Information & Ethics Committee on Nov. 27th, 2023

[Dr. Anatoliy Gruzd]

We can clearly see confusion in the future between what's authentic and what is not. Right now, we're not at that stage. In fact, in a number of studies that I've seen, when they ask human participants whether or not they recognize some of the deepfakes and other artificial intelligence-created artifacts, the humans still can recognize these.

We also see a time in the near future when it will be much harder to differentiate between generative AI content and authentic content or works. I think that's where the next battle is. We see some platforms exploring options requiring their content creators to first disclose whether any generative AI tools were used to produce that content. That's an important step.

The next step is perhaps to do some kind of digital certification or watermark on the content, so that we actually know how it was created. There is nothing wrong with generative AI, but if the content it may create is used for malicious purposes, that's of course problematic.

[René Villemure, Bloc, Trois-Rivières, QC]

Do you think that the use of artificial intelligence on social media platforms will help make the concept of truth vaguer and, consequently, make it difficult for people to trust their interactions with the platforms in question?

[Dr. Anatoliy Gruzd]

That is exactly right. It's about the trust between the content and the particular topics, and sometimes it's just enough to create confusion. If you have a state actor that may not be able to convince us here in Canada of certain narratives, it's maybe just enough to cause some confusion.

In my research, I focus a lot on Kremlin propaganda, and we see that strategy being used a lot.

[René Villemure, Bloc, Trois-Rivières, QC]

Have you assessed the impact of bringing QStar into the equation with social media? I understand it's early days, but do you know anything about that, or do you have any cautionary notes for us in that regard?

[Dr. Anatoliy Gruzd]

Can you expand and try to give me a bit of context?

[René Villemure, Bloc, Trois-Rivières, QC]

This is OpenAI's most recent project, called QStar, which, again according to OpenAI, would make the use of the technology dangerous.

[Dr. Anatoliy Gruzd]

Any unregulated use of AI, perhaps, can be misused in the future. I think right now that we are in this Wild West territory, where a lot of wrong steps will be made, companies will try to innovate, and bad actors will misuse the technology.

I am glad that this committee and maybe other committees as well are trying to look into this issue. It is right now like the Wild West. It is concerning, but as any tool, it can be used for educational as well as nefarious purposes.

[René Villemure, Bloc, Trois-Rivières, QC]

A hammer can hit a nail, but it can also kill, for sure.

There's a lot of talk about TikTok here today, but there are other social media from foreign countries that we are less familiar with. There's talk about social media from Russia, but there's also often talk about social media from India, Pakistan and Iran.

What are some of the other social media that maybe we should be looking at?

[Dr. Anatoliy Gruzd]

The Social Media Lab produces a report, every two years, entitled “The State of Social Media in Canada”, and we ask Canadians what platforms they use. Certainly most of the top nine platforms would be North American and U.S.-based, except TikTok is the fastest-growing platform. Around one-third of Canadians use it.

Another platform, which hasn't reached a 10% adoption rate in Canada, is Telegram. It is being adopted quite widely around the world. In fact, the rate in terms of the service rating I mentioned is, interestingly, B, so it's quite high versus E for the rest of the platforms. While it's privacy friendly or conscious of users, it's full of Kremlin propaganda discourse, so you pick your poison, unfortunately.

I would definitely keep an eye on Telegram and a lot of messaging types of apps.

I had a question earlier about WeChat and such. Those are really hard to study. Anything this committee can do to help mandate platforms to share insights on those platforms and their public groups, where most of that originates or is propagated, would be very helpful going forward.

[The Chair, Conservative, John Brassard]

Thank you, Dr. Gruzd.

Thank you, Mr. Villemure.

Mr. Green, you have six minutes. Go ahead.

[Matthew Green, NDP, Hamilton Centre, ON]

Thank you very much.

I believe that in your opening remarks you mentioned the efficacy or the rationale behind targeting one platform. I think you've just dealt with some of that in response Mr. Villemure's remarks around looking at these as tools. I've heard you say multiple times that Kremlin propaganda is present.

I want to be clear. Across all regions—east, west, north, south—would you not agree that all countries, including western European countries, utilize propaganda, be they state actors or private interests, and use these platforms for nefarious purposes?

[Dr. Anatoliy Gruzd]

In fact, we see that happening across the board. The case I mentioned earlier that was found by Meta was quite a sophisticated information operation run through multiple platforms, not just a single platform. In fact, that content included an element whereby fake websites were created, and of course, with generative AI it is quite easy to create something like that. Essentially, you create fake content that looks like a news organization and then use social media to get eyeballs to that content, or you use targeted advertisements to get eyeballs.

I think state actors will use any tools available, and any social media platforms that are popular in Canada would be a target.

