of artificial intelligence. And today, we're kicking things off with a conversation with Christelle about AI governance, what it means, and what it looks like in Canada and abroad. Christelle, can you introduce yourself, please?


Christelle Tessono: [00:00:33] My name is Christelle [Tessono]. I'm a doctoral student at the University of Toronto's Faculty of Information, and I study all things related to technology governance. In recent years, I've really been interested in AI governance [...] I also conduct that work at The Dais, which is a think tank based at the Toronto Metropolitan University. And it's a pleasure to be here with you today, Elizabeth.


Elizabeth Dubois: [00:00:55] Thank you so much for being here. I'm really looking forward to this conversation on AI governance. It kicks off our season on AI beyond tech, and I think we need to start with what do we actually mean by AI governance? So yeah, what's governance? What's AI governance?


Christelle Tessono: [00:01:11] Yeah. So that's a really good question. And I think it really depends on how we view AI. I personally view AI very largely and broadly because it's a highly contested term, whether you're in industry or in academia. Some people view AI as purely a technical thing, just like, you know, a set of models, computational techniques, and so on. Some people view AI as a socio-technical system, meaning that it includes the people as well as the machines that, you know, power these systems. And I think I fall more into that perspective. So I view AI, you know, as a socio-technical system that is materialized through, you know, the political, social arrangements between the networks of labor. So like the software engineers, the council working on, you know, ensuring that the company is respecting privacy laws, the data workers who clean up the data sets that are used to power the models, as well as the natural resources that fuel these tools. I'm thinking about the water, the fossil fuels, the land, the material resources as well. So the fiber optic, the computational tools, or even techniques from the algorithms, the data itself and the physical resources, such as like the data center facilities or even like the offices of the content moderators, for instance. And I can understand how, you know, this definition is very broad and confusing, but when we view AI in that way, it makes governance easier to understand because we're not only focusing on like the technical components of a system, but we're looking at like the resources, the people, the infrastructure needed to ensure that a model is operating. And so AI governance, in short, the way I would define AI governance is as a set of rules, norms, tools, institutional mechanisms, and practices designed to align AI development and deployment with the objectives of a given organization or a government.


Elizabeth Dubois: [00:03:13] That's really helpful. Thank you. And I, and I really like how broad you go with it, because I think it really does address the fact that these AI tools and systems are being integrated into so many aspects of our lives and are impacting so many different aspects of our lives well beyond just like the instance where we decide to interact with these tools or not, or some other person or institution has decided we are going to have to interact with these tools or not. So, that broadness, I think is really, really helpful. Another part, though, that's pretty broad in your definition, is [how it] aligns with the objectives of an institution or a government. What does that even look like? What do we mean by those objectives? How do we set those objectives? It seems really hard to think about what governance could be or should be when presumably there's a lot of different players with a bunch of different objectives.


Christelle Tessono: [00:04:06] Yeah. So you'll see governance initiatives emerging from, you know, companies or even within sectors such as in education. And unfortunately, what we're seeing as a trend is that the objectives are defined from a top down perspective, which is not the most democratic way of engaging in governance. But we'll also see community organizations developing their own approaches to AI governance. Maybe they're not as formal as what a government would do, but it is still a form of governance itself, because you'll see a community organization state that like, we're not going to use AI for these types of things. We're instead going to be using different types of technologies in order to address the needs of our community. That is a form of AI governance. It's not what we hear about in the news, but again, it's an example of it.


Elizabeth Dubois: [00:04:54] Yeah, that's really helpful. And it also starts to make tangible the different kinds of players involved in governance, right? So, I mean, in a lot of the rest of our conversation, I think we are going to talk actually specifically about what governments are and aren't doing. But there are so many players beyond just governments that participate in governance structures and systems. Are there a key set that you think are the most important, or maybe the most visible, at least at this stage in the AI governance discussions?


Christelle Tessono: [00:05:25] Yeah, so it depends on the sectors. But let's say let's start with the government. The most visible actors tend to be policy makers who are in sort of the industry side of government. So for instance, in the Canadian context, a very visible actor in AI governance is the Department of Innovation, Science and Economic Development,“ISED”, and they're the ones who are taking a lot of time with organizations and industry, as well as in other sectors to evaluate how should we go about AI governance. And they've also been the ones behind the drafting of codes of conduct, Bill C-27, which, you know, we can talk a little bit more later on. And they've also been the ones behind the national AI Literacy Initiative, as well as the Pan-Canadian AI Strategy. But other actors that are very important in AI governance are also tech companies, and they're also very visible as well. Tech companies tend to dominate the space in terms of defining what AI is, what their tools are able to do. And sometimes that's a lot of hype. That's a lot of marketing that is happening in those formulations and conceptualizations of what I can do. Less visible actors are, you know, the everyday people who are facing those tools. I'm thinking, for instance, of someone working at a hospital, a nurse, a custodial staffer who, you know, has to process their invoices or other types of payments through an AI system. They might not be aware that this is an AI system that they're interacting with as well. I'm also thinking about kids in schools who are now being evaluated with AI tools, whether it's for plagiarism detection or for, you know, simple grading of their assignments.


