Kernochan Symposium 2024: Panel 4
Panel 4: Collective Licensing and Antitrust Concerns
Transcript
Kernochan Symposium 2024: Panel 4
MODERATOR: All right. Our final panel of the day is going to be looking at antitrust concerns around collective licensing. Our speakers are Anu Sawkar from the Federal Trade Commission, Professor Johan Axhamn from the University of Lund, Judge David Strickler from the Copyright Royalty Board, and my colleague here at Columbia Law School, Professor Tim Wu. So we will get things started. Our format will be similar to the prior panel. Each speaker will speak for about 13 to 15 minutes, after which I will ask a series of questions. So to kick things off, Anu.
ANU: Thank you so very much for inviting me to participate in this important conversation today. Before I begin, I should note that the views I express today are my own. And they don't necessarily reflect the views of the commission or of any particular commissioner. I'm sorry.
So I want to use the next few minutes to share a couple of stories. The first is my personal story, so you can better understand the perspectives I'm bringing to our conversation. And the second is the story of the American music performing rights organizations and how they came to be subject to an antitrust consent decree regulating their behavior.
After my clerkships, I joined the United States Department of Justice's Antitrust Division as an honors attorney. And my first home there was the enforcement shop that covered many content creating industries. I worked on numerous book publisher mergers and broadcast television station mergers. I was also part of the team that brought the successful e-books price fixing case against Apple and the major book publishers.
You might remember this. Amazon was trying to enter the e-book space. Amazon was pricing New York Times bestsellers at 9.99, I think, and the publishers freaked out. And we alleged that Apple led a hub and spoke price fixing conspiracy to try and raise the prices of e-books. [COUGHS] Excuse me.
After that litigation, I ended up spending a lot of time working on music licensing issues, including working on the antitrust division's 2014 review of the ASCAP and BMI consent decrees, and I'll speak more about that in a minute. And I also joined the Copyright Office's music listening tour, as the office was working on a music licensing study in 2015. I eventually moved over to the Antitrust Division sister agency, the United States Federal Trade Commission. The FTC has a dual mission to encourage competition and to protect consumers. FTC shares antitrust enforcement duties with the United States, and-- [COUGHS] excuse me-- with the Antitrust Division of the Department of Justice. And the two agencies collaborate all the time.
As you probably noticed, both agencies are doing a lot of work in the AI space these days. And last year I worked on around table about AI and the creative economy, which elevated concerns from a particular subset of creators, including authors, musicians, artists, and actors. The FTC submitted comments to the Copyright Office's notice of inquiry on copyright and AI. That comment highlighted both the competition and consumer protection law have a role to play in the conversation we're having today.
One of the questions the Copyright Office asked in its notice was about the feasibility of licensing copyrighted works for use as training data, which was a topic we explored while working on the roundtable. Given all the time I've spent working on music collective licensing issues, I was, of course, intrigued about what role direct licensing and collective licensing can play in the AI ecosystem.
What lessons have we learned from our past experiences and how should we use those lessons to inform the next step we take as policymakers and competition enforcers? I think this is a timely and important conversation. And I'm really glad that we're all having it together today.
The second story that I want to tell starts in 1914, which was the year the American Society of Composers, Authors, and Publishers, which we call ASCAP, was founded. Also happens to be the year that the Federal Trade Commission was founded. And we just celebrated our 110th birthday this week. So happy birthday, FTC.
[LAUGHTER]
Let's see. So ASCAP's mission was to create a blanket licensing system for public venues, like restaurants, bars, and performance venues that play music for their customers. You can imagine that it was very hard for composers to monitor the use of their works as they were being performed all across the country in public spaces. So those venues weren't obtaining direct licenses from the music owners for the individual songs. And the composers weren't able to police the infringement activities that were happening nationwide. So a blanket license solves and simplifies all these problems for both the venues and for the owners of the compositions.
The venues can get a single license, and that license lets them immediately perform anything that's in the repertory. The owners of the copyrights are able to monetize and polices the copyrights, collects the fees, and distributes to the owners. ASCAP was the first PRO, performance rights organization, in the United States, and for over 30 years, it was the only PRO.
Selling blanket licenses to radio stations quickly became a significant source of royalties for ASCAP and its-- for ASCAP members throughout the '20s and '30s. So, for example, in 1932, ASCAP set the radio broadcast licensing fee to 5% of a radio station's ad revenue. As radio profits increased, ASCAP members argued that radio was depressing sales for physical copies of musical works and sought to increase revenues from public performance royalties. I feel like that's sort of a theme that we keep hearing over time in this space.
The radio stations became more and more frustrated about how the cost of the blanket license was increasing. And in 1939, they retaliated. The National Association of Broadcasters decided to start a competing PRO called BMI, Broadcast Music, incorporated. BMI was created to compete with ASCAP, and it was intended to open up the competitive licensing landscape and provide competitive pressure on ASCAP royalty rates. BMI quickly attracted members in newer music genres, including country and the blues, That couldn't join ASCAP.
In 1948, the ASCAP announced that it was going to demand a 15% royalty rate. That's a big increase over a short period of time. The radio stations responded by boycotting music from ASCAP members. The stations were actually able to avoid licensing from ASCAP, which had a large portion of the music, by playing music in the public domain and playing music from BMI. So at least at that time, ASCAP and BMI were basically able to function as substitutes-- and as a competition lawyer, that makes me happy-- at least for radio stations.
And this is where the Department of Justice's Antitrust Division first entered the scene. DOJ began to investigate allegations of anti-competitive conduct by ASCAP and eventually filed complaints against both ASCAP and BMI in 1941. DOJ charged that the blanket license was an illegal restraint of trade and that arbitrary prices were being charged as a result of an illegal copyright pool. In antitrust speak, DOJ was basically saying that ASCAP and BMI were price fixing and were unlawfully exercising market power acquired through the aggregation of public performance rights in violation of Section 1 of the Sherman Act.
I think this is a good time to pause for a moment and give a quick primer on the relevant antitrust laws. So the statute we'll be focusing on today is the Sherman Act. Section 1 of the Sherman Act prohibits agreements that restrain trade. The courts have interpreted the statute to prohibit unreasonable agreements, not just any agreement.
Section 2 of the Sherman Act focuses on monopolization and attempts and conspiracies to monopolize. And here the focus is on acting unreasonably to gain a monopoly position or to maintain that monopoly position. The FTC actually enforces the FTC Act. But any conduct that violates the Sherman Act also violates the FTC Act.
There are three different frameworks that the courts will use to determine whether an agreement is an unreasonable restraint of trade under Section 1. So some types of agreements, including horizontal price fixing, are per se illegal. You just can't do them. These are the types of agreements between competitors that are so obviously anti-competitive that no further analysis is needed.
Most agreements, however, are analyzed under a flexible rule of reason, where we analyze whether the benefits of the agreement outweigh any competitive harm. There's a third intermediate category that some courts have used called a quick look test or an abbreviated rule of reason. They may use that when the anticompetitive effect is clear but the restraint doesn't fall into that bucket of things that we know are per se offenses.
So under the rule of reason, we typically define a relevant market. We determine market power. And then we look for the existence of anticompetitive effects. If we are able to show anticompetitive effects, the burden is going to shift to the defendant to show that their conduct is procompetitive.
For example, with the PROs, there's a huge efficiency in transaction costs and benefits to the users. If the defendant is able to show procompetitive benefits, the burden shifts back. And we need to show that there you can achieve those effects through a less competitive means. So the courts do an elaborate balancing after collecting a lot of information, and this is a very fact-intensive inquiry.
Beyond that legal framework, I want to note that the DOJ and the FTC have issued IP licensing guidelines, which explain the analytical framework that we use when we look at this stuff. The guidelines note that licensing arrangements are generally analyzed under the rule of reason. And we don't presume that IP rights convey market power. I think this is a theme that's come up.
