THE STANDING SENATE COMMITTEE ON TRANSPORT AND COMMUNICATIONS
EVIDENCE
OTTAWA, Wednesday, April 22, 2026
The Standing Senate Committee on Transport and Communications met with videoconference this day at 6:45 p.m. [ET] to examine and report on the opportunities and challenges of artificial intelligence (AI) in the information and communications technology sector.
Senator Larry W. Smith (Chair) in the chair.
[Translation]
The Chair: Honourable senators, welcome to this meeting of the Standing Senate Committee on Transport and Communications. Thank you for your cooperation. My name is Larry Smith. I am a senator from Quebec and chair of the committee. I would now like to ask my colleagues to introduce themselves.
[English]
Senator Simons: Senator Paula Simons. I come from Alberta and from Treaty 6 territory.
Senator Mohamed: Welcome. I’m Farah Mohamed from Ontario.
[Translation]
Senator Cormier: René Cormier from New Brunswick.
[English]
Senator Arnold: Good evening. Dawn Arnold, also from New Brunswick.
Senator Lewis: Todd Lewis from Saskatchewan.
The Chair: Thank you, colleagues. I’d like to welcome everyone with us today, as well as those listening to us online on the Senate’s website, sencanada.ca.
We’re meeting today to continue our study on the opportunities and challenges of artificial intelligence, or AI, in the information and communication technology sector.
I would like to now introduce our first panel. From the Coalition for the Diversity of Cultural Expressions, Marie-Julie Desrochers, Executive Director, who is accompanied by John Degen, Chief Executive Director of The Writer’s Union of Canada. From Music Canada, we have Patrick Rogers, Chief Executive Officer, and from Music Publishers Canada, we have Margaret McGuffin, who is the Chief Executive Officer.
Thank you all for joining us today. Witnesses will provide opening remarks of approximately five minutes, which will be followed by a question-and-answer session with the senators.
[Translation]
I now invite Ms. Desrochers to give her opening remarks.
Marie-Julie Desrochers, Executive Director, Coalition for the Diversity of Cultural Expressions: Thank you very much.
For 25 years, the Coalition for the Diversity of Cultural Expressions has been the voice of Canada’s cultural sector across both English- and French-language markets. The coalition is made up of more than 50 organizations representing over 350,000 creators and 3,000 cultural enterprises. You have a fine example of that before you today, as I am joined by Margaret McGuffin of Music Publishers Canada, who testified at previous meetings, as have representatives of ADISQ, SOCAN and ACP, all members of our coalition.
Our work is grounded in the UNESCO 2005 Convention on the Protection and Promotion of the Diversity of Cultural Expressions, which recognizes the dual economic and cultural nature of culture. Canada played a decisive role in the development and adoption of this convention, and was in fact the first country to ratify it. Twenty years later, this commitment remains essential.
Generative artificial intelligence is becoming an integral part of our lives and is profoundly reshaping the cultural ecosystem. Some embrace it and integrate it into their daily lives, while others proceed cautiously or reject it altogether. People adopt it and feel comfortable with it to varying degrees, but that is not the core issue. Everyone agrees that it must be regulated.
Canada has acknowledged this. Last February, at UNESCO, the Minister of Canadian Identity and Culture, Marc Miller, supported adding a protocol to the convention, as did the 23 other countries serving on the committee. That was a significant step toward the adoption of a new binding international legal instrument aimed at complementing, strengthening and enhancing the 2005 convention in the digital environment, particularly in light of the growing challenges posed by AI systems.
At the National Summit on AI and Culture held last March, Minister Miller also stated that he does not intend to reopen the Copyright Act. Furthermore, he believes it is essential for creators to be compensated when their works are used. These are encouraging statements, but they still lack clarity and forcefulness.
If we allow our works to be used for free to fuel machines that will eventually compete with our own artists, we are methodically orchestrating our own cultural extinction.
This statement was delivered by a French senator in support of an innovative legislative proposal, but it sums up the spirit underpinning all of our demands.
I will now ask the Vice-President of the CDCE, CEO of The Writers’ Union of Canada, Chair of the International Authors Forum and an author himself, John Degen, to clearly outline our members’ expectations regarding the regulation of generative AI.
[English]
John Degen, Chief Executive Director, Coalition for the Diversity of Cultural Expressions: Thank you, Ms. Desrochers.
Thank you, chair and senators.
Our main asks of policy and lawmakers can be summed up by three words forming the acronym ART: authorization, remuneration and transparency.
For authorization, the Copyright Act obliges industrial users to obtain authorization from the rights holder prior to use, and this must remain the case. There is no need to introduce a new consent mechanism. We must simply reaffirm that our act operates on an opt-in basis and clearly state that Canada does not intend to change this, for example, by introducing an ill‑conceived text and data mining exception.
For remuneration, we are at the dawn of a new licensing market for valuable authors’ rights through voluntary individual or collective licences. We caution the government to not disrupt or destroy that market.
For transparency, companies developing and deploying generative AI systems should be required to disclose the training data used, explain how their models function and document their sources. Any content generated by an AI system should be clearly identified as such when it is made public.
As you know, this is not currently happening. In the book sector, for instance, we’re seeing a flood of AI-generated instant and fake books piggybacking on human work and genuine consumer interest in actual bestsellers. Because such fakery is currently not being honestly labelled, the U.S. Authors Guild and the Society of Authors in the U.K. have recently launched something called Human Authored Certification, a labelling program for books that provides consumer assurance the work is not the product of generative AI.
The Writer’s Union of Canada — my organization — will bring that program to our own members soon. My own upcoming publication, Seldom Seen Road — from Sudbury’s Latitude 46 Publishing and available in early May at fine bookstores near you — is one of the first books so labelled anywhere in the world, and that certification is based on my legal declaration that I did the work without using generative AI. To paraphrase what I have heard from my many thousands of writer contacts around the world; technology can check my spelling, it can check my grammar, it can help me organize and store my drafts, but I will not let it do the writing for me. I am a human author.
Thank you for your attention. I will gladly answer any questions I can.
[Translation]
The Chair: Thank you both.
[English]
I will now invite Mr. Rogers to give his opening remarks.
Patrick Rogers, Chief Executive Officer, Music Canada: Thank you, chair and honourable senators. My name is Patrick Rogers, and I’m the CEO of Music Canada, the trade association representing Canada’s major labels: Sony Music Entertainment Canada, Universal Music Canada and Warner Music Canada. Our members have offices filled with Canadians helping Canadian artists build careers and share their music with fans at home and around the world.
On AI, let me begin with a simple point: Our members are very excited about AI as a tool. It helps human artists elevate their creativity and find efficiencies in the recording process. My members are working with AI companies to develop new ways for fans to engage with the music of their favourite artists. But when it comes to AI as a tool, we have to get the rules right. The music industry knows from lived experience that the best way to do that is to uphold copyright.
Let me take you back for a moment. Not that long ago, the home internet created the possibility for peer-to-peer file sharing, and Napster and other piracy sites decimated the industry. It wasn’t the internet that decimated the music industry: it was the public’s willingness to become pirates. It was the breakdown of our public understanding that we — and I mean all of us — were stealing. I worry about that happening again today — or tomorrow — with AI.
The way we got out of the Wild West era of Napster and into the licensed iTunes era, where artists were paid, was to recognize and uphold copyright as part of digital innovation.
AI is the same. AI systems that are trained on music need permission from artists and rights holders. This is done by licensing, and creators should be able to negotiate the use of their music by AI in the marketplace.
I’m pleased to tell you that our industry has made massive progress on this front in the last few months. Major and independent record labels and publishers have announced landmark AI licensing agreements with AI companies. These licences ensure artists and rights holders are being compensated when their music is used in AI training. These agreements are proof of a growing marketplace for licensed AI that respects IP and artists.
I cannot overstate how pivotal these licences are. I am confident that, in a few years, we will look back on this period and see it for the game-defining moment that it was.
Parliament also plays an essential role in ensuring that this marketplace develops, and Canada’s artists and their creativity are protected. As this committee hears from stakeholders in this study, I hope that you will keep the following three points in mind:
First, the government should uphold copyright law. You will hear that AI companies need clarity on copyright, but that’s a euphemism for a new text and data mining exception so they can train on what they want without permission or payment to artists. The music industry does not suffer from a lack of clarity. Training without permission is theft. Full stop.