Elizabeth Dubois: [00:07:20] Yeah, absolutely. And then if we also expand our view with that sociotechnical lens that you mentioned, you know, you also brought up impacts on the environment and land and those sorts of things, right? So it's like farmers who maybe aren't thinking of themselves as really essential to the tech governance conversation most of the time actually do become pretty essential to the AI governance conversation, but aren't necessarily part of that wider existing tech governance community typically.


Christelle Tessono: [00:07:51] And like, there are other ways to like view actors within the I ecosystem. So there's like people who develop the tech. So the tech companies or even researchers at universities, the adopters of the technology could be, you know, a farmer or it could be, you know, a government department. There's also people who research those tools, whether they're in academia or they could be journalists or community organizations investigating how AI is being used in their respective communities. And then there are people who are evaluators, auditors of AI systems, who, you know, are doing the impact assessments of, of the tools to, you know, figure out whether they're able to actually conduct the practices that they claim to be able to do or whether they're doing them fairly or unfairly. What types of resources that they're using, whether there be environmental, economic, labor resources and so on. So it's big and [I] understand how that can be challenging, but it's viewing it in such a large way allows us to understand that there are different points of intervention for all the different actors depending on your level of expertise.


Elizabeth Dubois: [00:08:57] Yeah, that makes a lot of sense. And it's really helpful to do that kind of mapping of the different kinds of actors and their different kinds of roles, and then connect it back to, and now we base the governance choices and the regulatory choices on those different actors and those different roles. So let's jump into what the Canadian federal government has actually been doing so far. Can you give us a rundown on what that looks like right now?


Christelle Tessono: [00:09:20] Yeah. So I would say that what is happening right now in Canada is reflective of maybe like ten, 15 years of deep engagement on AI and broadly like open data practices. And it begins in the late 2010’s with the development of the Pan-Canadian AI Strategy, which sets, you know, millions and millions of dollars for three pillars first, commercialization. So ensuring that small and medium enterprises are able to adopt technologies that are going to increase their productivity. And also ensure that Canadian companies developing these tools are able to, you know, develop them in Canada and not have to rely on American companies to supply these types of tools to Canadian businesses. There's also standards development. That was one of the pillars of the Pan-Canadian AI Strategy, and standardization meant working towards ensuring the tools that are being developed and deployed follow the same codes and practices across industries, a level of interoperability, meaning that like if we're using this tool in this context, it should be able to work very well in another context, ensuring that like the data is being processed similarly for cybersecurity purposes as well as, you know, privacy and so on. And then the last tidbit is related to research. A lot of money has been invested as part of the Pan-Canadian AI strategy towards development of academic capacity in the country. This meant providing universities and institutes specifically with funds to support professors, doctoral students, graduate students, undergraduate students in studying AI and developing, you know, whether it's auditing mechanisms, assessment mechanisms, as well as own AI tools as well. And, you know, identification of harms related to AI as well. Those were all parts of the funding of the research. Yeah.


Elizabeth Dubois: [00:11:18] Okay. So that's the Pan-Canadian AI Strategy and what came next. I mean, recently we've had, you know, the minister talking about how essential AI is. And it just feels like in the last 12 months, there's been this explosion of effort, at least public facing effort from the federal government. What does it look like now?