[Matthew Green, NDP, Hamilton Centre, ON]

It would be your assertion, of course, and logic would follow that this study should be dealing with all platforms in all regions, including all actors, those deemed both Western friendly or more authoritarian regimes that might be around the world. Would it be a safe assumption that you would support a broad look at all platforms?

[Dr. Anatoliy Gruzd]

I agree that we need to look at this holistically, including not just one, single platform, unless your future witnesses clearly indicate or provide evidence as to why that particular platform is a special case.

[Matthew Green, NDP, Hamilton Centre, ON]

I'll ask you the question.

In your opinion, does spending all of our time and focus on one platform because it happens to be from a certain part of the world hit the mark, or does that miss the mark in terms of fully understanding the risks of algorithmic interference, profiling and direction?

[Dr. Anatoliy Gruzd]

I think it may put emphasis on just one platform, making it sound as though other platforms are safe when, in fact, they are engaged in similar data-harvesting practices and similar data misuse, or may be used by state actors. Definitely expanding the scope and looking at the strategies used for information operation would be a much more effective angle, unless—

[Matthew Green, NDP, Hamilton Centre, ON]

We've spoken here quite a bit around data capitalism, surveillance capitalism and algorithmic capitalism. I want to go back to these private companies, these western ones, these American ones: Meta, Instagram and Twitter, obviously, with Elon Musk and X. Is it not true that a lot of our information profiles—click information, geographic information, our tendencies, our preferences—are collected by private companies and then sold to third parties as, really, what the product is: the commodification of the user and not the actual platform? Is that not correct?

[Dr. Anatoliy Gruzd]

It's been happening, and some of the measures put in place are sometimes counterintuitive or counterproductive—even the simple example of going to a website and having it ask you if you accept cookies. Well, what is my choice if I want to visit that website? There is also the example I mentioned earlier about installing the TikTok app without having an account; somebody's trying to track.... I think it's a pervasive practice across the board and across the industry.

[Matthew Green, NDP, Hamilton Centre, ON]

I want to go back to that point to be clear, though.

Regardless of where these companies originate from, when they sell to third parties there's still a likelihood—if not a high probability—that the same sensitive information that's being harvested by TikTok via ByteDance and Chinese state-owned companies could still end up in the hands of oppressive regimes to get information on their citizens—to get information on the diaspora, on dissidents, on people who might not share the same opinions as these authoritarian regimes. Is that not correct, whether it's Instagram, Facebook or X?

[Dr. Anatoliy Gruzd]

It's not a secret. There are a number of different data-harvesting companies that would be sharing this type of information. There is also the whole dark web that would be collecting and sharing the information that was leaked or hacked through different repositories. Unfortunately, that's the environment we're living in. That has to be taken into consideration.

[Matthew Green, NDP, Hamilton Centre, ON]

Is it also not true that these other actors, these other platforms, have also, in the past, provided backdoor access to messages and to information, whether it's through quasi-legal or grey-area access as it relates to perhaps not having warrants and that type of thing?

[Dr. Anatoliy Gruzd]

Yes. As I mentioned in my opening remarks, I think it's important to recognize different types of interference. One can imagine having a platform where a state actor has direct access—like VKontakte, which ran from Russia and was, in fact, banned in Ukraine due to this threat that was determined, that it was actually run by the state—versus the risk associated with the general data misuse practices, whether by platforms or by third parties. I know that, earlier, you and other committees were referring to Cambridge Analytica. Sometimes a third party would get access to platforms and data through legal means, such as through their developers' applications. That's another form of interference.

[The Chair, Conservative, John Brassard]

Thank you, Dr. Gruzd and Mr. Green.

Dr. Gruzd, before we go to Monsieur Gourde for the second round of questioning.... You said something earlier about state actors using social media for polarization. I'm wondering if you could clarify how they would go about this. What methods would they use? Would they use the data to target individuals or groups who are sympathetic to a cause, for example? How would that occur?

[Dr. Anatoliy Gruzd]

Usually we're not talking about an individual like me or a colleague being targeted. It is in fact the sympathetic groups that state actors would be targeting. What happens is that they would look at political partisan views that may be aligned with their objectives. For example, when we're talking about pro-Kremlin content, that usually resonates very well with far right Conservative groups, especially in the U.S. An example would be Tucker Carlson, who was formerly on Fox News. He was channelling pro-Kremlin claims because some of those claims are aligned well with far right ideology.

Earlier I mentioned that while the state actors may create bot networks—and they did—those accounts don't have credibility; they probably will not impact individuals or groups. It's the goal of that campaign to impact somebody in power—either an influencer on TikTok or a politician running for office—and use a microphone to usefully share the same narratives and such.

[The Chair, Conservative, John Brassard]

Thank you, Dr. Gruzd.

Mr. Gourde, you have five minutes.