You may also hear that getting permission will only slow down AI innovation and investment. But let’s be real. By now, it’s pretty clear that no one is stealing music or the voices or images of artists to cure cancer, map the planets or improve crop yields.
Let’s have conversations on how to accomplish these important goals, but let’s not throw artists overboard to do it.
And when we talk about licensing, it’s critical that it be direct and voluntary, without rates or terms being imposed on creators. Academics will tell you compulsory licensing will give everyone certainty. But stripping artists and rights holders of their right to negotiate the economic value of their work is no solution.
I know that asking you to uphold copyright sounds like we are asking for no action. But that’s not the case. Our government should issue clear statements that training requires permission and compensation.
Senators, I hope that you will recommend doing just that.
There are other things the government can do, which brings me to principle No. 2. The government should require record-keeping and transparency by AI systems and require labelling of purely AI-generated content. AI developers should be required to provide meaningful information about training data so IP rights holders know whether and how their copyright was used to train AI.
You may hear that this is just too hard or too burdensome, but we know it’s possible, and good actors are already doing it. Labelling of AI-generated content helps Canadians know what they’re engaging with and protects human creators and their artistry.
Finally, the government needs to get serious about deepfakes. Not just the very worst kinds — the sexually explicit ones or ones that interfere in our elections — but all of the harmful ones because they put all Canadians at risk, including our kids. There are models for Canada to look at, like the U.S.’s NO FAKES Act that has bipartisan support and is backed by both the music industry and online platforms.
If Canada gets these principles right, it can be a leader in both AI innovation and creativity. I look forward to answering any of your questions.
The Chair: Thank you very much, Mr. Rogers.
Ms. McGuffin, you’re on the floor.
Margaret McGuffin, Chief Executive Officer, Music Publishers Canada: Good evening, chair and members of the committee.
I am the CEO of Music Publishers Canada, or MPC. I am here to discuss the need for the ethical and transparent development of AI models as Canada embraces the opportunity of AI.
Music publishers discover and develop Canadian songwriters and have made significant investments in the songs and scores that are heard every day on radio, on streaming services, in video games, in film and television productions, and on new, emerging platforms around the world.
Music publishers, composers and songwriters are already using AI tools, both in the studio and across their businesses. It allows businesses to unlock new efficiencies and scale their operations in an increasingly competitive global market. AI has the potential to support the work of human creators while also strengthening Canada’s cultural output and creative economy.
Unfortunately, the music industry has also experienced the mass theft of copyright-protected songs by AI companies, both on the input side for the purpose of training AI models and on the output side for the development and publication of unlicensed generative AI models. This poses serious risks for Canada’s creators and the companies that invest in them.
Strong copyright ensures that MPC’s members, songwriters and composers retain control over their work and are fairly compensated for its use.
When an AI company trains its models on scraped music without permission, rights holders lose both control and the ability to realize value from their works. The commercialization of these models is already creating market distortions and raising concerns about fair competition. This means Canadian songs are being used, without consent or compensation, to train an AI model that then generates new songs that replicate a creator’s distinctive sound, style and creative identity, effectively stealing their intellectual property.
Music Publishers Canada works with the International Confederation of Music Publishers, or ICMP. Evidence collected by ICMP over the past three years has shown that many of the world’s biggest tech companies have scraped copyright-protected music — created by millions of songwriters, composers and artists — to train generative AI systems without permission or licensing. To put it in perspective, nearly every song ever written by a Canadian songwriter has already been scraped and stolen by these AI companies without consent, credit or compensation.
The music publishing industry routinely grants licences to technology companies. AI developers should be no different. The emerging market for licensing music to AI developers should be encouraged, including by requiring AI companies to disclose and maintain records of all their training data.
So how should Canada proceed? The Canadian government must reject any calls for watering down a copyright system with a text and data mining, or TDM, exception. The Australian government made the announcement that they will not implement a TDM exception in December, and the U.K. government recently stated that a TDM exception is no longer their preferred pathway.
It is imperative that Canada approach generative AI in a manner that respects creators and incentivizes human expression. Doing so will not only support creators but also strengthen Canada’s creative economy. This will benefit not just creators, but Canadians as a whole.
Thank you.
The Chair: Thank you, Ms. McGuffin.
Would you like to introduce yourself?
Senator Miville-Dechêne: Sorry. I had a steering meeting in another committee. I was delayed. My name is Julie Miville-Dechêne.
[Translation]
I’m from Quebec. Welcome.
The Chair: Thank you, senator.
[English]
You’re not guilty. It’s okay. We still love you. Here we go.
Now, we have our list of questions.
Senator Lewis: Thank you for your appearance here this evening. You all talked about transparency in AI and how important it is for your members to be able to know that their work is being used.
Are there some ways that other countries have used the transparency requirements that Canada should follow?
Mr. Rogers: I think it’s really important, because on this topic, we run into this idea of, wow, that’s really hard. It’s billions of data points, and how would we even do that? I would recommend that AI companies that are looking to solve all these things should start with creating AI that writes a bibliography of all the stuff they use. I would point in that direction.
As far as other countries — Margaret, I don’t know if you have other thoughts on this — this is a big, million-dollar question and part of the global space. I think recognizing that it is part of copyright and that this is how we get to the place is the answer, and I think Canada is in a position to be a world leader on that.
Ms. McGuffin: If I may also add, a lot of A is being used to make these systems more transparent. Over the last month, I’ve seen two different technologies that are using AI to figure out — if you know what’s going in, to figure out what is coming out, by the recording, by the melody, by the lyrics separately, to a percentage point. So the amazing thing is that AI is going to help us make this more transparent.
[Translation]
The Chair: Ms. Desrochers, would you like to add anything?
Ms. Desrochers: I might add that the European Artificial Intelligence Act contains provisions requiring transparency and requires a sufficiently detailed summary for now. However, a report recently tabled in the European Parliament and adopted by an overwhelming majority proposed improvements to these provisions to ensure transparency by way of a detailed list enumerating every piece of copyright-protected content used for training purposes. These measures are already in effect in Europe and will improve there.
[English]
Senator Simons: I was a journalist for many years before I became a senator, and so I felt a particular chill today when I read the story about McClatchy, the big American newspaper company that has announced it will be using AI to take its reporters’ stories, rewrite them, reconfigure them and put them out with the reporters’ bylines attached. The vice-president of McClatchy said, “Journalists who are defiant will fall behind.”
This raises questions for me for writers. Maybe this is a question for Mr. Degen.
Many writers don’t have copyright of their own material, but they have a reputational connection to the material that they’ve written. What do we need to do to protect writers from reputational damage if AI is repackaging, regurgitating and dumbing down content that they have written that has their name attached to it, even though they may not hold the copyright?
Mr. Degen: Do you want me to respond to that?
Senator Simons: Yes, that is why I asked you.
Mr. Degen: Okay. Well, within our Copyright Act, we have a system of moral rights which deals with attribution and keeping your work intact. I think we would have to stress the strength of that part of our laws. I share your discouragement about that kind of development.
I think in my part of the business, which is the book publishing industry, I’m seeing 95%, above 95% of authors, who want to have nothing to do with this technology in the creation of their work.
There’s some small segment that wants to experiment here and there and thinks of it as an interesting tool, which is fine.
But I can’t see that sort of full partnership that you’re describing in journalism being something that would be at all attractive to my sector.
Senator Simons: “Partnership” would not be the word that I would use.
We heard testimony yesterday from Michael Geist, who is a professor of copyright law, among other things, at the University of Ottawa, who flagged a concern for all of us that, if we imposed strict copyright rules on the scraping of models, that Canadian content might be disadvantaged because we wouldn’t hear from Canadian voices, and particularly for Ms. Desrochers, we wouldn’t hear from minority Canadian voices because we would get mainstream American content. AI would become blinded to the diversity of Canada.
I’m going to ask Ms. Desrochers this question, perhaps Ms. McGuffin or Mr. Rogers.
Mr. Rogers: Yes, please let me in.
Senator Simons: You can be in my neighbourhood. What do you think about Professor Geist’s concerns?
Ms. Desrochers: Do you want to go first?
Mr. Rogers: I would love to.
Ms. Desrochers: I don’t want to.
Mr. Rogers: I take note of Professor Geist’s comments on transparency, on which I think we agree. I think the fact that AI is a copyright issue as a whole, I’m glad to see him agree on.