Individual copyrights aren't generally going to convey market power because there are often substitutes available. But when we aggregate those copyrights, that can create significant market power. And because the blanket licenses provide for a single fee the right to play many separately owned works, they eliminate the competition that would otherwise exist among those works if they were being licensed directly.
And well, maybe at some point in the past, different PROs acted as substitutes, I don't think we can say that anymore. And for many users, like streaming platforms, the PROs are complements. You need to have everyone to have the coverage to operate in the market. So DOJ ultimately settled using consent decrees, which is basically a court-enforced contract that was meant or is meant to prevent the anticompetitive exercise of market power while, at the same time, preserving the efficiencies of the blanket license.
And I think people have talked about some of those important provisions. One, they can only do performance rights, can't expand into other things, so they can't leverage their market power. Two, only blanket licenses, you can't do single licenses or mixtures of licenses, so you can't leverage the market power from that. The licenses are nonexclusive to the PROs, super important.
If you look at the licensing guidelines, we really want to see that direct licensing is available outside of the pool to check the power of the pool. Also, you have to license everyone who wants a license, nondiscrimination. Similarly, situated licensees get the same rate. And of course, everything happens in the-- excuse me-- in the shadow of the rate court.
So the way we've consumed music has changed over time. We talked about how the formats have changed. And with that shift, we've seen that the math for the various stakeholders has changed, too.
In the 1930s, we saw copyright owners were concerned that broadcast radio was depressing their sales of physical copies of music. But by 2014, we were deep into the transition into streaming, and we were hearing very similar concerns. So the music publishers thought that they would be able to obtain higher rates from the streaming services if they negotiated direct licenses outside of the PRO.
So they petitioned the DOJ to open up a review of the consent decrees with the intent that we would allow partial withdrawal of the rights so they could pull the streaming rights, independently negotiate that, but leave the blanket licenses for brick and mortars in place for efficiency reasons. So the overall purpose was to try and increase the licensing rates through direct licensing. We can talk about this more, but that consent decree review process ultimately left the decrees in place without modifications.
And our closing statements concluded that, although stakeholders on all sides of the conversation have raised concerns with the status quo, our investigation confirmed that the current system has well-served music creators and music users for decades and should remain intact. In 2019, we opened another degree review. And at that time, the administration was undertaking a massive project to review almost 1,300 legacy consent decrees.
So they came up again. And in 2021, we left it-- well, I'd left the agency at this point, so I was not involved. But DOJ left the consent decrees in place, again, noting that there is significant reliance on the decrees within the licensing community. Throughout the division's investigation, many licensees expressed that their view that the decrees were largely working.
So why have these lasted so long? I think because our market power concerns still remain and the consent decrees continue to offer important safeguards to prevent the exercise of that market power. So I look forward to continuing our conversation today about how licensing can be used in a procompetitive manner to save AI.
[LAUGHS]
Just kidding.
Thank you.
[APPLAUSE]
MODERATOR: OK, next we have Johan Axhamm, who's going to--
JOHAN: Thank you very much, an thank you very much for the invitation. I'm very honored to be here. I was a visiting scholar a few years ago. And it has also been a privilege to listen to the very long discussions that we have heard before lunch and also after lunch.
I am a professor at Lund University. Lund is a minor-- it's a small city in Sweden with a long history. The University was established in 1666.
I also have a background as a special government advisor to the Minister of Justice in Sweden. So I have, for example, negotiated in Brussels EU directives, including one directive which I will refer to in my presentation, which is the EU Directive on Collective Rights Management. I also have experience from representing the Swedish government in WIPO negotiations, most recently the diplomatic conference in May on genetic resources related to patent applications.
This is the outline of my presentation. I will begin talking a little bit about, say, machine learning from a copyright perspective. And it concerns the scope of the right of reproduction, when do we at all have reproduction, and then talk about-- which might be considered the main part of my presentation-- the so-called extended collective licensing model.
It's a type of licensing arrangement that was established in Sweden and the other Nordic countries in the 1950s. And it has been used and is still in use in many areas of so-called mass use. That is to say, where there is a large amount of works and a large amount of right holders, it might be difficult to foresee beforehand exactly what will be used and to what extent.
And then I'll also talk a little bit about antitrust concerns and this directive that I mentioned. This directive is, to a large, extent based on EU case law on antitrust related to collective management organizations. So that directive is, one could say, a codification of EU case law on abuse of dominant position when it comes to collective management organizations.
The right of reproduction, it's not entirely clear the scope, when do we have a reproduction. In one of the major EU directives on copyright is called Infosoc Directive from 2001, includes a definition of reproduction, which includes not only permanent reproductions but also temporary. And this means that, to the extent that machine learning entails a temporary reproduction, then it's something that the rightholders as a starting point can control.
And this is not clear, or you cannot-- this does not follow from the Berne Convention, and it's not entirely clear either from the WIPO Copyright Treaty. This is actually one of the areas where I think it was a vote during the diplomatic conference in '96 on the agreed statement to Article 1(4) in the WIPO Copyright Treaty concerning the scope of right of reproduction in the digital environment, where the United States and the then European communities had different views.
And the European community wanted to make it clear that the right of reproduction at international level covers also temporary forms of reproduction. As far as I understand, the US did not want to have this explicit reference. So if you read the agreed statement or Article 1(4), you will see that it's not entirely clear at international level that the right of reproduction covers temporary forms of reproduction. But within the EU, we have settled this with the Infosoc Directive.
Some years ago, the European Commission initiated an update to the EU copyright framework to, say, make it more fit for the digital environment. And for example, the European Commission proposed what was than to be the Article III in the adopted version of the directive, an exception or limitation for text and data mining for the purposes of scientific research.
However, during the negotiations, there was a strong push from certain stakeholders to introduce a more general exceptional limitation for text and data mining. And that's the one that you find in Article IV. And what's the difference between these two? One difference is that Article 3 is specific, so it only covers specific users, that is to say, for in the scientific area, whereas article 4 covers all types of text and data mining. Or it's not limited to specific users.
In addition to that, Article 4 opens up the possibility for rightholders to opt out. But the starting point is that, to the extent that the rightholders have not opted out, they are, say, covered by this exception or limitation for text and data mining. So their works can be used for text and data mining unless they opt out.
So the article 4, say, turns the presumption around and says that, unless you opt out, your work can be used for text and data mining purposes. But it's not entirely clear if you compare the definition of text and data mining with, say-- and this was also touched upon before lunch. Is text and data mining the same as machine learning?
It's not entirely clear. You have a definition there of text and data mining in Article 2(2). And this was also mentioned before lunch that, in the recently adopted EU regulation on artificial intelligence, there are provisions on transparency, meaning that the one who carries out text and data mining or operations in relation to copyright protected works must be transparent but not on a work-by-work basis but on a more general basis.
But this of course, gives the right holders the possibility to know whether their work has been used as or part of this text and data mining process. These transparency provisions in the AI regulation has caused some concerns for rights holders because it implies, more or less, that text and data mining is the same as machine learning, so some concerns from the perspective of rights holders when it comes to that, even if it's, say, positive that we have a transparency provision.
When it comes to extended collective licensing, sometimes, say, copyright is discussed in a quite, say, binary perspective. So we have exclusive rights. And then we have exceptions and limitations. But at least in countries with a long tradition of collective rights management, we deem collective rights management to be, say, something in between, so a fair balance between individual management, say, individual licensing, and exceptions and limitations.
Exceptions and limitations, as far as I understand, it would include provisions on compulsory licensing and statutory licensing. So rightholders holders have no control over, say, the permission or giving a license. They must provide-- or accept use, whereas collective management, there is normally an element of control, at least for the collective management organization.
When have we introduced provisions on extended collective licensing? In Sweden, we have had it since the 1950s, and this is the same for the other Nordic countries, Finland, Denmark, Norway, and Iceland. And I know that other countries in the European Union have similar, say, arrangements for the management of rights, sometimes referred to something else or something similar.
