Canadian Heritage Committee on Oct. 22nd, 2025

[The Chair, Liberal, Lisa Hepfner]

I was just thinking the same thing. His comment was very important.

[Bernard Généreux, Conservative, Côte-du-Sud—Rivière-du-Loup—Kataskomiq—Témiscouata, QC]

—what he said is totally relevant, and it was worth saying.

If possible, I would like him to have his full six minutes.

[The Chair, Liberal, Lisa Hepfner]

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

[Martin Champoux, Bloc, Drummond, QC]

That is very gracious, and I appreciate it. Thank you, Mr. Généreux.

I would like to thank my friends from Meta for joining us, and my first question is for them.

Two years ago, Parliament passed Bill C‑18 regarding online news. Meta decided at the time to pull Canadian news content from its platforms, probably respecting the law in the process. Countless small regional media outlets in Quebec and Canada, as well as large media companies, that used Meta platforms to share their content were tremendously harmed by this decision.

During the study of the bill, it was shown that, because consumer habits have changed considerably over the years, Meta played a major role, probably unintentionally, in delivering news content in Quebec and Canada. While it didn't change anything, I showed that Meta had a social responsibility to continue publishing content from news businesses in Quebec and Canada.

Meanwhile, for the past two years, Google has accepted the law and agreed to commit $100 million a year, helping small media outlets that would probably not have survived otherwise.

Would Meta reconsider its position and, in the short term, allow Quebec and Canadian news content on its platforms?

[Rachel Curran]

Look, we would love to bring news content back onto our platforms. I think I said that two years ago, as well. We are hopeful that the government will take another look at that legislation, which we think misrepresents the value exchange between publishers and our platforms.

We are not like search engines. Search engines scrape the Internet for news content, and they present it in the product that people see. When they do a Google search, they expect to see news content in that search, so they use news in a very active and proactive way. We do not. We just host news passively. News publishers place their content on our platforms because they get increased distribution and then they can monetize the clicks that they receive as a result of that distribution, so we think we are in a very different situation from Google.

That said, we would love to put news back on our platforms, and we're hopeful that can happen.

[Martin Champoux, Bloc, Drummond, QC]

If I understand you correctly, your position has not changed and you are using the same arguments you used two years ago, despite the situation the news industry finds itself in. Is that correct?

[Rachel Curran]

I would say that we have a new government now. I think that the new government is more open to these kinds of discussions, so we're hopeful we can make some progress with them.

[Martin Champoux, Bloc, Drummond, QC]

Thank you.

Let's go back to artificial intelligence. Mr. Chan, you mentioned earlier that Meta planned to develop AI in a way that would benefit everyone.

Have you considered that content creators, artists and copyright holders also need to benefit from artificial intelligence? There are a lot of them here today. If so, how will you ensure that copyright is respected and that content creators are properly compensated for the use of their work?

[Kevin Chan]

We think artificial intelligence is so important that it should be made available throughout the economy and for all levels of society, and that includes creators and artists. We think this technology will be very helpful to creation and creativity, and we are willing to make our models available for that.

In terms of what you're asking, which is about the act of model building and training models, as it is with the entire industry that is building these AI models, we do not see how learning about information and developing the patterns and relationships to build these models touches on copyright interests. We believe that this is very much an act of trying to get a tool to learn and develop very powerful models that are going to be very useful for society and the economy. In our case, we are giving away those models free so that everybody may benefit from them.

[Martin Champoux, Bloc, Drummond, QC]

I understand you're making your models available. I also understand your willingness and the concept of open-source code. Meanwhile, content is being generated thanks to work covered by intellectual property or copyright laws.

When this government or another government decides to legislate to rein in companies that develop AI tools or tools that use easily accessible content on the web, will you welcome that legislation, or will you react as you did when other laws were passed limiting what you can do?

[Rachel Curran]

Yes. I think Mr. Chan's answer indicated where we're going with these models.

Our models don't store or reproduce copies of any content. They have been trained on publicly available information across the Internet, billions of pieces of data. Any one piece of data is only marginally influential in the overall performance of the model. What they do is extract what we believe are unprotectable facts, statistics, patterns and relationships from that data. They're not extracting the protected expression from that data and certainly not reproducing it wholesale or in part, nor, again, do they store it. We don't believe that training these models implicates copyright interests in that respect.

That said, Meta, and I think other companies too, have entered into licensing deals where it makes sense to do so, and I think we will continue to do that. I think our fundamental argument—and we made this argument in a submission to the government when they conducted their copyright consultation—is that model training does not implicate the interests of copyright legislation as it stands now.

[The Chair, Liberal, Lisa Hepfner]

Mr. Waugh, you have the floor for five minutes.

[Kevin Waugh, Conservative, Saskatoon South, SK]

Thank you, Madam Chair, and welcome, everyone.

Last month, in Canada's national newspaper, an article stated, “jobs aren't disappearing because of AI, but rather, they're being redefined.” That was an interesting article, because I have heard doom and gloom on AI and employment. I think we would see a dip on AI and then acceleration, but this is the first time.... We're not going to replace human connections. The jobs aren't going to disappear; they're actually going to be redefined, which is kind of interesting when we know that a lot of our young people in this country are suffering right now because of unemployment. I say that because—and I think you would all agree—young people are the ones who are going to drive AI in this country.

I'm going to start first with Ms. Davy.