I completely disagree with the premise of that argument in that Canada makes some of the best music in the world. The top global stars are Canadian all the way down the list.
Today’s number one viral video in music is a French Canadian group making crazy rhythmic music.
This idea that will leave us behind; I completely reject the argument that we won’t be able to help artists if we don’t sell out artists.
[Translation]
Ms. Desrochers: As mentioned, I would add that these systems have already used virtually all of the available content. First, they are capable of using it. Second, if content has been used at this point, that’s because it has value.
Just recognizing that value does not mean that, all of a sudden, it no longer has any.
Now, I think we have to ask ourselves this question: If there are minority groups that are not currently represented in the results that come out of these machines and compete with their real work, are they really in a hurry to be there for free?
I reject that as well. Of course, we want all voices in Canada to be heard. These voices are human, they have value, and we don’t think recognizing that value will cause us to hear them less. Quite the opposite.
[English]
Senator Simons: Thirty seconds for Ms. McGuffin.
Ms. McGuffin: There is a problem. If there is a bias on the data going in, there will be a bias coming out.
We represent women producers and run an accelerator mid‑career for them. If only 6% of the songs you hear now are fed in, and those are the ones produced by women, you are going to have that bias coming out of the output.
[Translation]
Senator Cormier: My questions are for Ms. Desrochers.
The report produced by the Standing Committee on Canadian Heritage states that the Government of Canada must establish a clear opt-in consent requirement for the use of copyrighted works in the training of artificial intelligence systems, ensuring that creators’ works may not be used for text and data mining or model development without their prior authorization.
If I understood you correctly, I believe you said that wasn’t necessary. Can you clarify that statement?
Ms. Desrochers: Yes, of course. In fact, in its report, the Standing Committee on Canadian Heritage begins by emphasizing the principles of “art, authorization, remuneration and transparency”. This language suggests that Canada needs a consent mechanism, but the Copyright Act is already a consent mechanism. We don’t need a new mechanism, and saying that we do suggests that there isn’t one.
It’s just in the wording of that recommendation. It’s important to remember that, in Canada, our law is already a voluntary system, and it must remain so.
Senator Cormier: Thank you. In the Standing Committee on Canadian Heritage’s report, recommendation 6 states that the Advisory Council on Artificial Intelligence should add two more members from the cultural community and undertake work to determine the threshold of human intervention required to grant copyright to an AI-assisted creative work.
How do we do that? What criteria would apply?
Ms. Desrochers: A lot of people are interested in that issue. I’ll start by saying that we had some reservations about that recommendation because the council includes presidents of major AI development companies. If we want to discuss these things, there will have to be sectoral subcommittees.
It’s different for music, audiovisual and book publishing, and I should point out that certain guidelines are already in place. There’s currently case law taking shape with respect to the Copyright Act.
I know it’s very tempting. We all want to know what percentage of a work is human and what percentage isn’t. Those conversations will take place within cultural communities with the people affected. Anyway, we had some reservations about that recommendation.
Senator Cormier: Obviously, Canada is not alone in this. Are there any models out there?
Ms. Desrochers: There’s already a lot going on, even here. If you go to the Telefilm Canada website, for example, you’ll see that they’ve already adopted an initial policy to explain what they’re asking creators to do. It’s important to understand that determining the degree of human intervention also depends on the context.
For example, if you want to award copyright or reward someone, the definition at a gala will not necessarily be the same. If you want to set quotas, it won’t necessarily be the same definition either. Unfortunately, there’s no simple answer to that question right now. However, we shouldn’t oversimplify things by suggesting that a council can look at the issue and resolve it once and for all.
Senator Cormier: Mr. Rogers, you talked about direct and voluntary licensing. I’m a bit obsessed with understanding exactly how the licensing issue works. Can you tell us more about that?
[English]
Mr. Rogers: Absolutely. In today’s market, the copyright is you’re allowed to directly negotiate with anyone who wants to use your music. This is a very important sort of foundational piece of our copyright law. And it allows artists and rights holders to withhold their music from services right now.
You will remember that, at the launch of iTunes, not all of the world’s greatest music was originally available. Some of the biggest bands of all-time held out. That led to changes to the platform until they came on.
In that case, that made the services better for both consumers in that the world’s best music was on them, and better for the artists because now the world’s greatest music was truly on them. We were able to raise prices and do these things to create a better back and flow with the artists.
You will hear some people say, “Oh, but that is too hard. They might hold out. I won’t be able to build my AI model if I don’t have the world’s greatest people. They may not let me use it. Oh, my goodness.” That’s part of the negotiation.
The compulsory licence is when the government steps in and says, “We will set a rate. Don’t worry, AI company; don’t worry, any platform. We’ll set a number” through the Copyright Board Canada or somewhere else, and as long as the service pays that price, then you get to use it. Historically, we provide this data all the time. This leads to lower prices for artists and a worse experience overall.
Right now, the people who have scraped the world’s music are still inside the bank vault. We are negotiating with the robbers while they are in the bank vault, and if you say they can get out by just paying a fine, then they will. However, right now, we have them where we want them. We’ve proven they’ve scraped it. We know it is a copyright issue, and so we need to do this.
I take a slightly more generous interpretation of the Heritage Committee’s first recommendation, which you asked about earlier. I do so for two reasons. The first is the history of this. I have now been talking to government, officials and political decision-makers regarding AI for years. We said the other day that we wrote our first ATIP in 2019 about what the government thought of AI, so we’ve been working on this for seven years.
Only recently has it been agreed in government circles that it is a copyright issue. It wasn’t that long ago that government-sponsored AI forums started off with a warning: “We’re not talking about copyright.” So, thanks to litigation led by my members in the U.S. and around the world, we have proven beyond a shadow of a doubt that AI is not a big giant brain that listens to music like all of us do. It is, in fact, the rearranging of ones and zeros, and that rearranging has a name, and that is copyright.
I give generous interpretation to that first recommendation of trying to say, in parliamentary language, not copyright lawyer language, that this is a copyright issue. The work they did there should be applauded.
Ms. McGuffin: I agree completely.
We’ve been licensing services. I have been in the business for too long, even before Napster. A lot of the things you’re hearing now, companies were saying those things 30 years ago at the dawn of Napster: It was too difficult; there wasn’t going to be enough money; there was too much data. All of those things are being said now.
But we’re in the business of licensing. We’ve been doing it. The TikTok deal looks different than the YouTube deal, which looks different than the Spotify deal. We adapt. We use valuation techniques, and we can get the right deal customized to the service. That is just going to continue into AI. The people who are telling you that it can’t happen are people who have never come to talk to us. They are hiring lawyers for litigation and to avoid licensing as opposed to hiring lawyers who know how to do the licensing.
Senator Mohamed: Thank you for being here. For full disclosure, I used to be on the Music Canada board, so I have known some of these characters for some time. I say “characters” because I know about Mr. Rogers.
I’m interested in three things. First, in copyright law — and please correct me if I’m not right — there is no minimum threshold around content before you are infringing against and — I am going to use your word — stealing somebody’s product. In the music business, with AI, should that still be the case; should there be a minimum threshold? I think we all agree that artists should be protected and there should be no theft, but does it make sense to have a minimum threshold? That is my first question. Ms. McGuffin, I’ll come to you first for that.
Second, assuming you have the right framework in place — so you have a regulatory stop, the right laws and the ability to track when companies are using, if we decide to go with the idea of tracking databases and material that goes in — should Canada consider having an AI ombudsperson?
Ms. McGuffin: I will talk about a minimum threshold. I deal in songs, not recordings, and we will license the very smallest five seconds into that favourite commercial that gets you and hooks you. Very small bits of licensing are key to how songs are licensed now. That should continue.
Also, the valuation changes. If there’s a larger amount of the output that is music, it is a different type of deal than ones that have a smaller amount of music. This has already happened.
Senator Mohamed: So could that be put in compensation? I understand there is a sliding scale, but if you think about a company that is doing this — as you said, it is billions of pieces of data. How do we wrap our heads around this?
Ms. McGuffin: We find that the attribution companies that are being developed and companies themselves that are responsible and ethical are developing them within their system.
This is something for which we can use technology, and we have used technology in the past to ensure we can process the data in a way that those little micropennies add up into a real payment every month or every quarter for a songwriter.
Senator Mohamed: In the event that you have everything right and you want to make sure it is enforced, an AI ombudsperson — does it make sense?
Mr. Rogers is itching.
