Information & Ethics Committee on June 12th, 2018

[Jeff Silvester]

We make sure that all of the ads we run are attached directly to the candidate and to the campaign. If they're publicly espousing something that other people might take issue with, they can take it up particularly with that campaign. To your specific question as to how I would know whether it's misinformation versus just being negative, my response is that sometimes negative advertising is simply pointing out the truth about that person.

I can think of specific advertising. It was an ad saying look at what this person has done. Once you clicked on the ad, it brought you to a website. We didn't control the website, though it had factual information on it about some of the things that the particular candidate had done in the past that people might find unsavoury. That was negative advertising, but it also appeared to be correct.

[Jacques Gourde, Conservative, Lévis—Lotbinière, QC]

Indirectly, advertising could lead us to other sites containing disinformation about an election campaign. This might not be your fault, but these stratagems and means provided by your company probably allowed people to spread disinformation during election campaigns, despite your good intentions.

[Jeff Silvester]

I don't believe that we've ever done that. It is theoretically possible, except that the different advertising networks also check the landing pages, so if you're directing something to a landing page, I guess the company could ask us to advertise. We could start and, pointing to web paging at the time we began, everything looked fine. Then they went back and changed that web page afterwards, and we didn't know they'd changed it. Yes, it is theoretically possible. We have not seen that happen in any of the work that we've done, but I suppose it would be possible.

[The Vice-Chair, Liberal, Nathaniel Erskine-Smith]

Thank you very much, Mr. Gourde.

If nobody minds, I'll take the next five minutes from here. I want to start with a bit of a summary of what I understand, just so I've got it all clear in my mind.

Mr. Massingham testified previously that he was aware that if any additional expenditures had gone through the Vote Leave account, they would have been over their limit. Enter 22 year-old Darren Grimes, who starts the BeLeave campaign just weeks before the referendum with no resources and no data. They get in touch with you somehow. We don't know exactly how they became aware of you, but they did presumably through Vote Leave. Mr. Massingham has added to a Google drive. He testified before us that he had no idea what was on the drive, even though this was one of the largest campaigns you've ever run. Mr. Grimes makes purchase orders in large sums mere days in-between, or close together to his receiving the same large amounts from Vote Leave that weren't the exact right numbers of the purchase orders he made. We learn that 1,400 ads were placed by AIQ between the two campaigns. Almost £700,000 was spent in total. We're directed to Slack messages, and on those Slack messages we see statements from Mr. Massingham to the effect of “your soft stuff will play better here versus the hard stuff”. Presumably that's Mr. Massingham speaking to the Vote Leave materials and to differentiate between the two. But nowhere in those Slack messages is there an indication of how many ads to run, which content to use exactly for those large number of ads, or how much money to put behind the ads. That was all done verbally from what we know.

Oh, I forgot to add, but there was absolutely no collaboration or coordination whatsoever. Is that right?

[Jeff Silvester]

To your point, Mr. Massingham said he didn't know the entire contents of the Vote Leave drive, so when BeLeave sent us a link, they sent us a link to images and stuff they'd used previously. I just wanted to clarify that one point.

[The Vice-Chair, Liberal, Nathaniel Erskine-Smith]

But that was the only clarification you wanted to make—

[Jeff Silvester]

Well, no. The information that was given to us to do the advertising was certainly initially communicated by Mr. Grimes verbally. The actual direction on the ads did come from Slack. We did have a reporting tool up at the time that allowed BeLeave to see how their ads were doing, and how things were going, but we don't have that anymore. That was shut down at the end. We provided a full report to them at the end of it to show exactly where everything went, and they were happy with what we provided.

[The Vice-Chair, Liberal, Nathaniel Erskine-Smith]

Right.

So moving to a different country, we've got the Ripon tool that was developed. You've indicated to the U.K. committee that, “I want to be completely clear, the Ripon tool, the political customer relationship management tool, did have some of SCL's personality scoring in it, but we do not know the datasets they used in order to arrive at that.” And you've said similar things today.

You are under oath, so looking back at the tool that you created, the psychographic scoring as it were, and knowing what you know now about how SCL obtained information through Cambridge Analytica improperly through Facebook, is it your belief that that information did feed into that psychographic scoring?