Christelle Tessono: [00:11:40] Yeah. So between the Pan-Canadian AI Strategy and now we've actually had a renewed version of the strategy. I think it's “AI for All” that is now being called. And that initiative seeks to really revitalize efforts for AI development in the country. One of the pillars [is] related to sovereignty, ensuring that the tools developed are actually like from an infrastructure perspective, actually within Canada. AI is a transnational tool, meaning that we have resources and expertise spanning, you know, the borders of Canada. We have researchers, we have water labor resources that are way beyond the country's capacity. So the AI for all strategy seeks to ensure that the development is made in Canada and benefits the Canadian economy. Between this renewed AI strategy and the older strategy, there was also Bill C-27, an attempt to create a regulatory framework for AI. And unfortunately, that bill didn't pass. It also didn't receive the usual care afforded to bills. There was no public consultations before the bill had been introduced. It came as a surprise on like a summer in June 2022, if I remember correctly. And so a lot of people, you know, pushed back on that. They said, we need more time to actually figure out how we want to go about it in Canada. And there's also the Sovereign AI Compute Strategy, like there's a lot of small strategies, voluntary codes here and there. And in short, the way I would define AI governance in Canada at the federal level, it's really a patchwork of different initiatives. We do have things in the private sector [...] that focus on ensuring that we have a lot of funding going into industry, but we also have a federal strategy that seeks to identify how the federal government should go about using AI for the delivery of public services.


Elizabeth Dubois: [00:13:45] Right, right. It's interesting to hear that you see it as this patchwork, because a lot of the kind of publicity around “AI for All” and everything, it makes it feel like, oh, this is like the thing that's going to solve all of the AI governance problems and conundrums. As an outsider looking in, it can feel like it's meant to be just one sweeping set of choices. But in practice, that's also not really how governance of any tech has worked, right? Because there's all of those different actors and it can seem a little bit like, oh, it's a patchwork is necessarily a bad thing. But I also see that there's potential benefits to having, you know, like a lattice sort of lots of different things coming together.


Christelle Tessono: [00:14:27] Yeah. I agree with you because there are so many different sectors where AI is being used. It wouldn't make sense to have one big AI bill that addresses what's happening in education, in agriculture, in healthcare. Like that doesn't make sense. But unfortunately, because [...] one department dealing with a lot of the AI regulation, it means that like other departments or other sectors, are maybe not well supported in ensuring a cohesion in the regulatory framework.


Elizabeth Dubois: [00:15:03] Right. And I imagine we could think about this as there's high-level, kind of abstract values, moral decisions around how we want to use these tools, what we think governance and particularly regulation should look like that might apply broadly, that need to kind of get decided on but are hard to agree on. And then there's the more specifics of, well, how exactly we want that to be implemented. And then in all of these different domains, like there's a lot of pieces. And that's just thinking at the federal government. I want to turn now to the provincial counterparts and to territorial counterparts. What does that look like? Are provinces and territories doing much? Does it complement what's being done federally? Is it in conflict with it? What does that look like?


Christelle Tessono: [00:15:49] Yeah, I would say that provinces are now developing their own frameworks. But for like, I would say in the late 2010’s, provinces were sort of waiting on the federal government to take leadership. And then as new technologies became more popular, such as, you know, ChatGPT and all that, provinces were like, okay, we cannot wait. We need to figure out our own approach. And you'll see that Quebec decided to go about AI governance by updating their privacy laws. And then, if a decision is made with an automated decision making system or AI, however you choose to name it, people have a right to know why and what type of information was used to make that decision. And that is part of Law 25, if I remember correctly. But there's also the province of Ontario, which passed, I think, last year, Bill 194, which is a bill that seeks to regulate how public sector organizations are using AI. And there's a privacy component to it because they update how the information of young people in, particularly in education and child welfare, is being used by AI systems.


Christelle Tessono: [00:17:05] It's also like a cyber security bill because it, you know, requests departments and agencies to, you know, have impact assessments for how they're using AI and so on. But, it's still not sets of laws that prohibit a certain type of AI being used. It's bills that seek to just essentially like, okay, if we're going to use AI, we need rules to, just make sure that we're doing the assessments. And I think Manitoba also recently passed a bill this summer that again, looks at the public sector use of AI and essentially, you know, sets a bunch of frameworks in place for protecting the data of people in the province of Manitoba, as well as, you know, basic cybersecurity practices. You'll see other provinces have codes, but they're voluntary. They're not bills. And they focus as well again, on how are public sector institutions using AI? Very few of them are looking into regulating how industry and other actors are using AI. It's very difficult to do that now.


Elizabeth Dubois: [00:18:10] Yeah. And it's so interesting that the focus is, okay, how are we as government institutions going to use AI? What does that look like when a lot of the big concerns and risks that get talked about more publicly right now are not usually about whether or not the government is using these AI tools, right? And we, I think about the kinds of risks to the environment, to our own personal data, but then also what does it look like to support an AI industry and like tech development, and the strength of a lot of research and development work being done in Canada. How do you support that economic system, but also create a system that encourages responsible development and approaches that are going to be both economically viable, but also preserving of the public good. And so it's interesting that that's not where most of the provinces are doing their work at this point.