Mr. Rogers: I will add to that.
I think the output sampling conversation is a bit of a red herring, because in order to get to that output, you needed to ingest the song in the first place. At that point, it doesn’t matter; you used the song. That is the licensing that our members are busily working on.
Then, you know, we are excited to see what future licensing looks like in terms of, then, determining output value. But that first part of ingesting it, that is where copyright happens.
On the ombudsperson piece, I would say this goes back to our core argument, which you have heard throughout the day: Today’s laws and setup take care of this if we make full use of the law.
I wouldn’t be doing my job if I encouraged sliding away from that, but I genuinely think that what would be covered by an ombudsperson would be covered by just enforcing the Copyright Act.
Senator Arnold: My questions are for Mr. Degen.
When you said that you put a human-authored certification on your book and that you are the first one, do you feel there are others around who are also interested in doing that?
Mr. Degen: Absolutely. We have advertised this service to our members. There has been a lot of feedback from writers wanting to be able to so certify themselves.
This brings us back to the idea that all of this and the transparency piece are too difficult for the technology industry.
I work in a very tech-savvy business, but we produce analog products as well as technology products. We managed to do it, and it wasn’t hard. We would much prefer that the labelling be on the artificial intelligence side. We were there in the marketplace first, so the products that are competing with us and trying to pretend to be us should be the ones that have to label.
Senator Arnold: How big of a problem is that, though? As far as the AI-generated books that you see, such as the weird ones on Amazon and stuff like that — are they really a threat to you? They are usually garbage, aren’t they?
Mr. Degen: It is a huge problem because of the lack of transparency. I agree with you that the quality is not there, but you do not know that before you buy it.
What happens is this: You bring a book out, and it has a very short window of sales potential before it goes into the long tail of the rest of its life. You want to maximize your impact during that short window. Say you go up on Amazon, start to rise in the rankings, people notice you are a bestseller, and it is starting to work for your book. Then, suddenly, there’s a fake book in the listings beside you that looks an awful lot like yours, looks like it covers the same subject matter and is directly competing with you for sales, and that book was created in an afternoon while your book took five years to write. It’s not a fair fight. This is happening, and people are losing valuable income because of it.
Senator Arnold: That’s very interesting. Thank you for that.
As we delve into this, I’m finding I don’t even know what I don’t know. Are there any issues around audiobooks?
Mr. Degen: Sure, it’s the same thing. We’re dealing with fake books that have been created to essentially look like the genuine article, and it’s just quickly packaged up through AI. It’s just audio rather than print. It’s the same issue.
Senator Arnold: Have there been any effective lawsuits or copyright infringement cases of people taking people to court over this, or are they in the bank vault negotiating with the robbers? I’m just trying to understand.
Mr. Degen: I love that expression. Thank you for that.
You know what, we spend an awful lot of our time in court, and we don’t want to spend our time in court. We want to spend our time making books and selling them to enthusiastic readers. So, yes, we will take legal action where we can.
A lot of this is — you hear the expression all the time — like playing whack-a-mole. You find a fake book. You do a takedown request. It pops up somewhere else, or something like it pops up somewhere else. We spend, frankly, far too much time chasing these things down. If there was a proper labelling system and if our Copyright Act were properly enforced, I think we would deal with this problem a lot better.
Senator Arnold: Thank you.
[Translation]
Senator Miville-Dechêne: I wasn’t here for your opening statements, so forgive me if I make you repeat yourself.
The last time we spoke, when you were here, it was about Bill C-11. The situation was pretty tense because of the whole algorithm thing. You were very worried about how visible francophone music would be, especially in the public square, and about the listening habits of young people in particular.
Unless I’m misinterpreting your position, you seem much less worried about artificial intelligence because you’re hoping to get copyright back. I don’t think it will be easy, though.
Are you more optimistic, or do you think the advent of AI will have a greater impact on a culture like ours, the francophone minority on this continent?
Ms. Desrochers: Thank you for the question. It gives me a chance to follow up on what I said earlier to Senator Simons about members of minority groups.
I would say it’s about governance. It’s not just about making data available so it can be processed properly.
Going back to C-11, the fact that francophone Quebec music is available on all platforms does not guarantee that it will be highlighted. It’s not just about providing content. That was the gist of my answer earlier. We need to make sure this content is regulated so we know how it’s being used.
I would also encourage you to reach out to minority groups. We certainly have some in the coalition who have specific demands that complement our own.
Going back to the potential of generative AI, when we talk about AI, we’re talking about a number of different things. I recently had an exploratory conversation with someone from Mila about a project to improve content discoverability using AI. In this case, it wasn’t about using copyrighted works to generate competition; it was really about improving searches. For example, if you ask for a list of the top 10 films without any specifics, you get 10 American films. However, if you ask the same question, but you’re using an agentic tool to point research to local content, you’ll get a different list of 10 films.
I’m not saying I’m optimistic, but AI can open up some possibilities.
Obviously, our main demand is for remuneration. As I said earlier, without remuneration, you’ll get extinction because that’s key. If creators are no longer being paid to do their primary work, they’ll disappear because they can no longer make a living from their art.
The coalition held a conference in February, and illustrators attended. Illustrators are among the first to be affected. I would also add to what Mr. Degen said about how a book may have been written by a human, but the cover was generated by AI.
There are all kinds of cases when we say, “This book is human.”
When people in Quebec realized this, there was a solidarity movement in the book industry. Several publishers agreed to join a movement to stop using AI and support human creators and illustrators instead. Those illustrators tell us that they’re already feeling the loss not only of contracts, but also of jobs right now. It’s real. All the entry-level jobs people could do to make a living as an illustrator are disappearing. They used to do advertising work in addition to comics, but the value of that is declining across the sector.
The Chair: Thank you. We’ll now begin the second round of questions with Senator Cormier.
Senator Cormier: My question is for Mr. Degen. I’ll try to be clear. I’ll add to what Senator Simons said earlier about publishers reconfiguring reporters’ stories, if I understood correctly.
My concern is about ensuring linguistic diversity. We know that one of the great things about the French language is how diverse it is. Our regional cultures are expressed through words, through texts. When people use AI, such as ChatGPT, you can tell the language is being flattened.
Are you concerned that generative AI will flatten language and eliminate minority or regional cultural specificity? Can you comment on that?
[English]
Mr. Degen: Thank you for the question, Senator Cormier. A quick answer: Yes.
You mentioned fears of things happening, like the potential for the flattening of language and the potential for the loss of certain specialties. I don’t fear that’s happening; it has already happened. The profession of being a translator or interpreter was one of the first to be disrupted by AI, and it has had a net negative effect in our industry.
We want to see the same kind of commitment that Ms. Desrochers was talking about. For instance, for book covers, we want to see the same kind of commitment for translation of works. Translation of works is a lot more than just dumping a work into an AI and saying, “Make it French.” You need a human understanding of a culture and the language to produce anything of quality.
So, yes, I do fear that it will happen more, but I’m afraid to say that it has already happened.
[Translation]
Senator Cormier: Did you have something to add on that?
[English]
Mr. Rogers: I think the path forward on this actually rests with all of us in realizing that it’s not a magic box that suddenly makes it French. I think that is part of the sales pitch. As we all come to grips with what is going on and realizing that it is the rearranging of what it has been given, then we can understand that.
I think about music, and I think about the evolution of music through time. When everything is licensed and legal, maybe we will be able to have a debate: I like this AI-generated, licensed, legally AI-generated music, or I don’t. I like this French translation because it’s good, or I don’t like this French translation because it is bad. In the meantime, while everything is scraped and stolen, it’s all copyright infringement, and we should be concerned about it.
[Translation]
Senator Cormier: I was in a taxi in Mexico not long ago, and the driver was a musician, so I asked him to play me some contemporary, modern Mexican music. He played some tracks for me, and I asked if he could play me something really contemporary. I was with friends, and we all thought it was really great. I asked him who wrote the song, and he said he did; he created it with AI. That’s been bugging me ever since because I thought it was so good. It actually reminded me of Jovanotti in Italy.
This is happening. From an ethical standpoint, what do we do about it?
[English]
Mr. Rogers: I was a guest lecturer to a university class, and a similar instance, where a young artist came up and said, “I make music now with a 100-piece orchestra, thanks to AI. You’re calling me a pirate? I’m not a pirate. I’m an artist.” And I said, “Well, this is why my companies are busy trying to license this so that one day, you can make that music using the 100-piece orchestra, generated by AI. But in the meantime, you need to understand that all the orchestras from all recorded times have created that music for you.”