[Jeff Silvester]

With the information that has come out now, it looks as though they may have indeed used that Facebook information to come up with their scoring. I don't know that to be the case.

[The Vice-Chair, Liberal, Nathaniel Erskine-Smith]

But it's likely with what we know now.

[Jeff Silvester]

It appears to be the case.

[The Vice-Chair, Liberal, Nathaniel Erskine-Smith]

Right.

In terms of additional data, you've got the psychographic scoring, and you say that you know this number, but you don't know what it's based on. Then Mr. Vickery comes along and somehow stumbles into this hub database, and he's testified before us that it's pulled from a master dataset of.... He sees some RNC Trust information, some election list information, some information from the Koch brothers, and more. There are many data points for different individuals, including data points, such as that they lead a biblical life. However, you've indicated here that you don't keep data, that you destroy data, and that you don't share data across campaigns.

How do we reconcile this information that we have? Why is all of this information in a database from AIQ that is pulling it in from so many different sources?

[Jeff Silvester]

It's not one database. The Git repository, though it does have its own database to keep track of what's there, is essentially a file storage location. When those databases were backed up, some of those back-ups were intentional. For example, they were supposed to back up the list of users of a website. However, with respect to backing up lists of people who had signed up to a website, those weren't supposed to be backed up.

Now we've put in place measures to make sure that doesn't happen again. When I was here, I described in detail how that happened and what we've done to change that, but we're going to make sure that doesn't happen again. We're still investigating how Mr. Vickery got access. I said before that I would report to the committee when we're done. I'm committed to doing that, as well, as soon as we're done with the report.

[The Vice-Chair, Liberal, Nathaniel Erskine-Smith]

Let's take the example of leading a biblical lifestyle. Would you target potential voters based on that criterion?

[Jeff Silvester]

Some of the information that's in that repository is not information we created, used, or did anything....

If a client had information on their server, and we backed up the code from that server, we may have pulled some of their information inadvertently. That's part of the problem with what happened originally.

With respect to that specific criterion, I have never run ads for anybody targeting “living a biblical lifestyle”, but I know that in the work SCL did, they certainly had a broad range of issues and all kinds of things they were providing to campaigns.

[The Vice-Chair, Liberal, Nathaniel Erskine-Smith]

In retrospect, knowing what we know now, do you think you should have exercised more due diligence in taking information from SCL and simply converting it into advertising and targeting, and have been more aware and perhaps asked more questions about where the data came from?

[Jeff Silvester]

We did ask questions about where it came from, but the information we got was that it was from public data sources, and there are tons of them in the United States. We were unaware they were obtaining information improperly at the time. I don't know that we could have asked more questions. It's difficult to look back retrospectively and say what I would have done at the time, were I to know what I know now, because we didn't.

With respect to everything that's transpired after having worked with SCL, would I do it again? I probably wouldn't, given that I've now been here twice. It's not that I mind, but I have been here twice, and there are all the investigations, media attention, people camping outside our houses, horrible messages sent to us, and all kinds of things. Do I want that? Would I wish that on my enemies? No.

[The Vice-Chair, Liberal, Nathaniel Erskine-Smith]

Thank you very much.

With that, we go to Mr. Baylis.

[Frank Baylis, Liberal, Pierrefonds—Dollard, QC]

Thank you, Mr. Chair.

What's data harvesting?

[Jeff Silvester]

What is data harvesting? It would be collecting data from sources, I suppose. It just depends on which context you're speaking of.

[Frank Baylis, Liberal, Pierrefonds—Dollard, QC]

What's caller ID spoofing?

[Jeff Silvester]

In your office, for example, you might have four phones, but no matter which one you dial from, the main office number is the one that comes up when you make the call. It's not the actual number you're calling from, but when people call, they get back to the office you called from, essentially.

[Frank Baylis, Liberal, Pierrefonds—Dollard, QC]

But what's caller ID spoofing?

[Jeff Silvester]

That's what it is. When you make that outgoing call, it places a number that's different from the number that's calling on the call display, so that when someone sees the number and responds or calls back, it will go to the number that came on their call display. It's not the one the call originated from.

[Frank Baylis, Liberal, Pierrefonds—Dollard, QC]

Right.

Has your company written programs that do data harvesting?