But say, the legal effect of an extended collective licensing is that the agreement between the collective management organization by provision in law is extended to cover all right holders who are not members of the organization or who has not given their mandate to be represented by the organization. The provisions, the statutory provisions, they also give rightholders the-- so the ones who are covered by this extended effect-- the possibility to opt out. So I don't want to be part of this extended effect.
To the effect that right holders are covered by the extended effect, they have a right to equal treatment when it comes to, say, remuneration. And according to Article 12 in this EU directive from 2019, there is also an obligation to inform rightholders that there is this license with an extended effect.
This article 12, say, legitimizes the extended collective licensing model. So in EU law, it's now an accepted form of what one could say an arrangement concerning the management of rights. So the extended collective licensing model is not an exception or a limitation but rather an arrangement concerning the management of rights.
But you can only use this model in situations where it's practically impossible to give an individual license or individual permission. So one can say that we use extended collective licensing in these mass-use situations.
And examples of that could be prime-- so user works for primary broadcasting, use of works for educational purposes in, say, universities and schools. It could be use of works in archives and libraries. And it could also be, say, for reprography and similar situations. These are situations where I gather that the provision of fair use in US copyright law would cover some of these uses but maybe not all of them.
Yes, two minutes. This is actually one of the hot topics when it comes to copyright in the European Union at the moment. Rightholders are not, say, too comfortable with this article 4 on the exception limitation for text and data mining in this directive from 2019. So there is a discussion on, say, can we standardize-- and so they opt out of that exception-- and limitation and then meet the market demand with an extended collective license.
And what would be the purpose of that? The purpose is, of course, that then the right holders can control the extent of the use and also be remunerated for the use because, otherwise, article 4 being an exceptional limitation, then its use with no remuneration to the rightholders. But this, of course, is preconditioned in practice that there are collective management organizations that are ready and available to provide a license and that this is also something that is accepted by the market.
And I will touch upon this on the antitrust concerns. We have a provision in Article 102 in the Treaty of the Functioning of the European Union on prohibition, on abuse of a dominant position, that is. And we have quite a lot of case law related to that article as applied to collective management organizations concerning licensing conditions, so the, say, the relationship between the CMO and users on, say, how tariffs should be calculated, et cetera.
There are also case law concerning the scope of the mandate that the organization can require, or when it comes to rightholders, should it be all territories and all users and all works? Or should it be more, say, limited in scope? And we also have case law related to the relationship between collective management organizations.
In general, and EU law makes a distinction between, say, the existence and the exercise, so the existence of the ECL is not put into question but rather the exercise. And so when an organization is exercising an extended collective license, it must be in line with this case law on abuse of a dominant position.
And it's now, say, simplified in practice because this EU directive on collective rights management codifies this case law. And I will provide some examples of this.
But first of all, so to make it clear, I got a question from the organizers. So the directive includes a definition of what is meant by a collective management organization. And that definition makes clear that a collective management, it doesn't matter whether the mandate to represent rightholders is based on, say, an explicit mandate or whether it's something that is provided by law. So this reference to by law here includes extended collective licensing, the extended effect, as I mentioned.
And this is the last slide. One example, licensing conditions, so this is where-- so the directive codifies this case law. It says that when collective management organizations and users negotiate, they should conduct these negotiations in good faith.
Licensing terms shall be based on objective and nondiscriminatory criteria. And tariffs for exclusive rights and rights to remuneration shall be reasonable in relation to, for example, the economic value of the use of the rights in trade, taking into account the nature and scope of the use of the work and other subject matter. This is almost word by word from case law in this area.
To sum up, could we have an extended collective license for machine learning? Yes. But we must have a simplified mechanism for writers to opt out from this article 4 on text and data mining. And there must be market conditions when it comes to the structure and, say, the acceptance of collective rights management. Thank you.
[APPLAUSE]
MODERATOR: Do you want to speak from here, or?
JUDGE STRICKLER: No. I'm going to go up there. I thought you were going to go up there.
MODERATOR: No, no, no. I'm just going to turn off the slide.
JUDGE STRICKLER: OK. Good afternoon. Thank you for the privilege of having me speak here today. It's been enjoyable to listen to the other panelists, and I'm glad to make my contribution.
As I think has been made known from some questions and answers previously, I am the economics judge on the US Copyright Royalty Board, and we preside over trials, which, among other things, set royalty rates for sound recordings, also the underlying musical works. Now, earlier in the panel before us, my colleague Judge Rui went through some of the licenses, all of the licenses, really, quite comprehensively with regard to those licenses for which we set royalties.
I want to get under the hood a little bit in that regard by explaining how we have addressed those issues and how we have set royalties in actual cases, which is not to say that it will always be that way in future cases. Every case is decided on its facts. But economic principles tend to have a level of constancy to them. So to the extent they apply in one case, they may well apply in another case. But the facts may show that there are new economic concepts that didn't need to be addressed or were not addressed in a previous case, which you'll see as we go forward this afternoon.
So my focus today is from a more general sense to talk about how economic regulation or price regulation in lieu of what one might call traditional antitrust relief and remedies, makes sense as an adjunct to that more traditional antitrust concern, particularly as it relates to oligopolies in particular-- and I'll explain the terms as I go through-- complementary oligopolies and the presence of potential tacit collusion as opposed to explicit collusion among oligopolists. These are things that, to the extent they may be economically deleterious, are not well suited for antitrust relief and remedies because antitrust law, as opposed to antitrust economics, if you will, is more focused on conduct and behavior rather than necessarily just the economic result.
That is to say, high prices do not mean that there's a monopoly problem or a problem for antitrust or the existence of tacit collusion can either be proven or, even if proven, is worthy of a remedy. But separate and apart from antitrust, there may be an appropriate place for economic regulation.
And this is a really old concept. This concept actually goes back at least as far as 1968, '69, when Richard Posner, before he was even a judge on the Seventh Circuit, wrote an article where he spoke of the fact that antitrust is not well suited for these concepts, and it would make more sense to either just to perhaps-- he buried this. I'll use the word bury. He certainly did not-- in a footnote, the idea that there might be price regulation as an alternative, rather than antitrust activity in that regard.
And that concept is what animates the economics of what we do. So that's my general focus, to talk about economic regulation as an adjunct to antitrust. And I want to do it through the focus or through the lens of our CRB rate regulation with regard to two particular cases, Web IV which and Web V now. All jargon, and jargon is going to be lost on people who are not familiar with the jargon, so allow me to explain.
Web means webcasting. So webcasters or Webcasting IV was the fourth iteration of the case law, the cases that developed. Webcasting V followed after it.
When you hear webcasting. Think internet radio. Think, for example, Pandora. One of the other panelists, Mr. Marks, has represented Pandora before us, pursuant to these types of rate regulations.
And sometimes I wonder why there's so much jargon when we have three different names for the same exact concept. And the statute, internet radio is noninteractive. It's also considered, as I say, internet radio. It's also sometimes called, somewhat inaccurately, lean back radio. So there's a lot of synonyms, which I think are-- I don't know if they're meant to confuse, but they have the effect, very often, of confusing people.
Anyway, so when we set the Royalty rates for internet radio, we have a statutory standard that we must apply. And that's the willing buyer, willing seller, marketplace standard. That's a paraphrase, but that's the shorthand version of it.
Now when the judges have set such a standard, particularly developed in Web IV, again, in Web V, we've set what we consider to be-- got from the evidence and the expert testimony in the case an effectively competitive standard. We give you another synonym for that-- workably competitive.
Economists have long called it workably competitive. Lawyers have called it effectively competitive. It's the same thing. It basically means meaningful price competition, and I'll get into that in a little bit as well.
And I'll do that by explaining the standards and how we applied that standard in these two cases. And let me tell you something else in case you don't know this-- and many of you may well know this, but it helps to understand the cases.
When it comes to noninteractive-- I'll use that purposely, that statutory phrase for webcasting. The way it works is, if you have a Pandora account, you seed a station, S-E-E-D, seed a station with a song or with an artist. And then the algorithm takes over, and the algorithm does the work, trying to pick out the song that it thinks would be next, most likely-- that you would most likely want to hear based on your listening habits or based on that particular music and moving forward through the songs in that order.