His point, he was like, “No. I have pushed the button, and it makes the music.”
I think that we really do owe it to ourselves, not just here in committee rooms, but in our everyday lives, to point out the difference between what a human made and what the robots made.
Senator Cormier: How can society learn to recognize it?
Mr. Rogers: This would be my answer to this. On my drive to school with my girls Grace and Rose, we listen to Splash’N Boots, and there’s a violin lick in one of the songs. I work at Music Canada with Music Canada’s creative culture advisor, Miranda Mulholland. She does the violin lick on that recording. It’s up to me to tell my girls, “We know the person who made that sound. It wasn’t a keyboard press.”
As the son of a journalist, we have to do that with journalism, we have to do that with music, and we have to do that with books. We have to celebrate the artist behind it and make a point that we recognize it. Because if we don’t, then we will lose the understanding of what is what.
Senator Simons: I have a slightly different question. This is for Mr. Degen. I have a lot of writer friends who are being harassed; I would almost call it by AI bots that are promising them marketing, that are saying, “Oh, I represent a book club; come speak to our book club.”
And it’s all a scam to make them pay for something. In their inboxes, they cannot get away from this. I wonder if you can speak to that part of the equation. What can we do to protect writers from the AI scammers out there, from the kinds of fakes that Mr. Rogers was talking about, that are designed to defraud writers of what little money they have?
Mr. Degen: You’re really pressing my buttons with that one, because this is the kind of thing that comes across my desk an awful lot at The Writers’ Union of Canada. We have been dealing with an awful lot of fraudulent behaviour, especially across email, aimed at writers, which to me is terribly ironic. They’re trying to squeeze money out of writers, who, frankly, don’t make a whole lot of money to begin with. So it’s a weird target, but it is a genuine target of these scammers.
Look, we talk about transparency; we talk about labelling. What you’re seeing there is a scam; it’s a fraud. It should be illegal. We should have an enforcement mechanism to pursue the illegality of it and make it stop. At the moment, we don’t. At the moment, it’s really my job, and the job of The Writers’ Union of Canada, to educate our members and to say, “This is what this looks like; we’ve seen this before; I know it’s going to make you feel really good because it sounds like somebody read your book, but that is probably not the case if it comes at you as an email asking for money.”
There’s a lot of work that we do on that end. We would dearly love to get out of that business, and we’re quite resentful of the technology for having put us in that place. But you’re right, that’s where we are.
We all hear about the wonder of what AI can do for us and can deliver for us, curing cancer, and taking us to Mars and everything. The reality is that it is sending us scam emails. It’s producing revenge porn. It’s negatively affecting our elections. It has not been a very good rollout, I will say that. And I think the government has a serious role in making sure that our consumers are protected and our creators are protected.
Senator Simons: Thank you.
The Chair: We have finished the first group of witnesses, and we thank you very much. It was an outstanding session.
This is a fantastic introduction here. We have from GetFact.ca, Mr. Dinnick, you’re the Chief Executive Officer and Co‑Founder. Give us a little introduction to your company, if you would.
Wilf Dinnick, Chief Executive Officer and Co-Founder, GetFact.ca: Thank you very much. My name is Wilf Dinnick. That title is slightly elevated. I represent a group of Canadians. We call ourselves GetFact.ca, or Laura AI, named after Laura Secord. And we’re building the tools and the networks for Canadians to fight against mis- and disinformation. We’re a mission-driven organization, all volunteers, and we’ve been in action for about a year and a half.
The Chair: Laura Secord — it must be a sweet gig. It’s my last night. Mr. Dinnick, would you love to give us your five-minute introduction to position what we’re going to talk about?
Mr. Dinnick: I think for every file on this committee’s agenda, and frankly, probably every ministry in this government comes down to the same filter, what do Canadians believe? I think I can speak for most people in this room, and I know I can speak for millions of Canadians, who, when they get an email, open their phones and look at social media, they have to ask the question, “Is this true? Do I believe what I’m seeing?”
The fact is if Canadians don’t have a shared set of facts on the critical issues in this country — crime statistics, public health, education, climate — if we cannot agree on a shared set of facts, it’s going to be very difficult for this country to go forward and deal with the challenges that we’re facing.
At GetFact.ca, we look and tackle mis- and disinformation, but the reason I’m here, and a lot of our members, is I spent a career as a foreign correspondent. I worked as the Middle East correspondent for ABC News, I worked at CNN International and here in Canada for Global National, and I spent a good part of my time in conflict zones, war zones and crises. I will tell you, there’s a pattern in all of them, and that is that citizens lack access to trusted information in those environments. The institutions that they trust start collapsing, polarization begins, and that leads to violence. It would be very foolish for us as Canadians to assume that we’re immune to this.
So we have to address mis- and disinformation in this country. Before I get to the solutions, which I will say are cost-effective and easy, we have to also admit the environment that we’re working in. I’m very proud that we have a government talking about sovereign AI and Canadian compute, and we have the Vector Institute and Mila. But we have to be realistic. The real merchants for our information for the foreseeable future are large AI companies based in the United States, and increasingly in China. They are private corporations that have no obligation to pull back the curtain and tell us how they got to a piece of information when we searched it to decide who we’re going to vote for or looking up something about Canadian public health or learn something about Canadian education; not only what they found, but also what they’re hiding or what we’re not seeing. That won’t happen. And we know that because we have a history of it in Canada dealing with those big AI companies.
Also, recognize that they’re not in the business of verification. They don’t want to check facts. Not only politically are facts not really their mandate, but they’ve made it clear that as accelerationists, those companies, the guardrails are off. It’s off the leash. They just want to grow. They’re not aligned with Canada’s needs. We have to address this problem here in Canada.
Now, other countries have done this, and we’re asking for Canada to adopt some of the same mandates.
The first one is standard verification of information. That sounds like a huge ask, but really it’s quite simple. That is when you see a piece of information, a video, or a claim, you can click on it, and it goes to the provenance of that information. You can see the source of it. This has to be seen as the ingredients or the nutrition label on packaged goods. You can like it. You cannot like it, you can ignore it, but it’s there, and we can all go to it.
The second thing is to now prepare for how we are accessing information, which is through bots and through AI agents. For instance, every ministry in Canada should have marked down files and APIs which our agents and bots can access easily. I can speak to it in detail after, but what it does, it cuts out the middle person. It cuts out anybody, like the platforms, et cetera, who is trying to manipulate us with that information or warp information in order to achieve political gains. And we know this is happening on a daily basis with malevolent foreign actors, and now, sometimes, the call is coming from inside the house here in Canada.
Let me sum up by saying this is not a technological problem. This is just making sure that Canadians have easy access to the source of the information that they are looking for, and they are looking at and it is absolutely easy to do, and it’s cost-effective.
The Chair: That was an excellent introduction.
Senator Dasko: Thank you for being here. I’m going to ask you to explain how it works. How do you verify? Let’s say you’re looking at something in a newspaper or something comes on Google, or you’ve used an AI connection to learn something, or what about television? What if I’m watching CBC or CTV and I want to verify something?
Mr. Dinnick: Let me say first, this is not out in the public domain, but about a year and a half ago we got some engineers, volunteers, military veterans, JTF2, and AI engineers who volunteered their time to help build one of the models to prove this would work.
We started with news. You drop in a news link, and it goes and searches and finds all of the information around the claim in that news article. So you can see the research based on that claim.
Let me give you a little example; a 250% tariff on dairy going into the United States. You click on that fact, and it goes to the source, and it shows there is actually a tax, but guess what? It’s never been excised. Then Canadians can quickly see that.
This exists. Universities have built this model, we built the model, and yes, it’s not widely in the public domain. And this is not conspiratorial, but the large platforms will not build this. When we went to talk to these platforms, one of the big engineers said to us, “Please build it, because no one else will.” The reason is the platforms do not want to get into the verification game, as I said. If Google, Meta or Open AI were to check a fact and claim a fact, they are open to litigation. They’re not going to do that.
Second, they’re all racing for us to love one of the models, so they’re always telling us what we want to hear. So we’re not always getting the facts. We’re not seeing how they make the decisions to get those facts. They’re not interested in sharing that because they’re private, competitive AI models. None of this serves the commercial benefit of these Open AI models. They’re not going to build it.
