Canadian Heritage Committee on Oct. 29th, 2025

[Vicky Mochama]

At the core of any decisions around AI is the question of public trust. I'm certain you see it in your offices where people become angrier, more frustrated and less accurately informed. That has been in progress since well before AI intervened in that process; what we're now seeing is an acceleration of that with the use of AI tools.

Can we meaningfully intervene in what AI is doing but also take a step back, look at the larger picture and look at the landscape of trust? How does trust factor into our digital information ecosystem, whether it's your ability to log on to the Toronto Public Library website—as many Torontonians could not do, because that website was held ransom for almost a year—or your ability to log on to ChatGPT and ask it a question about who Vicky Mochama is? When I ask that of ChatGPT, I've never met that woman in my life. We have a set of unreliable factors. That doesn't mean that everybody is a bad actor in the mix; it's simply that we do not have a coherent information ecosystem that we can point to and say to Canadians, “You can trust that.” I think the trust question is primary for our publishers: How can we ensure that they can trust that?

The question of attribution is an important one when it comes to AI summaries or anything that AI software is going to use. If you're going to pull from a reporter's work, the public should be able to replicate that work, find that reporter and ask if that was true. I think the human-in-the-loop paradigm is one that our publishers rely on, especially when it comes to public trust.

[The Chair, Liberal, Lisa Hepfner]

Thank you.

Mr. Champoux, you have the floor for six minutes.

[Martin Champoux, Bloc, Drummond, QC]

Thank you, Madam Chair.

Ms. Guèvremont, I first want to congratulate you on the book entitled Intelligence artificielle, culture et média, which you published with Colette Brin. It's absolutely fascinating. I found a lot of answers and a lot of interesting questions in there. It provides extremely relevant material to fuel discussion. I'm pleased to have you here today.

There's certainly a lot of talk about the impact of artificial intelligence on the cultural sector. However, alongside that sector, there's the journalism sector, which we're discussing today with our guests. I'd like to take you down that path, then, because it's an integral part of your research. I found it quite interesting to see that not everything is doom and gloom in the conclusions drawn from your research, and I'd like to hear your thoughts on that.

Is it possible that journalism might not be a victim of artificial intelligence, but instead benefit from it? I'm talking about journalism as a profession, but also, more broadly, journalism in the sense of its societal role and its essential nature for democracy.

Do you think it's possible to use artificial intelligence and look to the future with a bit more optimism?

[Véronique Guèvremont]

Mr. Champoux, thank you for your question as well as your comments on the book I co-edited with Colette Brin.

I'm going to disappoint you, because our expertise is shared between Colette Brin and me in this book. Colette Brin, who's a Laval University professor in the department of information and communications, and director of the media studies centre, is really the one who would have the expertise to answer that question, specifically in relation to media and journalism.

I'll say a few words, because we do work together. Colette Brin is one of those researchers who recognizes the potential of artificial intelligence, on one hand, depending on how it's used by journalists and in newsrooms, particularly in terms of increasing productivity. Journalists and media rooms have been using artificial intelligence for a very long time now, and they can benefit from it in some respects. However, there are obviously also some risks associated with its use.

I'm afraid I don't want to dwell on the topic, since it's her area of expertise.

[Martin Champoux, Bloc, Drummond, QC]

That's absolutely fine. We can also refer to this book, which contains answers, particularly the results of investigations conducted by Knight Lab and Knight Foundation, which reveal some interesting and encouraging things for the media, including those with limited resources.

Anyway, we may be able to come back to it later if we can have more studies and hear from Ms. Brin at another meeting.

You also talked about culture and raised an issue that I find extremely concerning: the homogenization of cultural content. That's an issue that's all the more concerning because we have a culture to protect, particularly that of Quebec, which is surrounded by around 400 million anglophones.

Given Canada's francophone culture, specifically Quebecker culture in Quebec, as well as our cultural ecosystem, how can we play our cards right and convince the major players, who don't necessarily need to care about us, to respect that difference?

Is that an impossible dream? Are we always going to be swimming against the tide, or do we have the technological ability to play our cards right?

[Véronique Guèvremont]

Thank you again for your question, which is excellent.

Some actions have already been taken. First, homogenization is a phenomenon that isn't unique to the era of artificial intelligence. Certain cultural powers, such as the United States or other governments in other parts of the world, have long had enormous power to penetrate our markets in light of globalization. That's already leading to a certain homogenization of what we're exposed to as an audience, so homogenization isn't a phenomenon that's unique to artificial intelligence.

Now we're seeing that the use of artificial intelligence at various stages in the value chain can amplify that phenomenon. When it comes to using artificial intelligence in analyzing user behaviour and algorithmic recommendations, studies have shown for some time now that content similar to the content previously viewed and enjoyed by users gets featured. That leads to a certain homogenization of the content consumed and to the filter bubble phenomenon that exists in the media but is also observed in the cultural sector.

