Evidence of meeting #139 for Industry, Science and Technology in the 42nd Parliament, 1st Session. (The original version is on Parliament’s site, as are the minutes.) The winning word was facebook.
A video is available from Parliament.
4:25 p.m.
Laurentides—Labelle, Lib.
That's right. James Rhodes played Bach on his piano and uploaded the video to Facebook, and Facebook's Content ID equivalent—I don't know what you call it—
4:25 p.m.
Laurentides—Labelle, Lib.
—triggered and said, “I'm sorry; Sony owns the copyright on Bach,” which is patently false.
4:25 p.m.
Head of Public Policy, Facebook Inc.
That's interesting. I'm not familiar with that particular example. I think you are right that—well, grosso modo, how it works is that for any reported piece of content, as I mentioned in the opening statement, certainly an individual at Facebook would review that report and ensure that, first, the information's complete for the reporting, and then that it's a legitimate or a valid request for removal.
4:25 p.m.
Laurentides—Labelle, Lib.
How many reports do these people have to go through in a day?
4:25 p.m.
Head of Public Policy, Facebook Inc.
As I said, we had close to half a million reports, I think it was, in the first half of 2018, and that resulted in about three million copyright takedowns globally. I think there are some other things that we do have to complement that, which are our automated systems. Potentially, again, not knowing the specific case, I couldn't say, but—
4:25 p.m.
Laurentides—Labelle, Lib.
It's just an example. I've seen this kind of case many times, coming across my Facebook feed, of all places.
4:25 p.m.
Head of Public Policy, Facebook Inc.
Right. Potentially, this could be an automated system example. I would obviously also want to add—and I'll turn it over to Probir in case you have something to add, Probir—that the concern for us is always, as you point out, sir, false positives. We want to be tough to ensure that we are protecting the rights of rights holders, but on the flip side—and that's why I mentioned it—as a platform, we always want to be careful about how we balance this. You don't want to be so aggressive that you accidentally take down something legitimate. I would never say that we're perfect, but again, not knowing the specific case, I couldn't give you a satisfying answer on this one.
I don't know, Probir, if you have some thoughts on it.
November 26th, 2018 / 4:25 p.m.
Probir Mehta Head of Global Intellectual Property Policy, Facebook Inc.
Thanks very much.
I would add that with every system, we're always looking to make it better. The engineers and the personnel who work on our systems are constantly sitting down with rights holders and users to try to....
Again, if I understand this correctly, if it was Rights Manager, this is a system that relies on input and feedback. I'm not familiar with the specific case, but again, as with everything, we're constantly trying to make it better.
4:25 p.m.
Laurentides—Labelle, Lib.
Can anybody in the world, and this applies, again, to Mr. Kee....
I don't know if you have any comments, Mr. Kee, on the previous question before I go to the next one.
4:25 p.m.
Public Policy and Government Relations Counsel, Google Canada
I would like to just quickly comment on that.
You actually highlight a challenge that we have when we implement systems like this. They certainly go well above and beyond our minimum requirements under United States, European or even Canadian copyright law. We have a court challenge in terms of balancing those rights. In the case of Content ID, that's actually why we have an appeal system, the idea being that if something gets claimed that shouldn't be claimed, you appeal it, and then ideally the claim will be released.
In this case, it should never have been claimed in the first instance—
4:25 p.m.
Laurentides—Labelle, Lib.
But it's still a “guilty until proven innocent” system for both companies.
4:25 p.m.
Public Policy and Government Relations Counsel, Google Canada
In terms of the assumption that was being made because a match was being made, that's correct.
4:25 p.m.
Laurentides—Labelle, Lib.
Can anybody in the world upload their...?
Mr. Chen, does Facebook have a name for Content ID, so I don't have to keep referring to Content ID?
4:25 p.m.
Laurentides—Labelle, Lib.
Rights Manager, Content ID—great; I have two terms. Thank you.
Can anybody in the world upload their content to these two systems, or is it only companies or larger copyright holders that can do so?
4:25 p.m.
Public Policy and Government Relations Counsel, Google Canada
I'll start.
Actually, only larger companies can, because it is an extraordinarily powerful tool. It requires quite a bit of proactive management. We actually have 9,000 Content ID partners. Generally these are larger entities that have large libraries of content that require this kind of protection, and they also have the dedicated resources to manage it properly, especially because you can control the way the system manages for each individual territory in a very nuanced kind of way.
We also have other tools that are available at other levels and are available for other creators—for example, independent creators—or they can work through, frankly, Audiam, for example, which can actually manage the Content ID system on their behalf.
4:25 p.m.
Laurentides—Labelle, Lib.
By having only 9,000 creators—and I'm assuming it's fairly similar for you, Mr. Chan, from what you said here—if it's only 9,000 creators who are permitted to submit to the system, does that not necessarily hurt the smaller producers?
4:25 p.m.
Head of Public Policy, Facebook Inc.
Sir, if I may, I'll just refer to my colleagues. We have a slight nuance for rights management.
4:25 p.m.
Head of Global Intellectual Property Policy, Facebook Inc.
It's application-based, to enter into the Rights Manager system. It's typically large, commercially minded rights holder groups, but also we have recently been testing for smaller creators as well. However, again, ultimately it's a needs-based assessment.
4:25 p.m.
Laurentides—Labelle, Lib.
Do these systems currently handle Canadian fair dealing exceptions in their enforcement?
4:25 p.m.
Public Policy and Government Relations Counsel, Google Canada
Essentially, no, effectively because fair dealing is a contextual test that requires analysis on each individual case. On any automated system, no matter how good the algorithm, no matter how sophisticated the machine learning that we're applying—and we are doing that—basically, we'll never be able to ascertain that. This is why it's critically important that it has an appeal system: it's so if a video that is a clear case of fair dealing is allowed and then gets caught by the system, they can appeal that decision. It will basically be determined and released.
4:30 p.m.
Laurentides—Labelle, Lib.
Then in both of your situations, why isn't the system set up to say, “You have had a flag; please respond within 24 hours, and then we'll take it down”, to make it a system where one is innocent until proven guilty instead of guilty until proven innocent?
4:30 p.m.
Public Policy and Government Relations Counsel, Google Canada
In some instances, that does happen. It depends on what policy the rights holder has chosen to enact and how they've selected to do so.