We built one; Mila has one. They’re quite simple; you can put them as a web extension or a web browser. They need to be governed, of course, and they need to be overlooked, but they can be used by all Canadians to quickly see the source of information.
Again, this is not about technology. This is about changing our understanding that we cannot have gatekeepers anymore to decide about pieces of information. If there is a claim, as you say, in a news article or a social media post, Canadians should be able to quickly see the data behind that claim and decide for themselves whether that is true or not.
Senator Dasko: I would think there would be a demand for this kind of verification, given the environment we’re moving into with all the disinformation and misinformation. Isn’t there a demand?
Mr. Dinnick: That is a very good question, senator. We launched our first one, and we discovered something quite astonishing. We called it GetFact, and we put it out there, and people started using it but more as a tool that started to inflame the social media environment, pointing out somebody was wrong, somebody was right, et cetera.
It is a bit inside baseball, but we are at the forefront of this. We are launching a second model, which is much more agnostic, which goes right to the source, and, like the nutrition label on a bottle, you get to see it, and you get to decide. We are building that.
I am not here to shill our model or try to pitch. I’m here to say, whoever does it — there are many models — the Canadian government should start supporting this and advocate that Canadians have the tools to be able to access the source of any information they see.
Senator Dasko: Thank you.
Senator Lewis: Thank you for your comments. You mentioned it can be done with a manageable cost. Who pays for it? What is the revenue source?
Mr. Dinnick: The government could pay for it, but let me give you an example. We built one, almost all volunteers. I would say that the thing is worth $300,000 to $400,000. Google generously gave us $200,000 in compute, and we have not gone through nearly that. You are talking a small amount of costs in order for usage.
I’m not sure that I could put an exact number on it. In terms of government spending, it’s almost nothing.
Also, there is a private-public tool here. If you are a corporation, you want to ensure that the information about your sector, the information that your juniors are feeding up to you, the people who are writing your speeches, are able to check the facts and ensure that is reaching the top. There are models around software to service that this can be used.
Senator Lewis: Are there other jurisdictions pursuing this kind of model?
Mr. Dinnick: There are. However, ours and some of the university models are unique. Many of them do something which goes along and points out who is right and who is wrong, who is lying and who is not.
What we have learned and are seeing to be very successful, and what we have learned needs to happen is it is agnostic. We have to let people decide for themselves what they like.
This is the major problem. We have spoken to many different ministries and different leaders in government. What we have come away with and what we understand, which is absolutely true, is there is such a low esteem of institutions in this country that the moment an institution starts declaring facts, it backfires.
This really needs to be in people’s hands. That is why we’re advocating for this model. This model is in users’ and peoples’ hands, and, again, they can look at the facts and decide, “Okay. That is where that climate statistic came from. I don’t like it,” or, “I believe it. At least I know that is where it came from, and it is not somebody else warping those facts.”
I’m sure you’ll all remember the ostrich cull and what a big news story that was. As you know, that story was huge not because people cared about ostriches and not because there was a lie in that: it was a trigger point. People gamed the algorithm and targeted some soft and touchy issues: government overreach that people felt upset about and bird flu, closely aligned to COVID. It triggered a lot of people about the way that the government handled COVID, like it or not. A thing like bird flu, you cannot explain it on social media. That takes a little time to understand.
These are things that I think that Canadians, when you have to click to a source so you can quickly see and understand the story, or at least know that it is there, I think that diffuses those kinds of issues.
Senator Lewis: Thank you.
Senator Simons: I hate to rain on this parade.
Mr. Dinnick: Please do.
Senator Simons: I was a journalist for 30 years. I’m dubious of the human capacity to tell a fact from a fiction. I’m even more dubious about AI.
Months ago, I watched a movie that had a bittersweet ending. It was an old film. I went online to see who was in the cast. I noticed that the precis of the film in Google had the wrong ending. It said the couple lived happily ever after, but the point of the movie is that they didn’t. I thought, well, that is a fairly benign thing to get wrong, but it is wrong.
I notice this now all the time with the Google summaries that you cannot make go away. They are just wrong. Sometimes they are precis of my own articles, and they get the point of my article wrong.
I don’t know how you can rely on AI to tell what is fact and what is fiction. You have been a war correspondent. How on earth can AI tell me what is happening in Gaza when nobody else can?
Mr. Dinnick: It is a great challenge. Thank you, senator.
It addresses the exact problem you’re talking about. Those little summaries, that is not what we’re relying on, and we’re not relying on a large language model. We’re relying on the domain expertise of editorial guidance to say that this claim around climate comes from this research paper, and the tool only takes you — when you ask where that came from — to that research paper in a second.
Right now, if you are dubious about it, it takes you three searches on OpenAI or Claude to maybe get to that research paper. What we’re doing is saying not to trust these large language models to wrap it up in a nice bow and make it all pretty and make you feel as if you got the right answer. We’re going to take you to the source of that information. The source of that information you may not agree with, but that’s where it came from. That is simply it.
We agree with you.
Senator Simons: Snopes.
Mr. Dinnick: It is along the lines of Snopes, but there is no human there looking through it. It is the technology that says, “You are asking — I see a claim here. Where did that come from? This is the research. They got it from here, or, “This is what it is based on.”
Senator Simons: I did original reportage for years. I would go out, interview people and report what they said.
Mr. Dinnick: Yes.
Senator Simons: There is no external source. You have to rely on your belief that I’m not going to make up quotes from people. You have to rely on your belief that I was a credible, professional journalist and that I did not, intentionally at least, put mistakes in my copy. But there is no way to fact check that. The New Yorker fact checks people by calling up and seeing if that is actually what they said in the interview.
Mr. Dinnick: But, senator, you worked for a news organization that demanded that excellence and probably checked your facts. Am I right?
Senator Simons: No. Daily newspapers don’t check facts. The fact check is that hundreds of thousands of people read the newspaper, and if I screwed up, I certainly heard about it.
There is no objective correlative to original source reporting. If you went into a war zone and witnessed something and reported on it, there is no source. You are the source.
Mr. Dinnick: Yes. What I’m trying to get at is if there is a claim out there and there’s no backing for the claim, you are going to see that. These models, you click through and you ask it, and, actually, no, that doesn’t exist. That claim that is being made, that doesn’t exist, or, yes, here is where the source of that claim comes from. That is what the tool is, but, yes, it does demand a little human effort to say I don’t believe that or I believe that.
Right now, it is much worse because there are filters there and platforms that have malevolent actors, people stirring it up, so this is a much better tool, and it puts it in the hands of people.
Senator Simons: I have many more questions and wish to go back on the list.
Senator Arnold: Thank you for being here. This is fascinating. That quote is in my head. “If we can’t agree on the facts, democracy dies.”
I’m not as good a researcher as Senator Simons. I have been thinking about this a lot lately. I was harkening back to about 2016 when Barack Obama was still the president, and we were filled with hope, and then everything kind of went really bad during COVID. All of this stuff came out. All of this negativity. All of these stories. I was an elected politician at the time, and it was, like, where do you start? I felt overwhelmed by all of the fake news that was coming out.
I think something like this sounds really great. I do know that law firms are using software like this right now.
Mr. Dinnick: It is almost the exact same thing. Sorry to interrupt, senator, but they are using databases around legal parameters, and this one does the same thing except it looks at sources — government sources and trusted sources, like news, where we know they have double-checked. Sorry to interrupt, but that is exactly it. It is like Harvey and other systems.
Senator Arnold: Don’t say “Harvey” because that reminds me of Suits. That’s what AI was doing. It was quoting John Grisham novels and stuff, not actual facts. My question to you: Your newer version —
Mr. Dinnick: Yes.
Senator Arnold: — that is more agnostic —
Mr. Dinnick: Yes.
Senator Arnold: — is it better?
Does it create the same kinds of issues in a social media context?
Mr. Dinnick: Again, our learnings have been so valuable. It is better, but here is what we learned: People don’t bother checking facts. There was too much friction in our first model, so we are going to, hopefully, get to no clicks or one click. Our credo is we’re making fact-checking as simple as spell-checking, so you will be able to quickly see what claims might be in question. Then you can click on it, and there is a tool tip to take you in.
The last model we had, the engineer — again, thank you to our military because these are former Joint Task Force 2 AI engineers, and it’s absolutely brilliant what they built. You click on it, and then you can ask the audit trail a question. You can ask it to take you to the source. What are the contrary sources? How long has it been around? The technology is there. It is absolutely doable.