The way to combat those filter bubbles is to ensure that recommendation algorithms are reversed, meaning that they don't recommend only similar content, but open the field of possibilities and spark the audience's curiosity about other types of content than those they would have previously enjoyed or viewed.

Legislative measures have been developed in that area. In Canada, the Online Streaming Act talks about promoting Canadian content, which has to be anglophone, francophone and indigenous at the same time. As you know, a bill on the discoverability of French-language content is currently being studied in Quebec. That covers the stage of distributing and disseminating works.

The challenge now is to also combat homogenization at the content creation and production stage, given that there can be more and more reliance on artificial intelligence.

I'll go back to what I said in my opening remarks. It's really important to work on the training datasets so that they are diverse, don't include biases and can generate a certain diversity of content.

It's important to be careful, though. This isn't about generative artificial intelligence that would generate content on its own. If that were the case, we would necessarily be dealing with a form of homogenization. Indeed, depending on the functionality of those systems, which are based on calculations and probabilities, we're inevitably moving toward the production of similar content. Where we have some leeway is when—

[The Chair, Liberal, Lisa Hepfner]

Thank you.

[Véronique Guèvremont]

Yes. I'm sorry.

[The Chair, Liberal, Lisa Hepfner]

That's okay.

Thank you.

[Martin Champoux, Bloc, Drummond, QC]

Thank you, Ms. Guèvremont.

[The Chair, Liberal, Lisa Hepfner]

Mr. Diotte, you have the floor now for five minutes.

[Kerry Diotte, Conservative, Edmonton Griesbach, AB]

Mr. Geist, I was a long-time journalist, so I have a special interest in this regard.

I believe you mentioned something about how over-regulation of AI could spark a replay of what we've seen happen on news sites or something to that effect.

Could you expand on that and say what the risks are?

[Michael Geist]

It's reasonably well known that I was quite critical of Bill C-18 or the Online News Act. Many of the concerns that were raised at that time have played out, with respect to the blocking of news links now for a couple of years, which hasn't entirely eliminated the existence of news on those Meta platforms. It has eliminated much of the quality news. We've seen substitutable content that has undermined the reliability of it and diminished the amount of news because, quite frankly, people are happy not to view that news. It has ultimately hurt many of the news organizations that were beneficiaries of the referral traffic and the like that was coming from that platform.

Even more, I think Bill C-18 has also had the effect of undermining the independence of many media outlets. We've now seen some outlets that previously were opposed to taking the Google money now saying they can't effectively compete if they don't take the Google money when their direct competitors are.

We've created a system where the vast majority of organizations are now increasingly dependent on regulatory models that result in these kinds of payments. I don't think that is a healthy place to be with respect to where we're at.

I referenced that in my opening remarks because it seems to me that one of the risks we face is if we adopt an approach that increases the costs.... Basically, if we say that the only way you can include certain Canadian content in a large language model is if you meet these new regulated costs, we're going to loop in some of these same players and we run the risk of running a replay on this all over again.

It's not that these companies are going to exit per se, although some might. The Quebec legislation that was just referenced—Bill 109, I think—very likely leads to the potential exit of certain streaming services, given some of the demands, especially in the context of AI. We have services that will basically say that if the choice is that they have to compensate in a manner that is inconsistent with what the business models and the global standards are or exclude their content, they're going to exclude the content.

That leads us to a place where many of the benefits we've heard about from this are lost. Even more, the concerns we have about the AI systems reflecting our culture and reflecting our interests may be lost.

You have to ensure that you are in the game, it seems to me. We face real risks if we make choices that lead some of these organizations to say that Canada is a market that isn't AI-friendly and they can't effectively compete in this space.

As I said, we've seen that play out with respect to Bill C-18. I don't think we'd want to see that play out again with respect to generative AI.

[Kerry Diotte, Conservative, Edmonton Griesbach, AB]

Excellent.

Dr. Geist, who determines if AI is impartial or objective about the news it creates? We've heard, obviously, that it scrapes news. Who determines and how do we know that news is objective?

[Michael Geist]

People ask that same question about all sorts of news organizations, not just the AI-generated versions of that content.

We don't have someone who reaches those determinations. If you take it as a given that the majority of the public is interested in being able to access accurate news, whether that's from the original source or a summary of what is available by gathering up sources from a number of places and putting that back, if you end up with an AI system that consistently provides information that is found to be inaccurate or biased, I suspect that the majority of users are going to say that it is not an AI service they want to rely upon. We've seen it play out with Twitter's AI service, Grok. There were a number of instances when some of the results it was generating caused a lot of concern. People then abandoned that service because it was not reliable.

The market has some disciplining effect here. If the kinds of outputs that are being generated are the sorts of outputs people can't rely upon, if the summaries themselves don't reflect what is happening in their own community or their own interests and the like, they simply aren't going to use those services, it seems to me.

The market plays a bit of a role, but we don't have—nor should we have—a government agency or otherwise going on and checking this out to say, “Is this accurate or not?” That's not how the system functions.