And there are restrictions as to what can be played. You can't repeat one artist too often. You can't repeat a song too often. These are called the sound recording performance complement is the jargon in the statute, to limit the amount of repetition that can be done.
And all that's important backdrop, not just for the sake of letting you know how that works, if you didn't. But it's important backdrop for how effective competition can be created and was created according to the judges in Web V and Web V.
So Web IV was in 2016. And I'm happy to say that the DC Circuit affirmed our decision in 2018. And Web V was decided in 2021. And the DC Circuit affirmed that in 2023.
And let me tell you, we all breathe a sigh of relief when they affirmed it, not because-- just because we're happy that they agreed that it was reasonable and not an abuse of discretion. But as Judge Rui pointed out, if they don't agree and they kick it back, it's a hell of a lot more work for us than we had anticipated, although the particular case that judge Rui I don't think mentioned specifically was a different case, the phono records license case, where I had dissented from the other two judges.
Judge Rui was not on the panel, by the way, back then. And I had dissented. And the DC Circuit essentially agreed with my dissent and kicked it back. That was quite a laborious process, proving the old adage that no good deed goes unpunished.
[LAUGHS]
So anyway, back to our story, Web IV, the evidence we relied on was twofold, and it was in the nature of benchmark rates. What we look at are representative contract rates or, arguably, representative contract rates that the parties put forward that say are analogous to the rate that you're trying to set, in this case, the noninteractive rate.
We had two types. One was the interactive rate, which is on-demand music. So on-demand music the relates to the royalty-- or to the music that's played by streaming services, where you can pick any music that you want at any time.
And the labels negotiate with the streaming services. Think Apple, Amazon, Google, Spotify, Deezer, and a host of relatively smaller ones in the United States. Those transactions are not regulated at all. Those are pure market transactions.
And if you're a laissez faire fan or a laissez Faire economist, you love that. You say that's the way the market is supposed to go. That's the way this should be done. Just let the market work. Get government out of this.
And sometimes when I talk to people and bore them at cocktail parties, when they ask me what I do and I do this, they say, I didn't even government was involved in such a thing. And then they go look for an interesting conversation.
[LAUGHTER]
So then that was one type of benchmark that we received in the case. The other one was rather unusual, and it was a rate that was not from a different market but was from this very market, from the noninteractive market itself. There were two contracts that were presented in evidence.
One was a contract that Pandora had with an entity known as Merlin, which is a consortium of some of the larger indie records, independent record labels, sometimes referred to as the fourth major because it's so big. And also there was a contract between iHeart, a noninteractive service, and Warner, one of the bona fide, big three record labels. And I'll talk about each piece of evidence in that regard, starting first with the interactive contract rates, again, interactive meaning on-demand music.
You need all of the music in order to be able to play. So when the label-- to have a service. So when the labels negotiate, they have this tremendous power. It doesn't do you any good if you're a noninteractive service to have an agreement with Warner and Sony but not Universal, which may have, for argument's sake, 30% of the market because, if you're missing 30% of the market, nobody's going to want to listen to your service because 30% of the time those Universal artists are the ones that you're going to want to hear and you won't get it. So you have a serious problem there.
So that doesn't make it a good-- really good benchmark for the noninteractive service because you have this competition problem. The major labels do not-- because of that power, because of that structure, do not need to negotiate price because they are going to "take it or leave it" mode. They have what's essential for you. And when somebody is offering you something that's essential to you, you really don't have much bargaining power at all.
So you have a competition problem. And in the economic jargon of this, what the labels are complementary oligopolists, not just your ordinary oligopolists. But they're offering complements that are perfect complements. You must have all three. So in economic terms, they're perfect complements, and therefore, there's a tremendous amount of power.
Now, if your goal is to set an effectively competitive rate, if you think that's what the statute requires, that's not going to be-- going to do the job for you. So somehow that has to be adjusted. And by the way, more than a parenthetical but before I turn to the other evidence again, how do the record-- what did the evidence show that the record companies would do to ensure that they didn't have price competition?
Well, the evidence showed that they set most-favored nation clauses in all their contracts. They had nonsteering clauses, antisteering clauses. And they acknowledged in testimony that they did not engage in price negotiation with the streaming services in this benchmark interactive market.
So it wasn't just theoretical that that's the case. It was actually acknowledged and supported by the evidence. So there needs to be-- if you're going to use that, which we did for, ultimately, for the subscription service of Pandora and iHeart or any noninteractive service, you needed to make some-- you needed to make some sort of adjustment.
But let's turn to the other evidence, the noninteractive rates, the Pandora, Merlin rates and the iHeart Warner rates. This was unusual because it came out of that very market, which we don't usually see. Pandora and Merlin and iHeart, Warner, the rates in these contracts were lower than the existing statutory rates.
Now, at first blush, that seems irrational. Why would a licensor who could get, say, I'll just use a number at random, 20 for its royalty be willing to take 15 instead when it could get 20? Well, if that was all there was to it, they wouldn't. But what was utilized in those contracts was a concept known in antitrust circles and in this particular circles as steering.
These companies, Pandora and iHeart, said, if you-- I'll just use percentages. That may not track the case but for purposes of elucidation. Pandora would say, we'll give you-- we'll drop the rate by 12% to you, but we'll give you 15% more plays. So you will drop the price, but we'll increase the quantity.
So why wouldn't any company take that? They're going to get more revenue that way because it's a higher percentage increase in quantity than there is a decrease in price. For those of you fans keeping score economically at home, that's an elasticity concept. And that concept animated the self-interest of Merlin and in its contract with Pandora and Warner in its contract with iHeart, which was similar in economic effect.
So that was the reason why you would actually see rates that were lower. Now, how could Pandora do that? How could iHeart do that? Well, they had the technological capability because of steering to be able to do that. So that, as I said, it wasn't just to give background on how these noninteractive services work.
Once the algorithm takes over, it will rank the songs that it thinks you want to hear, and those songs will be played. But these noninteractive companies injected into their algorithm a pricing function or a pricing variable to say that if, for argument's sake, if it was the iHeart, Warner contract and the fifth song in the algorithm would have been a Universal song, the algorithm may drop that down to 15th, for argument's sake, and bring up to the fifth spot, the cheaper Warner one, so that ultimately the Warner songs would be played more often, reducing the overall royalty.
So Warner would get more plays, and it would be happy. iHeart would have to pay less revenue on those plays, so it would be happy. But to do that, it would have to substitute away from those that were insisting on the statutory rate, the 20. So those are the losers. There's no free lunch in economics.
The services benefited the steering approval of Warner and of Merlin meant that they were going to benefit as well because they got more quantity than they lost in price. But there's no free lunch. So the losers were the other labels that had to be steered against in order to make that happen because you can't increase the total number of plays.
That's a function of overall demand. It has nothing to do with the steering function. So that had to be done.
So what's interesting, curiously-- and maybe it's because I knew I was going to be giving this presentation today. When I was teaching earlier today-- I teach this stuff, for lack of a technical term. I teach this stuff both to NYU undergraduates in the Steinhardt Music Business Program and also in the law school today. This semester was the Steinhardt turn.
And I tried to-- and I asked the students. I said, well, what if you-- what should the record companies that are steered against, what should they do to? And I was met with silence, as I often am when I ask the class a question.
And they're thinking, professor, it's 8:30 in the morning. Don't do this to us on a Friday, no less. But I did something with them today. And this wasn't going to be part of my presentation, but it was so noteworthy to me because it's something I will say to people.
And I never really tried it in class. And I said to them, economic concepts immediately tend to intimidate a lot of people because they don't feel comfortable with it. They took an economics class in college.
They hated it. They hated their professor. And they don't want anything to do with it, so they're immediately frozen by it.
So I said, there's an easy way to deal with this. When you deal with it, be the person, be the person who's going to make money or lose money. Be the person who's responsible for making the label money or losing money and think that you're going to be called on the carpet by the CEO and says, we're losing money because they're steering against us.