To answer your question, senator, the real challenge has been making it more usable for Canadians and for it to be simple and frictionless, so that fact-checking does not seem like a heavy lift.
I would like to say one thing as a journalist — and Senator Simons, you may agree with this — I think one of the big problems with journalism over the last few years is that we have held behind us and closed the door on how hard we worked all day to check facts. When I worked at ABC News, you could have the fattest contract and best job, and your buddy in New York could make sure that you were okay, but if you got one fact wrong, you were sitting out for the rest of the contract, and you were not coming back.
Fact-checking at those organizations — I do not think that Canadians, especially younger ones, understand the importance of facts and the heavy work that journalists did.
Back to your point, I think that we have to get Canadians, who maybe don’t get it, to understand that facts play a really big role in how we all operate, and this tool makes it easy for Canadians to get there.
Senator Cormier: My question is: How can this model be applied to culture? Could you give us some examples of how it would apply to, generally, culture, novels or movies or whatever?
Mr. Dinnick: Senator, I’m not sure that I can answer that because it is in its infancy.
I’m not sure if this answers your question, but it starts to get there. We built it in any language, and because we were concerned that when somebody just looks at AI and wants to find out some cultural thing in Nova Scotia, it is not going to be easily found in that noise of AI, which we also know is being intentionally polluted by bad actors like Russia and China. That is another issue.
What we’re trying to do is create databases that are out there but are uniquely Canadian and easily searchable.
I’m not sure we’re at the stage where we’re talking about specific different cultures on literature, film, et cetera, but our goal is definitely to make it accessible for uniquely Canadian data because, again, if you look at something like health care and vaccine information, boy, we do not want to go on Make America Healthy Again and have people get those answers about vaccines and measles. We want them to be able to access Canadian research. Again, I don’t think we’re there yet, but the idea is that eventually we would like to be.
Senator Cormier: Talking about data, what road would you take to get the data?
Mr. Dinnick: There are data sets. What we have learned is that the best thing for this is AI governance that is open and transparent. That is what we are.
One of our models now helps climate NGOs create sets of data that Canadians can then look at and find out more about climate issues very quickly with this tool.
Those climate NGOs will have a board of governance over that data that is transparent to everyone so they can see what is there and what is not there.
Senator Cormier: Thank you.
Senator Mohamed: I am really captured by this idea that you could put in a URL and up comes basically a source, a way to say whether or not this is fact. In any model, because you are using a model that has AI as a part of it, there is often a bias of some sort. How do you account for a bias? How do you make sure that we’re not doubling down on a bias that may exist in that reporting, which then comes through your platform? That is my first question. I will let you answer that. I’m troubled by that doubling down and the implications.
Mr. Dinnick: Again, we’re not training a large language model. That is not what we’re doing. Think of it as 500 fact-checkers working very fast when you ask a question. That is what it is like. It is going out, and it is searching everything.
To answer your question specifically, on our advisory committee, we have some great Canadians — people like Kathy English, public editor of the Toronto Star, Wendy Freeman, former president of CTV News, and Kevin Newman — working on a weighting system and with engineers to look at weightings — I’m not an engineer — to look at what bias is and the judgment that then goes to pull out the facts based on it. Then, it shows you a list of the other sources that have mentioned those facts.
I will send you all of the tools so that you can look at it. It is a weighted tool. Again, it is open and transparent so that you can see it. It isn’t just “go find this fact.”
Senator Mohamed: I have been on the tool. As you have been talking, I was on the tool, and I know Wendy Freeman. She is a very bright, strong, opinionated person who does the right stuff.
Mr. Dinnick: Yes, very careful.
Senator Mohamed: Having said that, I just imagine you put it in there, and then speed becomes the name of the game. I’m not asking you to choose, but speed versus trust is often the tension. I’m having a hard time thinking about it from a young person’s point of view. Is this a part of the challenge? It is not that I do not think that young people do not investigate or interrogate, but they are moving fast, fast, fast. To get them to go to a place like this, how do we make that a behaviour where people are getting the information so fast and trust becomes almost irrelevant? Explain that, please.
Mr. Dinnick: Yes.
Senator Mohamed: I know that you are not here selling this, but you’re selling the idea of verification.
Mr. Dinnick: First, let me say if someone comes up and makes this faster than we are, fantastic. I would love to pack up, go home and be happy about it.
No. It is what I said to Senator Arnold. Our next iteration is not dropping in a URL. It is a tool tip. It is a web extension. It is in your workflow. It is one click or no click, so the idea is that it is right in your feed, right there so you can see it.
As examples, there are Bluesky and Gander, which are new social media platforms. We’re working with them to be their verification. Before you load something on their social media, you’ve clicked it. Our little button is there, and it says that, before I load this, I’ve made sure that I have double-checked that it is true.
To your question, how do we make it faster and better? A little bit of money. We’re a volunteer organization, but we’ve learned enough of the model that we’re going to grow it properly and that it is frictionless. That is the lesson. We’re here to share what we’ve learned because they have been expensive and important lessons, and we want Canadians to benefit from that. It has to be frictionless, and speed and accuracy can come with money and GPUs, et cetera.
Senator Mohamed: I have many other questions, like my colleague, but I will refrain.
Senator Miville-Dechêne: I was also a journalist for 25 years at Radio-Canada. I am a bit puzzled. I’m intrigued and puzzled by your model because it is not a fact model; it is a source model. Obviously, I left journalism a few years ago. But the fact checking at Radio-Canada was not what it was at ABC. We made mistakes, and then we would correct them. There is a whole system of correction. So sources can be bad sources. I’m not speaking about Radio-Canada because, in general, it is quite good, but there are mistakes.
So if you send somebody on your system with a source that is not reputable, how do they know that the source is not a great source? It is not because you have a source that you have a fact.
Mr. Dinnick: Yes.
Senator Miville-Dechêne: Where do you go from the source?
Mr. Dinnick: I can answer that in two ways. The first one is, let me reiterate, this is a tool. Human in the loop, absolutely. Our whole credo, our whole mission, is that we’re just getting that information to you faster. What you do with that information, how you analyze it, is up to you.
To answer your —
Senator Miville-Dechêne: So if it is a newspaper that you know there is a bias there, you will take that into account, but the machine doesn’t say that?
Mr. Dinnick: Yes, it does.
Senator Miville-Dechêne: It does?
Mr. Dinnick: It does. But the bias is not important to us; it is the fact. Because you can have three facts and one untruth or lie in there. Our job is to unpack that one untruth and to point out the other facts. So absolutely it is. Bias, judgment and truth, that is for Canadians. We’re helping you find the facts around any piece of information and the source of it. Again, yes, as a journalist, we sit in the newsroom at the end of the day, and we sometimes argue whether those facts even matter to that story, agreed.
This tool is just like a fact-checker. It helps fact-checkers. It is helping people get to the source of that information and those facts. Does that make sense?
Senator Miville-Dechêne: Maybe I’m tired.
Mr. Dinnick: No. This is why we don’t have venture capital money yet.
Let me give you a great example. We took it to a minister. We had 45 minutes with the minister. We thought, great, okay. This is great. I will leave the person nameless. They spent the whole time looking at the CBC’s same story and the Rebel News story, and they were so angry because it had the same results on facts. How can the CBC have the same results as Rebel News on facts? Well, Rebel News uses some facts, CBC uses some facts, and they had very different angles on the story. That is okay. Our tool only sourced the facts and pulled the facts out and showed them.
At the time, our tool was rating — which we don’t want to rate for a whole bunch of reasons that we’ve learned — but our tool showed — this minister could not get their mind around it, sucked up the 45 minutes — that those facts were the point of the tool. In all that noise, in all the wrapping and all the other stuff, here are the facts. Against these claims, here is the source of those facts. That’s where it went. That’s all the tool does.
One thing we are also trailing is with journalists in small and regional newsrooms who are under such pressure, right? Talk about CBC; they have fact-checkers and stuff. These are one- and two-person newsrooms, and they serve the most important role in journalism in this country, where people are dealing with crime, public health, schools and education, things that matter to them most. Their local news organization is so under pressure and inundated with misinformation, so this tool that we’re trialling is going internally into those little newsrooms where it is simply helping journalists check the facts. When they quickly see something, they can put it in. Okay. I know where this came from. That is how it works.