Christelle Tessono: [00:19:13] Yeah. And I think it's a reflection of the federal system in Canada. Provinces have such limited amount of power on the things that impact their populations. And what's also interesting is that AI governance is now about believing that if we have ownership on how we develop the tools, then we're going to make sure that the tools are safe. But investing in AI development doesn't mean that the tools are being safe, because we're still not outlining what, you know, our limits are we going to be developing, you know, initiatives where funding is available to develop surveillance tools that are AI based? Like those conversations are not happening and they're very important in AI governance, I think.


Elizabeth Dubois: [00:19:55] Yeah, yeah, yeah. Because it's one thing to say, okay, we need to have control and access. And if we're going to be able to develop tools that are going to serve our population, we need to set up the infrastructure for it. Like that all does make sense. But yeah, that next step of, but how do we ensure that the development actually looks like the kind of development we would want? How do we ensure that the tools being created are actually ones that serve people for the better, or at a minimum, don't undermine entire democratic systems or public health or safety, those kinds of things. And we've got a ton of episodes this season that start digging into some of those different problems. And what better policy and regulation might look like. For now, though, can you talk to me a little bit about what's happening at the international scale? Are other people already doing this better? Are other jurisdictions ahead of us in terms of outlining what looks good, responsible, ethical, whatever term you want to put in front of AI development?


Christelle Tessono: [00:21:02] Yeah. So in the AI governance space, we've often looked towards Europe as an example of a jurisdiction that is doing things slightly better than here in North America. But I'll put a big caveat that I'll address later on. But essentially two years ago, the EU passed the EU AI Act, which sets responsibilities and also limitations on AI systems based on their levels of risk. It's a risk based approach that prohibits, you know, AI systems that will go against human rights, if I remember correctly. And it sets responsibilities for the actors adopting those tools as well as the actors developing them. It's quite big as a body of legislation, and the EU is a really huge jurisdiction with actors, you know, doing also their own things. And so it is still too early to say whether they're being good about their approach to AI, because there are a lot of bodies that are being developed right now for oversight and enforcement. And the enforcement piece is really important because how are you going to enforce this large body of legislation to maybe see a company that does business in the EU, but is not based in the EU. Is a fine going to be enough? Are the legislators, are the policy makers and regulators able, with the current tools at their disposal, to conduct investigations to find out whether, you know, a specific company contravened to the EU AI Act. Those are still questions that are really important and in the following years, we'll be able to see whether it was effective or not. Yeah.


Elizabeth Dubois: [00:22:43] Right. And so there's that enforcement question. Another one that comes to mind for me immediately is like, how good are we at actually predicting the risks? Right? How good are we at actually setting those thresholds? Well, is there like flexibility built into the EU approach? How do we deal with the fact that the technology is going to keep changing, and how we evaluate risk might also need to keep changing?


Christelle Tessono: [00:23:09] Yeah. And again, [it] depends on the independence of regulators. Right now the EU has set up boards that are somewhat independent and like yes, somewhat resource[ful] with funding, with people with expertise in public interest technology. But it's still too early to figure out whether that is enough. And another set of initiatives that we're seeing in the international context is the emergence of AI safety institutes. Canada is part of an international network of AI air safety institutes which seek to study and investigate AI tools. But those institutes are more about research. So even though they're affiliated with policy maker institutions, they're still about conducting research and having access to the tools. But what if you're an institute and you find out that a tool is incredibly dangerous? Like what is the next step available to you in order to ensure that the tool is no longer available? These are things that we're grappling with right now. The Canadian example is the unfortunate tragedy that happened in B.C. Tumbler Ridge, a community affected by how, you know, someone was able to, use ChatGPT to engage in very harmful behavior. I know that, like OpenAI is being currently investigated and that the Canadian Safety Institute was also pulled into this in order to investigate, you know, the symptoms and how, you know, they sort of allow very toxic conversations to continue with the teenager. But what is the next step available to us in Canada is still unclear. And in Europe those are the same types of discussions that they're having and grappling with.


Elizabeth Dubois: [00:24:59] Right. And what are the next steps once you notice that? But also like, are there early detection approaches to prevent harms before tragedies happen? Those kinds of things are really, really difficult to figure out. And some say, you know, well, a healthy AI environment would support the development of AI enabled tools that themselves would help with those kinds of things. But what I'm hearing from you is there's a whole bunch of unknowns that in an ideal world, we'd have tons of time to research and really think through. But in practice, there's a whole bunch of different AI tools that are out loose in the world and having impacts on daily lives, and there's only going to be more of those as the tech keeps evolving. How does governance keep up with the pace of tech?