What are we going to do? How are you going to fix this problem? What would you do?
After another short delay, a hand went up. And the student said to me, she said, she said, well, why don't we offer to steer as well? She was exactly right.
That's the point is that the threat of steering, when you know that that's out there, you're going to-- you, as a label, are going to say to Pandora or to iHeart, no, no, no. Don't steer against us. Don't steer against us.
We'll match it. We will match that lower price because otherwise we're going to lose revenue relative to where we otherwise would be if we just agreed and kept the old percentages, the old quantity percentages. Now that sounds esoteric, and steering sounds like a difficult concept.
But all that is price competition. It's just simple price competition. If you were a buyer for Macy's and you had to buy shoes that you were going to put on the floor, you play one seller off against the other, trying to get the lowest price you possibly could, and make that decision. And steering is no different. You're steering the business the way that you can.
And then I ask the students, I said, but if you wanted to do it differently, and it was somewhat nefarious, what would you do? I don't need that. I'm going to keep going.
And one student said, I would ask the other record companies to stop steering. I said, yeah, you could do that. That's the right answer.
But the problem is, if you go over and ask them and you ask them by telephone and it's being audiotaped by the Justice Department, you're going to go to jail for price fixing, which is exactly what happened in the Cargill case with regard to vitamin supplements in the agricultural field. It was actually, in that case, two oligopolists, one from Cargill I think it was, and I forget the other company.
One of them actually said on the audiotape, gee, I hope the-- I hope the FBI isn't listening in on this. And the FBI was listening in on it. So it wasn't just evidence. It was evidence of a confession.
So steering is price competition. And the judges recognize this. And we took the subscription rate, which had no steering involved, and we adjusted that down by 12% to reflect the value of steering. And in the ad-supported rate-- for the ad-supported rate, we just adopted the rates that were in those two benchmark agreements Pandora, Merlin, and iHeart, Warner.
Now, labels push back very hard on this. And they said, well, judge, judge, they can make all these abstract arguments what might happen in the hypothetical market. Anybody can make a hypothetical argument. But in the real world, this would never happen.
Well, why? They said, well, we would insist on most-favored nation clauses or antisteering clauses, or we would insist-- we would say we're going to withhold our entire catalogs unless you refuse to-- you agree not to steer. Or we'll require that you pay upfront royalties, take the historical average of plays, pay it all up front.
Then if you start steering and messing around with it, you're going to pay something. But you've already paid for everything, so you be paying twice for, in essence, for a certain percentage of plays. And our response in our decision was, yeah, we recognize you would do that.
But doing that is the use of the excess market power that we're trying to cure. So you're telling us we can stop you from making an effectively competitive market by doing things that are not effectively competitive. So that argument was rejected.
So is the last thing I want to say-- it'll be much, much briefer-- on the Web V case because it was the same economic logic, the same arguments being made, but one very important difference. And I was struck by the point Elliott made during his comments when his panel was up here, and he said that Apple is afraid of the majors. The majors are afraid of Apple. And that's exactly what animated the Web V case because, in Web V, the labels were-- the evidence clearly showed they acknowledged the tremendous power of Apple, Amazon, and Google.
And they were afraid of these tech rivals, who, if they had the markets to themselves, just the three of them, they could hold out. They could vertically integrate and start their own services. They could do a whole host of things with their tremendous power.
And as a consequence, it was important-- according to the evidence and as the decision said, it was important for the labels to keep Spotify going. They had to make sure that Spotify was there as a bulwark against these big companies. And they lowered their rate in the interactive market.
Again, this is the benchmark market. They lowered that rate for Spotify. And that took us a ways towards what otherwise would have been a 12% reduction. It did roughly half the work because it was a different power. It wasn't effective competition.
And this is where economics is a very large-- has a body of work. And if you can find the right economic analysis, you may well find the right answer. And in this case, it was a matter not of price competition but countervailing power. Spotify had a countervailing power that was thrust-- that it benefited from simply by not being part of one of these large ecosystem companies but by being an independent, sometimes some people would say pure play web caching, not so pure play anymore because they do podcasts and some other things as well.
But therefore, there was another power, another piece of economics generated both by the theory of-- economic theory of countervailing power, plus the evidence that showed, as one of the witnesses colorfully pointed out, for the record labels, big as they are, their entire business is a rounding error for Amazon, Apple, and Google. And so the countervailing power led to something lower.
And in that decision-- and I'll close with this-- we pointed out that it was this-- we had this interesting pitting of two different types of economic power, the economic power of the marketplace that the labels had as must have-- you must have the music versus the fear of what Apple and Amazon and Google would do if they used that power to squeeze Spotify to the point where it would leave the industry but Apple, Amazon, and Google could do pretty much whatever they wanted.
They could use music as a loss leader rather than an attempt to make money on this. And to make that point, and before I turn this over to Tim, we quoted specifically Tim and somebody who is 180 degrees counterpart to him, former FTC commissioner Joshua Wright. Tim had written in his book the Curse of Bigness about the distinction between market power and power derived from sheer corporate size.
And he talked about Apple. He wrote specifically in that book, just a handful of giants, Amazon, Google, and Apple, transcend the narrowly economic and that we pointed that out as a counterpoint to laissez faire economist, former FTC commissioner Joshua Wright, who criticized the emphasis on sheer corporate size as calling for nothing less than the complete dismantling of the consumer welfare standard and the consensus among antitrust practitioners, enforcers, and academics about how to promote competition.
Not so simple a position to take because the fact-- and that's the-- again, I'll close with the overall arching point. Price regulation by a regulator, particularly an administrative adjudicator like the Copyright Royalty Board or like the rate court, or also arbitrators can do the same thing on the CSAC, which is a small PRO and is not covered by the consent decrees, entered into an arbitration agreement after a private antitrust suit. And now their rates with licensees are subject to an arbitrator's decision.
So there's always a third party when you have market power. CSAC only has about 5% of the market, so it seemed to be implicitly suggested before they really don't have any particular market power. But they're a cream-skimming operation, and I mean that in the most positive sense.
They take high-value artists, give them a better return. And those high-power artists sign signed with them. And as a consequence, even they only have 5% of the market, they're a must have, too.
I have no idea offhand where Taylor-- who represents Taylor Swift, but if you only have 5% of the market and that includes Taylor Swift-- she's probably more than 5% of the market all by herself. But if you have Taylor Swift, you're a must have for anybody who wants to have a streaming service that has all the music.
So there's a question of bigness and power in the marketplace that are in tension with each other. And these types of cases are not solved very easily by ideologies and truisms. They're solved by detailed painstaking work, digging into the facts, ferreting out the economics, and coming to a determination. Thank you.
[APPLAUSE]
MODERATOR: Yeah. Take it away.
TIM WU: Thank you. Hi, everybody. Thanks for inviting me. I want to thank Kernochan Center and also students and Pippa and everyone who does all the work behind the scenes.
So my name is Tim Wu, and I teach copyright and antitrust, which seems to be why I'm on this panel, among other reasons. And I wanted to spend my time on a slightly related but slightly different topic, which is the prospect of using copyright licensing as a remedy in antitrust cases. I was thinking this is actually a slight-- a subject that hasn't in my knowledge gotten an enormous amount of attention but seems to be important.
This is usually-- and most antitrust, intellectual property attention focuses on patent. And the subject I'm going to talk about has gotten more attention in patent, but I'll mention a few ways in which it also comes to copyright. So I'm inspired in this topic by three particular cases or situations that inform it.
The first is the 1956 AT&T consent decree, which centered on a licensing remedy requiring the licensing of all of AT&T's over 8,000 patents, including the transistor patent. So intellectual property licensing has been used before. I also want to discuss the 1909 mechanical license, known well to this crowd, which in its own way was a settlement of an antitrust dispute or controversy.