I think that some people get lost on this because it seems as if it’s some magic box; it is not. It is just a tool to help check facts. Once you get to the facts, you can decide whether you like those facts, you agree with them, whether you think they are from a reputable institution, all those things. There is just so much noise that we need to cut out all of that noise and get to the signal. Thank you.
Senator Simons: I’m trying to get it to work. It keeps telling me it’s a processing error. I’m trying to put my own stories in.
Mr. Dinnick: I’m very sorry. Yes, we are a volunteer organization, and Google has been generous giving us — I mean, we have no money, so we are just all volunteers, basically throwing it in, so sometimes it does not work. We are building the new one, which will be replacing that one.
Senator Simons: It wasn’t a criticism; I was just disappointed because I wanted to see what it said about my reporting.
Mr. Dinnick: No, no. I know, it’s frustrating. Yes, thank you.
Senator Simons: Yesterday, I raised the issue in committee about a news story that reported that there were a cluster of fake accounts on YouTube creating AI-generated videos in support of Alberta separatism, and not just in support of Alberta separatism but in support of Alberta becoming a part of the United States.
Mr. Dinnick: Yes, I saw that.
Senator Simons: They were all full of wrong facts, that 65% of Albertans favour separatism. There is no poll that says that. That is fine. You can pull out facts like that and say that those are not facts. But the grievances of political grumblers are a lot harder. If somebody says Justin Trudeau tried to destroy Alberta’s oil economy. I’m a human being, so I can say, really? Because also he built TMX at a great loss to his political capital in British Columbia. But a lot of Albertans would say to me that is not the point, because he put in place emission caps that were devastating to the industry or that he had environmental assessments.
I spent my whole life compiling news stories that were, on the one hand, on the other hand. Oftentimes, there are facts that are in dispute, and I don’t know that a bot can give me assurance that the facts it finds are not — as Senator Mohamed said — going to be biased in how it goes to look for things.
Mr. Dinnick: Again, what our tool does is it doesn’t tell you what is right or what is wrong. It doesn’t tell you that you are right or wrong. It doesn’t tell you what your bias or judgment is. Simply, when there is a claim, it points to the source of that claim and whether there are facts behind it or not. Our tool will say that is not true, but here is the latest Ipsos poll on how Albertans feel about separatism.
I agree with you. There is no way our little organization is going to be able to appease grievances like that.
Senator Simons: Yes.
Mr. Dinnick: Fifteen per cent — I’m guessing on either side, the left and the right — they are probably in their worlds, and there is nothing that you can do about it. But there is a malleable middle that increasingly, because of AI which is creating huge amounts of content, is making it very difficult for Canadians to source real information. That is what we’re trying to solve. We don’t want conversations that end in, “That’s not what I heard,” or, “Are you sure? No.”
“I was able to click through, and here is the source of information; this is what I think.”
Where we’re really scared is you have to look to the United States and the most extreme arguments between MAGA and the progressives. They no longer argue about efficacy or policy; they argue whether something is true or not.
Senator Simons: Here’s a great example. The U. S. Centers for Disease Control used to be considered the gold standard for health information; people around the world would look to the CDC as an exemplar of scientific truth. Everything at the CDC has now been corrupted; they are sharing out health disinformation, and yet they still have the imprimatur of a respected organization.
It’s one thing for me to say that their claims about vaccines or about Tylenol causing autism are just not true, but a bot will go to the CDC website and say, “Well, here is a legitimate organization, and this is the fact it provides.” If the wellspring has been poisoned, how does the bot know what has become a reliable source of information? The Washington Post of Woodward and Bernstein is not The Washington Post of Jeff Bezos.
Mr. Dinnick: For the news one, we have data sets that we know are verified information, and then when we have those 500 agentic fact-checkers that go through, in the CDC it will say, “The CDC has been under criticism for this, so you can click through, and here are the Canadian statistics, et cetera, if you want to dig in.”
But back to Senator Mohamed’s point, we have to make that frictionless, and that is a challenge. Our little organization can’t be here to start talking about what people think of the CDC, et cetera. We just want to be able to give Canadians quick access to facts in an increasingly noisy environment.
Senator Simons: I don’t know if you would also consider, and maybe this is outside of your mandate, that critical thinking skills seem to be in short supply these days. People who have grown up in a world where if something was written down as text, it was perceived as being believable are having a very hard time adjusting to the new reality that everyone is trying to trick them.
My phone today sent me a message that said, “Dad, I’ve lost my wallet.” I get them all the time now; these people pretending to be my child. First of all, I’m nobody’s dad; second of all, even the ones that say “Mom,” my daughter never calls me mom with an “o,” so it’s easy for me to know these are fake. But for a lot of people, especially if they are slightly estranged from their children, they get these messages, and they seem real. I would think that anybody with a brain God gave a goose would know that this is fake, but unfortunately, it’s evident that a lot of people don’t have the brain God gave a goose to be able to tell what is fake from what is real. If we outsource our critical thinking to an AI, doesn’t that just make the problem worse?
Mr. Dinnick: We’re not outsourcing it to AI; we’re giving people a tool to go to the source of the facts. We’re doing the exact opposite; that’s our goal.
The Chair: Would you like to make a closing summation?
Mr. Dinnick: The more eyes on this, the better. Many businesses have supported us. Many people have helped keep the lights on, many retired engineers, military veterans, et cetera. We’re open to criticism and questions. My email is there. It’s wilf@getfact.ca, and we would be more than happy to entertain any questions and ideas beyond this session.
Senator Miville-Dechêne: Is it available in French?
Mr. Dinnick: It’s available in French. It’s available in any language.
Senator Miville-Dechêne: You just click.
Mr. Dinnick: You can click, yes. Of course.
The Chair: On that note, I’d just like to thank you for the discussion that we’ve had, Mr. Dinnick. As someone who is probably the furthest away from understanding technology, I found it really motivating. The motivation I get from what we talked about tonight is the importance of searching for more, and you did a fantastic job of positioning it without telling us, “Well, you know this,” or, “I know more.” You were very good at putting facts on the table so all of us could look at those facts and make our own judgment. Thank you so much.
Mr. Dinnick: Thank you so much.
The Chair: Let’s give a hand to Mr. Dinnick.
Colleagues, before we wrap up the meeting this evening, there’s a matter of housekeeping that we must take care of first. I would like to move the following motion:
Is it agreed that the Honourable Senator Wells (Newfoundland and Labrador), subject to his addition to the membership of the Standing Senate Committee on Transport and Communications, be elected chair of the committee effective April 23, 2026?
Senator Miville-Dechêne: Yes.
Senator Simons: We’re very happy to welcome Senator Wells to the chair, but we’re very sad to say goodbye to you.
The Chair: Thank you so much.
Senator Dasko: I’m ready to vote.
The Chair: Is it agreed?
Hon. Senators: Agreed.
The Chair: Thank you very much.
I got myself emotionally involved here. Hold on.
I would just like to take a few moments of your time to offer my sincere gratitude to all of the senators on this committee, but also to my fellow steering members, who have been exceptional in their support.
Thank you as well to our dedicated clerk, Andrea Mugny, who has been a rock star in helping us organize this committee. Thank you to our wonderful analysts Geneviève Gosselin and Marion Ménard for their extraordinary work, but also for their patience in dealing with all of us, and maybe especially me.
Finally, thank you to all the staff in the room with us and those behind the scenes — the interpreters, the stenographers, as well as the broadcasting group — for their professionalism and ensuring the good work we do here is accessible to all Canadians. Let’s give a hand to the folks in the background.
Finally, senators, thank you for putting your trust in me and for allowing me to sit in this very chair. I look forward to the excellent work that will continue to come out of this committee in the near future.
I sat on a lot of committees over my time here. I’ve presided as the Chair of Finance and the Deputy Chair of Internal Economy. I have been involved with the internal operation of the Senate, and this committee, for me, has been an eyeopener, and I would say an eyeopener for a variety of reasons.
We have a great team. You have a great team of people here and the ability to communicate.
[Translation]
The way we share and support each other is incredible.
[English]
I’m not trying to build anything up or exaggerate. This has been a real experience of a lifetime for me, and I thank you all so much.
[Translation]
Thank you all for your support.
[English]
Thank you so much. We’ll wrap it up and say, “Good night; stop crying; thank you.”
(The committee adjourned.)