[Kerry Diotte, Conservative, Edmonton Griesbach, AB]

Thank you very much.

[The Chair, Liberal, Lisa Hepfner]

Ms. Royer, you now have the floor for five minutes.

[Zoe Royer, Liberal, Port Moody—Coquitlam, BC]

Thank you, Madam Chair.

I'm really enjoying the conversation.

A recent Library of Parliament report highlighted the fact that many countries are now re-examining their copyright laws. The EU gives creators the right to opt out. That's really interesting, because it puts the onus on the artists. The creator's work has already been stolen, but that means that unless they opt out of having it used, it can be used freely. We've heard some witnesses at previous meetings on this topic talk about opting in to have their content used.

Could you speak a little bit more about that?

I know, Professor Geist, you touched on opting out, but I'd also like to hear from Vicky on this topic.

Thank you.

[Michael Geist]

I'll try to go quickly.

First off, I don't see what is taking place as theft—it's not being stolen—and it is not necessarily the case that it can be used freely. What's taking place, certainly in an AI context, is an interest in the underlying data itself. It's not being republished. It's not being commercialized in any way where people are taking that original source and trying to make that original available to them. They're interested in the underlying data itself. To the extent to which we see even these systems learning and then generating results that are similar, musicians and creators have been doing that forever. Many of their perspectives and their approach is based on seeing what others are doing and trying to adapt and put their own unique spin on it.

In that sense, this is not theft in the way we would conventionally think of theft, and it is not free use. The uses—at least what we've seen from the courts so far—are consistent with what fair use or fair dealing would be, and frankly, because you mentioned the EU, the EU has a text and data mining exception that specifically seeks to ensure there are certain kinds of uses that are appropriate for that kind of informational analysis purpose.

That's not to say we don't need to be thinking about copyright in this context, but I'm not wholly convinced the starting point here is to say that what's happening is that content is being stolen and then used with no limits at all. There are clear limits in the law right now under fair dealing and fair use, and the kinds of uses are not what we would conventionally see when we think of that.

[Zoe Royer, Liberal, Port Moody—Coquitlam, BC]

I guess, really, it's specifically to opt in and opt out.

Vicky, could you also respond?

[Vicky Mochama]

Creators should be empowered to leverage their data, their content and their creations as they best see fit. There are examples of singers who can no longer sing who have opted to participate with AI systems to recreate songs or music they want out there; they want to continue having their voice out there but conditions are such they can't do that. That should be an option that's available to you as the creator. If I want to resell a book I'm writing but I'm not going to finish the book and I want the publisher to finish it for me using an AI system, that should be within my rights. It shouldn't be within the publisher's rights to insist on that. I think it's about where the power balance lies and whether the creator is still made whole and sacrosanct in that process.

[Zoe Royer, Liberal, Port Moody—Coquitlam, BC]

Thank you.

My next question is for Mr. Sutherland.

You had talked about the right of personality. You mentioned this is within the U.S. Constitution, but we are seeing deepfakes come out of the U.S.

Can you speak a little bit more about that? Does their constitution protect them enough? What can Canada do in this realm?

[Chip Sutherland]

The protection is there. Whether you can enforce it or not is the question of the deepfakes.

We barely have it. Just to give you an example of what I mean by this and why it's separate, there was the case of the skaters Salé and Pelletier. A photographer went and took their picture skating, framed it and sold it for $500 at a store, and they were like, “Wait a minute; that's us.” You own the photograph, because you're the copyright holder. The copyright didn't protect them, but he was using their image to make money.

There are a few cases like that, and that's what I'm talking about. If somebody were to use the recording, it's such a fine line. What Professor Geist said is true: Musicians constantly reference other musicians they're inspired by. It's different from scraping the actual songs from a record and then using the images of an artist. Feist, who I manage, is very distinct. It captures her whole character.

That's what I'm saying. It wouldn't hurt for us to be exploring whether we can formalize this as a separate protection. It's not data. It's not that vague, “Well, they're just learning from the data.” If you're using their voice and their image, then you're definitely stepping over that line.

[The Chair, Liberal, Lisa Hepfner]

Thank you.

Mr. Champoux, you have the floor for two and a half minutes.

[Martin Champoux, Bloc, Drummond, QC]

Thank you, Madam Chair.

Ms. Guèvrement, I'll come back to you. Our American neighbour has its eye on the cultural sector. I think it really doesn't like the barriers that Quebec and Canada put in place to protect our cultural distinctiveness.

In your opinion, in the specific case of artificial intelligence, would it be possible to effectively regulate and oversee its uses? Is it a waste of time?

If so, how could we regulate the development of artificial intelligence to properly oversee it, with all the constraints that entails?

If not, do you think we should invest more in Quebec and Canadian businesses to encourage the development of artificial intelligence?

Which approach would prevent us from disappearing due to artificial intelligence?