And something very top of mind that prompts this discussion is the ongoing search for a remedy in the Google antitrust case, which is the subject of a lot of discussion right now, and the question of whether copyright licensing should possibly be a part of the Google antitrust remedy. And I think that's an open question that hopefully we can discuss. So let me talk about intellectual property licensing as an antitrust remedy, first through the 1956 consent decree.
This is a well-known topic among antitrust tech historians and practitioners but I think less in the copyright crowd. So in the early '50s, the Justice Department, which was then probably at its most aggressive-- I don't know. Now it's pretty aggressive too, but the '50s were pretty aggressive, so were the 1910s.
Anyway, they sued AT&T for monopolizing telephony and other things. And they wanted to break up AT&T but didn't. And instead, on the pain of breakup-- and there's a whole story in there-- AT&T agreed to two conditions. First, AT&T agreed to stay out of computing forever, which opened a lot of room for the development of a computing industry in the United States. That was independent telecom, a different Topic
But more relevant to this, to us, AT&T agreed to license over 8,000 patents, including its most valuable patent, the 1947 or '48 transistor patent, which was probably the most valuable patent of the midcentury. So I think most people who've studied this period think this was a successful remedy, partially because of the growth of computing in the United States but also the birth of the semiconductor industry in the United States, independent of Bell Labs.
What happened during that period, a lot of people quit Bell Labs and started their own companies. You have a Fairchild. You have Shockley. Eventually Intel ends up being born during this period. And the United States takes the lead.
There's been a few studies of the effects of the licensing, and they're very positive. So that's one model for when license can be-- copyright can be a remedy for antitrust cases. Another interesting model is the 1909 mechanical license for which this is obscure at some cocktail parties but not this crowd.
So I think copyright people think of that as a copyright thing. But if you've paid more attention, if you dig into that history, it was also motivated by antitrust issues. There was a concern at the time that there would be a monopolization of the, I guess, recorded music industry through a player piano company. I don't fully know how you pronounce her name, Aeolian, which had trying to do exclusive deals with all the publishers of music.
And so that and the Apollo decision led to the 1909 decision which created the compulsory license phonorecords, $0.02 a copy. Since then, it started being adjusted. So this, as people in this room know, the mechanical license turned out to be more popular than people thought.
And I think it was in 1961, the registrar proposed the abolition in the New Act, the 76 Act. They didn't know it would be 76 then. But people seemed to like it. And so the mechanical license survived the 76 Act.
And I don't know if I would-- know how popular it is, but at least it seems to have worked. And obviously, the compulsory license has been used for other areas as well. Now, that wasn't a remedy in an antitrust case, but it was kind of a congressional settlement of an industry dispute involving market power. Compulsory license has, obviously, been used since, often to settle disputes involving market power.
The cable license, compulsory license is a good example. It was of a situation where the concern was, combined with the broadcasting industry, combined with copyright holders who were loyal to the broadcast industry by collectively, to use a antitrust-- to use the antitrust word keep out or foreclose the operation of cable and the cable copyright, cable compulsory license has a lot to do with that.
As I recall, though, I can't find any sources for this. This is my third historic point. I think in the Microsoft antitrust case, there was at least some discussion of whether licensing Microsoft code should be part of the remedy. If someone remembers that better than me-- obviously, that's not what happened.
The Justice Department proposed ultimately a breakup along vertical lines or horizontal lines, depending how you describe it. And ultimately it was settled differently, which leads us to the Google case, which I'll introduce in case you're not an antitrust aficionado. So Justice Department sued the Google company for monopolization of general search, and there was a trial last year.
The core of the claim was that Google had basically paid Apple and other companies large sums of money. They ended up being over $20 billion. I think the total number paid per year last year was $26 billion or something like that. Google had paid this money to have them keep Google as a search engine and, therefore, excluded its competitor.
Justice also suggested that Google was paying off Apple to stay out of search with a $20 billion a year. And $20 billion a year is actually isn't chump change. I think it starts to become more of the entire record labels combined, so some big money.
So Google was found liable over the summer. And currently, the search is on for a remedy. And one of the things I think is interesting to think about is whether copyright licensing might somehow form part of that remedy. And let me throw out two ideas how. I don't know if they're any good or not.
But one little broader is to suggest that maybe Google should be asked to license its core code base as a remedy. It's a little strong. Google has about two billion lines of code at least, and they're overseen by a system called the Piper system. And it's a little bit like GitHub, owned by Google. So it's maybe a strong remedy, but this administration has been tough.
And the idea that, I guess, you could build on that to build competitors to Google in various ways. And it might end up to be a useful thing. A milder version of that would maybe include asking Google to license the search engine-related stuff. Maybe someone could build another Google. I don't know if that would work that well basically because it's a lot more than just the code that matters.
The other thing that has been discussed is the idea that the remedy in the Google case should be more forward looking as opposed to back. The competition for search using traditional technology is maybe over. And the question is, can you have a copyright panel or entry without discussing AI? That AI and artificial intelligence is more relevant.
So is there some way that Google-- I've seen proposals for the Google be forced to license or make available its AI-- related facilities. There's a couple problems with this. First, the valuable stuff for training is-- a lot of it is-- or that Google has, a lot of it is data, not-- or the valuable stuff going into making its products, some of it's data, which may be not subject to copyright.
And the degree it is subject to copyright probably belongs to somebody else other than Google. So I'm not really sure how that quite would work out. But I think it's worth discussing in a period where people are discussing remedies, whether copyright licensing of some form might form part of the antitrust remedy for the Google case.
It's a scenario that one might say needs more research, as we academics always say. Thank you again for inviting me to this panel. I look forward to your discussions and questions.
[APPLAUSE]
MODERATOR: All right. Thank you all. So we have about 12 minutes. This time, instead of my asking questions-- and I know that people didn't get a full chance to ask questions at the end of the last panel-- why don't I begin with seeing if there are any hands that want to go up and ask questions? And then I'll interject mine in the process. Yes, go ahead.
AUDIENCE: Thanks. Professor Axhamm, I hope I pronounced that correctly-- I'm asking this question based upon the case pending here in the Southern District of New York, New York Times versus Microsoft and OpenAI. I assume you're familiar with it.
So my question is, based upon the presentation you made, suppose, hypothetically speaking, that there was an AI company in Sweden. And this company, hypothetically speaking, copied the entire database of Svenska Dagbladet. What would be the result that you think what would happen in your country?
JOHAN: Yes. It would depend on the purpose of the copying. If we are talking about the new exception or limitation for text and data mining that I mentioned, then it could be that reproduction could be covered by that. But say, a complete copying, a permanent digital copying, that would constitute, well--
[INTERPOSING VOICES]
AUDIENCE: --exactly the same use as alleged in the pending lawsuit here in New York, in other words, the former large language module for the purpose of creating an artificial intelligence program.
JOHAN: Yes. But would you say that it would constitute the definition that I had on text and data mining. I'm not familiar with the details of this case that you are referring to, but it sounds as if reproduction is for a longer period of time than just for text and data mining.
AUDIENCE: Yes, it would appear to be so.
JOHAN: Yes.
MODERATOR: So it would be-- I think he's saying it would be infringement under Swedish law. But if I could just ask you a question relating it back to ECL, Johan, which is one of the things that you talked about, obviously, is the ECL having an opt out option. I'm wondering if you could tell us, as an empirical matter, how much opt out is there?
Have there been any studies that have actually indicated proclivities for opting out? Do we have a sense of what kind of categories do opt outs come in more frequently? Just anything on the empirics of how it works.
JOHAN: It's extremely rare. There are, say, anecdotal stories about Andrew Lloyd Webber opting out of the ECL for primary broadcasting. But then when he realized that then they will-- Swedish Public Radio will not play his music, he-- because they don't --
TIM WU: I'm not sure.
JOHAN: The Swedish Public Radio, they don't want to have an individual license with Andrew Lloyd Webber, so then they didn't play this music. So extremely rare. When this question is put forward to Swedish collective management organizations, they say no opt outs in practice.
