Canada-China Relations Committee on March 22nd, 2021

[Stéphane Bergeron, Bloc, Montarville, QC]

Thank you, Mr. Chair.

I would like to thank the witnesses for being here, for giving us their time and for enlightening us with their comments. I am clearly grateful to Mr. Parsons, but particularly to Mr. Sarts, given the very late hour. I will address Mr. Sarts first.

On February 22 of this year, the European Parliament held a meeting of the Special Committee on Foreign Interference in all Democratic Processes in the European Union, including Disinformation. You participated in that meeting and discussed how disinformation works. When we look at your findings on the 2019 European parliamentary elections and the 2020 American election, we have every reason to be very concerned about what's next.

You will be able to enlighten me on this, but one might think that all NATO countries are facing this same sort of foreign intervention in elections. But it is a bit surprising that, according to Greg Austin, who leads the Cyber, Space and Future Conflict Programme at the International Institute for Strategic Studies, China's cyber defence capabilities are far below those of the major western powers, including Canada. For example, Canada ranks 9th of 155 countries assessed, while China ranks 27th.

Aside from the fact that they rely heavily on private companies, why would western powers allow themselves to be put in this vulnerable position without reacting?

[Janis Sarts]

Thank you for the question.

We've been looking.... Most of these social media companies that we use for everyday life have become the agora for democratic process. Most of the elections actually play out in these platforms. We've detected that most—basically all—of those platforms are manipulable by robotic networks to put the messages and to game the algorithms—including during the election processes—to advance particular interests, including of hostile actors.

We've been measuring, every year, how well the platforms do in taking out these robotic networks from platforms, and the results have been very disappointing. Back in 2019, when there were European Parliament elections, we bought 55,000 different actions through robotic accounts on social media—of course, neutral effects—for 300 euros. During the EU parliamentary elections, 90% of that got delivered.

We repeated the same experiment during the U.S. presidential election, once again in a neutral manner. We were able to buy likes, shares, views, custom-made comments and all of that, but this time 300,000 for $300. About 70% of that got through. Basically, there was an option for outside actors to influence the discourse.

Most of the companies were incapable of eradicating that process. If I had to measure the companies, typically Twitter is the best at it. Facebook is less so. Last year, we measured TikTok for the first time. TikTok is basically defenceless. You can do any gaming of that system that you wish. Of course, the more potential electors there are out there, the more malign things can be happening.

Clearly, that goes back to Mr. Parsons's point that there is no way to oversee what the social media companies are doing. They're declaring great success, but when we turn to the vendors of these manipulations, it's cheap, available and effective. We have to have oversight to make sure that it is neither simple nor easy.

Thank you.

[Stéphane Bergeron, Bloc, Montarville, QC]

Thank you for that very detailed answer. But I would like to come back to the question I asked you, Mr. Sarts.

China is considered to have a cyber defence capability that is far inferior to that of most western powers. I suppose one could say exactly the same thing about Russia. In this case, why are western powers content to put themselves in a vulnerable situation? Why don't they use their superior capabilities to establish a system of deterrence, to discourage powers like China or Russia from engaging in these sorts of practices, or face strong retaliation?

[The Chair, Liberal, Geoff Regan]

You can give a 10-second answer, Mr. Sarts.

[Janis Sarts]

I think the U.S. just did it in 2020 by discouraging Russia, as we've seen from open-source reports, so it is possible to do if you have leverage, resources and the political will.

[The Chair, Liberal, Geoff Regan]

Thank you very much.

We'll go to Mr. Harris, for six minutes, please.

[Jack Harris, NDP, St. John's East, NL]

Thank you, Chair.

Thank you, both, for your interesting presentations.

Mr. Sarts, you just described, in the EU elections, the ability to buy data and then send messages to that data. Can you give us an example of where that would come from? Would this be private operators providing this information for sale to make money, or is this part of some other effort that might be done by a state actor?

[Janis Sarts]

Both. There are private actors that do it for sale. There are what we believe to be—by continuously following these networks—state actors. In some cases, there are hybrid networks where most of the time they would do business-related things for gaming marketing or influencer posts. About 10% to 15% of the time they would do the political impact. There are a variety of players in the field.

[Jack Harris, NDP, St. John's East, NL]

So you can influence an election by getting information about who likes what and then sending targeted messages. Is that exactly how it works?

[Janis Sarts]

That is one of the ways. Of course, you can use these automated robotic networks, which seem like humans, to create reactions en masse in the online media, gaming the algorithm—for instance, making some posts much more visible than others, making specific comments at the political leaders' social media presence discouraging, or making false appearances. There are many ways one can use that infrastructure for effect.

[Jack Harris, NDP, St. John's East, NL]

Is there any technical or technological defence to any of it? How do you deal with it?

I was reading a comment about when you testified at the United States Senate Select Committee on Intelligence in 2017, suggesting that public awareness campaigns could counter and influence operations that target populations.