Christelle Tessono: [00:25:47] Yeah. I think that's a really good set of questions. And on the one hand, people will say that we just need to evaluate the tools more. And that is a technical view of governance, which is not inherently bad, but it's one component of a bigger puzzle. I don't think you need to know how to code to know that a certain tool might not be good for you. I like to think about facial recognition technology, which is, you know, how a lot of people have begun their journey into AI governance. You don't need to know whether the tool is working accurately to know that, this is actually about surveillance because in order to operate, it needs a lot of images to build the model. It needs a lot of images in its database to make sure that they're identifying the right people. So that doesn't make sense. And I don't think you would need to do deep investigations to understand it. Like a surveillance tool is not useful in a lot of contexts. And I think that we need to expand AI governance discussions beyond the technical features. We can make tools marginally better day by day if we keep putting money [in], but is that a good use of money and resources? I don't think so. I would be more interested in AI governance initiatives, moving beyond what is happening in industry, and focusing more on providing resources for policy makers to coordinate across sectors and to have more public interest technology development, meaning that like we have people who are able to like, understand the technology at a technical level, we have people who understand how to implement or consider human rights risks, sustainability obligations that we have signed on in Canada, or also even like commitments towards indigenous reconciliation, that we still are a long way to go in ensuring that are being met. All those people, putting them together at a table, I think would make for amazing governance that is addressing current risks as well as issues that are plaguing Canadian society beyond just technology.


Elizabeth Dubois: [00:27:59] Yeah, yeah. And a way of trying to guard against just re institutionalizing and, and reinstilling some of the discriminatory practices that we know are very common. Right. And we know that a lot of tech development has been developed on very specific data sets and so serves certain populations really well and other populations poorly. We also know that collection of personal data about people in the land we call Canada has historically been used and weaponized against Indigenous populations. We know that there are these problems. And so if we have those different actors involved in developing governance approaches, maybe we can do it better than we've been doing it in the past.


Christelle Tessono: [00:28:46] Yeah. And what's really important is that often technologies [are] put in isolation of other issues. And a big marketing element is that we're saying, oh, we're going to put AI in this because it's going to increase productivity. It's going to increase access to public services. But AI is not a Band-Aid. If the infrastructure is weakened, the Band-Aid is not going to protect it. It's going to still crumble. And so having an inclusive approach to AI governance is the best way to tackle this.


Elizabeth Dubois: [00:29:16] That's a really helpful analogy. We're coming up on time, but I wanted to ask you the question. That's going to be our final question for every episode this season. From your perspective, what is it that you think policymakers might be missing right now in the AI governance conversations?


Christelle Tessono: [00:29:34] Hmm. I think that policymakers are underestimating the power of collaboration across departments. I would be interested in initiatives about AI governance coming from collaboration between people in health departments, regulatory bodies, agencies and so on, with a competition bureau, with a privacy commissioner. I want to see more collaboration in that end, because people in those separate departments and agencies are thinking about AI. They're just not able or empowered to have those conversations. So cool. Governance and cross-departmental collaboration, I think, is like something that really needs to be developed more in Canada, but also internationally. So many other jurisdictions.


Elizabeth Dubois: [00:30:25] Amazing. Thank you. That's a really, really helpful insight. I think we could talk for ages, but we are at time. Thank you so much for joining today.


Christelle Tessono: [00:30:33] Thank you so much for having me, Elizabeth. It was a pleasure.


Elizabeth Dubois: [00:30:37] Thanks for listening. That was our episode on AI governance. I hope it gave you a bit of a framework and mapping to use for the rest of the season. As always, we've got a bunch of links to different resources in the show notes, and you can head over to polcommtech.com for annotated transcripts in English and French that link out to a ton more resources and further readings. This season of Wonks and War Rooms is supported by the University of Ottawa's Office of Vice President, Research and Innovation, the Faculty of Arts, the center for law, Technology and Society, and my University Research Chair in politics, Communication and Technology at the University of Ottawa. I want to acknowledge that I'm recording from the traditional and unceded territory of the Algonquin people, and I want to pay respect to Algonquin people acknowledging their long standing relationship with this unceded territory.

Comments

Commenting on this post isn't available anymore. Contact the site owner for more info.
1/15

© 2026 Pol Comm Tech Lab
University of Ottawa

Sign up for our newsletter to receive occasional emails from the Lab.

  • PolCommTech on LinkedIn
  • @polcommtech on Instagram
  • @polcommtech.bsky.social on Bluesky
bottom of page