So the opt out is there. One could say that it puts a pressure on the organization to negotiate licensing conditions that are acceptable also for the ones who are covered by the extended effect. So the potentiality for opt out puts pressure on the organization to when it negotiates the license. But there are this Andrew Lloyd Webber situation, but otherwise, no opt outs.
MODERATOR: Roy.
ROY: Hi. Roy Kaufman, CCC. I'm not an antitrust lawyer, but as I approach some of this, a lot of our talk here, the historical cases, what's gone on in music licensing is about a horizontal agglomeration of creators, musicians, whatever. My understanding of antitrust law at the meta-- sorry, I shouldn't say meta-- at the macro level is that it's about power imbalance.
And when I look at, so who's in AI using stuff without consent? And what is our antitrust concern? Is it that the creators are going to get together and try to negotiate something? Or is it that companies, like Microsoft, with a $3 trillion market cap have decided, hey, we're big enough that we can infringe on everything and just fight it out?
So where is the antitrust dynamic there? What should we be looking at in terms of antitrust law? Should we be worrying about the creators trying to negotiate? Or should we be worried about the incredible monopoly power of Microsoft and its semisubsidiary, almost-nonprofit OpenAI company? Where should we be actually looking at this at a policy macro level?
MODERATOR: Tim.
TIM WU: I'll speak since I don't have to speak for the administration.
[LAUGHTER]
I'll save the FTC this question. I mean, I think the Justice Department and FTC are not particularly focused on market power coming from copyright when they're concerned about the tech platforms and their monopoly power. I think they are-- like in the Google case, they're worried about monopolization of search or other markets. So I don't know if I got it right.
When I think about a OpenAI, Justice, according to reports, is investigating the Microsoft-OpenAI relationship. And maybe they're thinking that it's turning into a merger or something along those lines. So I think that's where the main concern is.
When copyright has been an antitrust concern, which is relatively rare, I think it usually centers on the idea that so a dominant monopolist is going to, like any sort of essential resource, kind of, I guess, grab a hold of all of the essential input for a certain industry.
So a little bit like a US steel company getting the only iron ore mine, for example, and then being able to foreclose competitors, I guess the concern would be-- and like, back in the old days, there was only one monopolist record player company. And they somehow have exclusive deals with every single. Then you can't have other record companies starting. That's along the kind of lines. But I don't think it's a center of the administration. I think they're worried about the core platform power of the tech platforms.
MODERATOR: David.
JUDGE STRICKLER: Yes. I'd like to answer that question with reference to the--
MODERATOR: Take that mic.
JUDGE STRICKLER: I'm sorry. Yes. Here we go. I want to answer that question as it relates to potential statutory compulsory licensing, which either through an antitrust consent decree or through statutory initiative that creates a compulsory license might work. As a judge, in my role as a judge, I don't have any opinion on that.
As a-- nothing that would be within our jurisdiction now but nothing that I could say. As an adjunct professor, I teach at least one of the classes each year with regard to artificial intelligence. And in conjunction with that, I've looked at some of the comments that have been filed by licensors and licensees of music in connection with-- this is in response to Copyright Office inquiries.
And the one thing they all agree on, in no event should there be statutory compulsory licensing. Now, why do they both agree on that? The dynamics that I mentioned in my presentation would apply equally in that situation. If you own the copyrights, you have this great must-have power.
And if you have, this-must have power, why would you want compulsory licensing? You'd want to take advantage of your power. That's what businesses do. That's what they're supposed to do. That's not a criticism. That's an explanation of reality.
The AI companies say, no, we don't want compulsory licensing because, in fact, we want no licensing at all. We say there's a fair use exception. There's a transformative quality to what we do. And we shouldn't have to pay anything, so why would we license in a compulsory way or in any way if there's no property right involved here at all?
So should the answer be a consent decree in antitrust or an outright statutory conclusion that there should be compulsory licensing? Is one argument in favor of that? The fact that both sides don't want it, at least as of now, maybe, but I don't know.
AUDIENCE: Both sides don't want it because they don't know how the cases are going to come down. One side will want it very soon.
JUDGE STRICKLER: It is a game theory type of problem, no doubt.
MODERATOR: I know. I'm guessing you're not going to-- OK, all right.
[LAUGHTER]
Yes, David --
AUDIENCE: Can the panelists speak a little bit to the question of unionization? Coming back to professor Ginsburg's question at the beginning, which is, what about the authors? It's one thing to say, even Apple and Google and Amazon need protection from three record labels that have 90% market share. It's another thing to say they need protection from individual songwriters.
And so traditionally, songwriters, playwrights can't unionize because they keep their copyright, whereas film writers, television writers can unionize because they give up their copyright, which essentially makes copyright almost worthless to those that are not giving up their copyright. So just as a hypothetical, suppose there was a performing rights organization made up only of songwriters and not music publishers, does the analysis change?
JUDGE STRICKLER: You want to give it a try.
TIM WU: You go first. You go first.
MODERATOR: David, go ahead.
JUDGE STRICKLER: Thank you. I don't think-- certainly from the economic point of view, I don't think the analysis changes. The question, I suppose, is, why have songwriters given up any rights they've had or assigned any rights to music publishers? Presumably, because these are voluntary transactions, it's because the music publishers can do this better. And if they can do it better, then it's the power of the music publishers.
Now, if the songwriters decided to band together and do this, presumably songwriters don't have the skill, the commercial scale, the business skills. Their great skill is song writing, lyrics, and music. So they would have to contract out or bring something in-house and basically create the same type of structure that the music publishers have, and that would give them a power similar to what the music publishers have.
Am I understanding your hypothetical correctly? You're talking about music songwriters organizing on their own. Those are the authors you're referring to, right?
AUDIENCE: Presumably they could get together and hire a lawyer to--
JUDGE STRICKLER: A lawyer in business--
MODERATOR: : I take it David's asking a question more about the legal ramifications rather than the economics of it.
AUDIENCE: Yeah, because the question is, do individual songwriters really have market power over somebody like Apple and Google? It's one thing in the 1940s to say mom and pop movie theaters, which virtually don't exist anymore. We have five or six chains. We don't really have mom and pop movie theaters anymore. So the question is, do the music users really need protection from individual artists negotiating collectively?
JUDGE STRICKLER: Well, I think anybody who negotiates collectively creates a power for themselves. But if the parties they're negotiating against have their own power, well, that's the countervailing power I spoke of before. And those will, in some sense, neutralize one another.
So we're looking at power-- disproportionate power as a concern. If we get proportionate power, I don't think there's a particular cause for concern, which is what animated the Web V decision that I talked about earlier.
TIM WU: Can I just say a quick comment? There'a a question-- and interesting. And I guess, I think it's a difficult one for antitrust. It's the only thing-- this is a very general answer, but I think it's very important in these policy discussions to be very realistic about the economic power and where it's held in our current markets.
And most of it's with the platforms. I would say, I agree, that let's just face it. And the challenge is some of the setups slightly presume differently. But I think the reality is that whatever power rightsholders may have, the tech platform power far exceeds it by orders of magnitude. And I think we need to be a little more realistic about that.
MODERATOR: All right. We are at the end of our time, so please join me in thanking our panelists.
[APPLAUSE]
And I'm going to hand the podium back to Kate to close things up.
INTRO SPEAKER: So quickly, I just wanted to say thank you all so much for attending our annual symposium. I hope you learned something and enjoyed your time with us. We'd love to have you back. And a sincere thank you and hats off to our tremendous panelists, both here and in the audience, that shared their time and expertise with us today. Thank you.
[APPLAUSE]
And I'd also like to thank the leadership and the staff of the Columbia Journal of Law and the Arts, who you've seen throughout today. And we'll be publishing a symposium issue covering today's events. So I'd like to thank them for their work and, of course, our wonderful staff here at the Kernochan Center and Columbia Law, including Samara Weiss, who is the woman behind the curtain that made all of this happen today. So thank you, Samara.