Of course, that was 2017, and we've seen an awful lot happen since then. Are you optimistic? Was that an optimistic statement, that public awareness campaigns could be a defence against this, or are there other techniques that might be possible now?

[Janis Sarts]

First, public awareness campaigns are important, because if we don't do them, then we are even more vulnerable. But on the particular issue, obviously part of the defence is within the social media companies.

In our assessment, they are not doing [Technical difficulty—Editor] public discourses that are happening on these platforms, and therefore some kind of regulatory framework on our side would be necessary.

[Jack Harris, NDP, St. John's East, NL]

How would they stop them, Mr. Sarts?

[Janis Sarts]

Actually, it is very simple. You can create algorithms that see these things for what they are. For instance, if we buy this robotic accounts effect, you can see that account. We report that account to Facebook or Twitter.

Most of the time, their algorithm doesn't detect it. It's just a matter of upgrading their algorithms and being better at their jobs. That is not the case, at this point.

[Jack Harris, NDP, St. John's East, NL]

You're saying that these companies could actually police that activity, if they were motivated to do so, shall we say.

[Janis Sarts]

Yes, if there were a better business case for them, I believe they could.

[Jack Harris, NDP, St. John's East, NL]

You also talked about another tool that's used, called data scraping. I guess that's what we're talking about here, using artificial intelligence, algorithm and data scraping to influence behaviour.

You referred to some event actually influencing the behaviour of military operatives by obtaining information. Could you explain how that works and what the dangers can be in something like that?

[Janis Sarts]

Yes. Two years ago, we tried an experiment trying the hypothesis that open-source data can be used to influence human behaviour. We did an experiment together with the Latvian armed forces, during a military exercise, where we scraped the open-source data for the soldiers. Based on that data, we tried to impact their behaviour during the military exercise.

We succeeded in making soldiers disobey orders, making them leave the positions they were supposed to defend, just based on the data that was available. This basically underlined the future risk of big data that is available. If it's used in a malign way, it can not only bring, as it does currently, the marketing product; it can also shift beliefs and behaviours. In the wrong hands, it is a very dangerous tool.

In that respect, I would highlight the future risks of 5G. It's not only about the infrastructure; it's also about the data that is going to flow in that system. It is incredibly valuable, if you look from the hostile actor's perspective, to get access to that kind of societal data, because with certain AI capacity you could actually sway the behaviours of the other societies.

[Jack Harris, NDP, St. John's East, NL]

That's fascinating. I was in Riga at one time with the NATO Parliamentary Association; I think I've been at your centre. There was a lot of concern in those days about the disinformation campaigns of the Russians, trying to undermine the interest in democracy in the Baltic states.

Is that still going on?

[Janis Sarts]

Yes, it is of course still going on, but as we see, from a Russian perspective they have moved their eyes more towards other, bigger players. They're spending most of the resources there. Of course, the Baltics are still a target, but not the main target.

[The Chair, Liberal, Geoff Regan]

Thank you very much.

Thank you, Mr. Harris.

As we only have a few minutes remaining, what I propose to do is try to distribute it equitably, with three minutes to Mr. Williamson and three minutes to Ms. Zann.

Then, Mr. Bergeron will have one minute and 30 seconds.

Finally, Mr. Harris will also have one and a half minutes.

Mr. Williamson, you have three minutes.

[John Williamson, Conservative, New Brunswick Southwest, NB]

Thank you.

Mr. Parsons, I see you co-authored a report a couple of years ago, “The Predator in Your Pocket”. It touched on DNA sourcing and DNA transfers. You referenced 23andMe.

We now fast-forward a year and a half. The world's largest biotech firm, BGI, was given approval by Health Canada to offer COVID-19 testing in this country.

Do you have concerns about the transfer of the health information and DNA of Canadians to mainland China for use in that country?

[Christopher Parsons]

I haven't looked specifically into that case. Certainly what we have seen in our research and analysis of other parties' research is that DNA information is incredibly sensitive.

There is a concern, of course, when any highly sensitive data is moving to any country outside of Canada, and that would be inclusive of China.

[John Williamson, Conservative, New Brunswick Southwest, NB]

Can you talk about the information on 23andMe and expand on why you felt the need to highlight that in the paper and why that was a risk?

The difference here is that this is on an individual basis where that information is being collected and possibly shared, versus the BGI, which is possibly collecting it on hundreds of thousands of Canadians. Could you just address why for 23andMe and that transfer it's critical to have a handle on it?

[Christopher Parsons]

Yes, absolutely.

The first concern is how that data will be used to affect the individuals themselves. It's collected under a certain set of terms. Will that set of terms be applied on an ongoing basis?

The secondary concern is that while it does reveal information about the individual, it also reveals information about their entire family, including members who may not yet be born. The ability to use genetic information to drive information about your current or forthcoming next of kin is something that we really can't predict. Genetic technology is just exploding. The actual uses are pretty broad.