I've read your Work in Culture report, “AI for Administration in Ontario's Creative Industries”. I note you found that “79% of creative professionals report using AI tools in their work, with nearly half using them often or very often” and they continue to use them. That was an interesting find from your organization.

Do you believe AI is creating new opportunities in arts and culture and in the job market? What are those opportunities, if you do believe that? How is it improving the landscape of Canadian content producers?

[Diane Davy]

I do believe it's creating opportunities. It's quite a fantastic tool. It has to be understood and used in that context. It does not replace human activity or human creativity, hence the need for appropriate training and for adherence to strong copyright policies and regimes.

What kinds of opportunities? There are all sorts of things. We're using it in our own small organization. As I said, we're typical of art service organizations. We're four people, overworked and underpaid. We've been using it for everything, particularly for grants. Grants, as you may know, tend to let you write 250 words in a box, but you have 300. Often, we're using AI for that kind of thing. It does help.

You don't just do it and then use it. One of the things we learned was a concept of human in the loop. With anything you do, you check because AI has a bias towards positivity, and it hallucinates and produces slop, so it can lie to you, too.

One person defined it as having a really smart intern but an intern with no life experience. You don't turn them loose without supervision and rigour. Yes, there are opportunities.

[Kevin Waugh, Conservative, Saskatoon South, SK]

You're the first one who hit on training. How will that be delivered? You were the first one on the panel today who talked about training.

[Diane Davy]

Several people did talk about training.

[Kevin Waugh, Conservative, Saskatoon South, SK]

Yes, but you were the first one who flagged it. How are we going to deliver that?

[Diane Davy]

I think it will be in all methodologies that we use for training now, which are virtual, in person, mentorships, internships, everything that, for example, Work in Culture does.

We're a very highly educated group, very well educated and very smart but generally lacking business and entrepreneurial skills training. If you think of an average artist, they come at their world through their techniques, their creative genius. At some point, they might think, “I want to make my living at this, but gosh, I have to sell it, and I have to market it, and I think I have to invoice and take care of my taxes,” and things like that.

We're dealing with a sector that already doesn't have a great depth of understanding of entrepreneurial and business skills training. Over that, there are digital skills, which we've all been coping with. I would say that, again, because of time and resources, the average organization in the sector has deficiencies in their digital skills, and on top of that is AI. There needs to be ongoing, rigorous training addressed from a variety of avenues.

[Kevin Waugh, Conservative, Saskatoon South, SK]

Thank you. We'll leave it at that.

Thank you very much, Ms. Davy.

[The Chair, Liberal, Lisa Hepfner]

Those were great metaphors, Ms. Davy, I have to say.

Mr. Ntumba, you have the floor for five minutes.

[Bienvenu-Olivier Ntumba, Liberal, Mont-Saint-Bruno—L’Acadie, QC]

Thank you, Madam Chair.

Good afternoon everyone.

Ms. Morin, given how easily AI has found its place in a changing environment, what measures would your organization recommend to safeguard against AI tools' reproducing or amplifying cultural or linguistic biases? I'm especially concerned when it comes to the fair representation of minority communities, the diversity of training data and algorithm transparency.

[Marie-Christine Morin]

We work on different levels. My colleague mentioned the digital strategy for our sector earlier. We worked on that strategy with representatives of the Fédération culturelle canadienne-française and, on a larger scale, with members of the French-Canadian artistic and cultural ecosystem. One of the pillars of the strategy is to influence public policy. Our role, both in Canada and abroad, is to convey these messages, and to make sure governments listen to civil society and act accordingly.

Your question is an interesting one. There's a lot of conversation on the international stage around the need to protect the diversity of cultural expression. Everyone agrees, and there are many ways to go about it. One of the measures we support, which was proposed by the Coalition for the Diversity of Cultural Expressions, is to add a protocol to the 2005 UNESCO convention, which Canada has signed. This would allow us to have influence at an international level in relation to what you are referring to.

Obviously, there is a lot of training and support work being done on the ground. My colleagues were talking about the need to train members of the ecosystem. We face the same challenge. To help us meet the challenge, we created what we call a coaching pathway. We realized that training wasn't enough, and that artists, creators and organizations needed guidance during the digital transformation. They also need support in situations where AI is needed, because this technology is changing the way organizations work and has an impact on their effectiveness and efficiency.

[Bienvenu-Olivier Ntumba, Liberal, Mont-Saint-Bruno—L’Acadie, QC]

Thank you very much, Ms. Morin.

Mr. Chan, could you answer my question as well?

[Kevin Chan]

I'm happy to. Is it the same question you asked Madame Morin?

Open-source code allows us to train models using data that are relevant to a specific culture or community. We talk a lot about training an open model like Llama to speak proper Québécois. I think it is theoretically possible. We did something similar for indigenous languages with UNESCO during the International Decade of Indigenous Languages. We programmed an open-source-code model to translate about 200 languages, including indigenous languages. We would like to use this model to help, protect and promote indigenous languages. However, since it's an open-source-code model, any community, whether in Canada or anywhere else in the world, could train it using culture-specific data and teach it to speak another language.

Cultural diversity is very important in AI models. A clear way to protect and promote diversity using these models is to train them in other languages using other data.

[Bienvenu-Olivier Ntumba, Liberal, Mont-Saint-Bruno—L’Acadie, QC]

Thank you very much.