[APPLAUSE]
And of course, our amazing facilities team, AV, and catering who just made this a wonderful day. so thank you all so much. My final CLE reminder, please sign out, and then have a wonderful weekend. Thank you again, everybody.
[APPLAUSE]
MODERATOR: All right. Our final panel of the day is going to be looking at antitrust concerns around collective licensing. Our speakers are Anu Sawkar from the Federal Trade Commission, Professor Johan Axhamn from the University of Lund, Judge David Strickler from the Copyright Royalty Board, and my colleague here at Columbia Law School, Professor Tim Wu. So we will get things started. Our format will be similar to the prior panel. Each speaker will speak for about 13 to 15 minutes, after which I will ask a series of questions. So to kick things off, Anu.
ANU: Thank you so very much for inviting me to participate in this important conversation today. Before I begin, I should note that the views I express today are my own. And they don't necessarily reflect the views of the commission or of any particular commissioner. I'm sorry.
So I want to use the next few minutes to share a couple of stories. The first is my personal story, so you can better understand the perspectives I'm bringing to our conversation. And the second is the story of the American music performing rights organizations and how they came to be subject to an antitrust consent decree regulating their behavior.
After my clerkships, I joined the United States Department of Justice's Antitrust Division as an honors attorney. And my first home there was the enforcement shop that covered many content creating industries. I worked on numerous book publisher mergers and broadcast television station mergers. I was also part of the team that brought the successful e-books price fixing case against Apple and the major book publishers.
You might remember this. Amazon was trying to enter the e-book space. Amazon was pricing New York Times bestsellers at 9.99, I think, and the publishers freaked out. And we alleged that Apple led a hub and spoke price fixing conspiracy to try and raise the prices of e-books. [COUGHS] Excuse me.
After that litigation, I ended up spending a lot of time working on music licensing issues, including working on the antitrust division's 2014 review of the ASCAP and BMI consent decrees, and I'll speak more about that in a minute. And I also joined the Copyright Office's music listening tour, as the office was working on a music licensing study in 2015. I eventually moved over to the Antitrust Division sister agency, the United States Federal Trade Commission. The FTC has a dual mission to encourage competition and to protect consumers. FTC shares antitrust enforcement duties with the United States, and-- [COUGHS] excuse me-- with the Antitrust Division of the Department of Justice. And the two agencies collaborate all the time.
As you probably noticed, both agencies are doing a lot of work in the AI space these days. And last year I worked on around table about AI and the creative economy, which elevated concerns from a particular subset of creators, including authors, musicians, artists, and actors. The FTC submitted comments to the Copyright Office's notice of inquiry on copyright and AI. That comment highlighted both the competition and consumer protection law have a role to play in the conversation we're having today.
One of the questions the Copyright Office asked in its notice was about the feasibility of licensing copyrighted works for use as training data, which was a topic we explored while working on the roundtable. Given all the time I've spent working on music collective licensing issues, I was, of course, intrigued about what role direct licensing and collective licensing can play in the AI ecosystem.
What lessons have we learned from our past experiences and how should we use those lessons to inform the next step we take as policymakers and competition enforcers? I think this is a timely and important conversation. And I'm really glad that we're all having it together today.
The second story that I want to tell starts in 1914, which was the year the American Society of Composers, Authors, and Publishers, which we call ASCAP, was founded. Also happens to be the year that the Federal Trade Commission was founded. And we just celebrated our 110th birthday this week. So happy birthday, FTC.
[LAUGHTER]
Let's see. So ASCAP's mission was to create a blanket licensing system for public venues, like restaurants, bars, and performance venues that play music for their customers. You can imagine that it was very hard for composers to monitor the use of their works as they were being performed all across the country in public spaces. So those venues weren't obtaining direct licenses from the music owners for the individual songs. And the composers weren't able to police the infringement activities that were happening nationwide. So a blanket license solves and simplifies all these problems for both the venues and for the owners of the compositions.
The venues can get a single license, and that license lets them immediately perform anything that's in the repertory. The owners of the copyrights are able to monetize and polices the copyrights, collects the fees, and distributes to the owners. ASCAP was the first PRO, performance rights organization, in the United States, and for over 30 years, it was the only PRO.
Selling blanket licenses to radio stations quickly became a significant source of royalties for ASCAP and its-- for ASCAP members throughout the '20s and '30s. So, for example, in 1932, ASCAP set the radio broadcast licensing fee to 5% of a radio station's ad revenue. As radio profits increased, ASCAP members argued that radio was depressing sales for physical copies of musical works and sought to increase revenues from public performance royalties. I feel like that's sort of a theme that we keep hearing over time in this space.
The radio stations became more and more frustrated about how the cost of the blanket license was increasing. And in 1939, they retaliated. The National Association of Broadcasters decided to start a competing PRO called BMI, Broadcast Music, incorporated. BMI was created to compete with ASCAP, and it was intended to open up the competitive licensing landscape and provide competitive pressure on ASCAP royalty rates. BMI quickly attracted members in newer music genres, including country and the blues, That couldn't join ASCAP.
In 1948, the ASCAP announced that it was going to demand a 15% royalty rate. That's a big increase over a short period of time. The radio stations responded by boycotting music from ASCAP members. The stations were actually able to avoid licensing from ASCAP, which had a large portion of the music, by playing music in the public domain and playing music from BMI. So at least at that time, ASCAP and BMI were basically able to function as substitutes-- and as a competition lawyer, that makes me happy-- at least for radio stations.
And this is where the Department of Justice's Antitrust Division first entered the scene. DOJ began to investigate allegations of anti-competitive conduct by ASCAP and eventually filed complaints against both ASCAP and BMI in 1941. DOJ charged that the blanket license was an illegal restraint of trade and that arbitrary prices were being charged as a result of an illegal copyright pool. In antitrust speak, DOJ was basically saying that ASCAP and BMI were price fixing and were unlawfully exercising market power acquired through the aggregation of public performance rights in violation of Section 1 of the Sherman Act.
I think this is a good time to pause for a moment and give a quick primer on the relevant antitrust laws. So the statute we'll be focusing on today is the Sherman Act. Section 1 of the Sherman Act prohibits agreements that restrain trade. The courts have interpreted the statute to prohibit unreasonable agreements, not just any agreement.
Section 2 of the Sherman Act focuses on monopolization and attempts and conspiracies to monopolize. And here the focus is on acting unreasonably to gain a monopoly position or to maintain that monopoly position. The FTC actually enforces the FTC Act. But any conduct that violates the Sherman Act also violates the FTC Act.
There are three different frameworks that the courts will use to determine whether an agreement is an unreasonable restraint of trade under Section 1. So some types of agreements, including horizontal price fixing, are per se illegal. You just can't do them. These are the types of agreements between competitors that are so obviously anti-competitive that no further analysis is needed.
Most agreements, however, are analyzed under a flexible rule of reason, where we analyze whether the benefits of the agreement outweigh any competitive harm. There's a third intermediate category that some courts have used called a quick look test or an abbreviated rule of reason. They may use that when the anticompetitive effect is clear but the restraint doesn't fall into that bucket of things that we know are per se offenses.
So under the rule of reason, we typically define a relevant market. We determine market power. And then we look for the existence of anticompetitive effects. If we are able to show anticompetitive effects, the burden is going to shift to the defendant to show that their conduct is procompetitive.
For example, with the PROs, there's a huge efficiency in transaction costs and benefits to the users. If the defendant is able to show procompetitive benefits, the burden shifts back. And we need to show that there you can achieve those effects through a less competitive means. So the courts do an elaborate balancing after collecting a lot of information, and this is a very fact-intensive inquiry.
Beyond that legal framework, I want to note that the DOJ and the FTC have issued IP licensing guidelines, which explain the analytical framework that we use when we look at this stuff. The guidelines note that licensing arrangements are generally analyzed under the rule of reason. And we don't presume that IP rights convey market power. I think this is a theme that's come up.
Individual copyrights aren't generally going to convey market power because there are often substitutes available. But when we aggregate those copyrights, that can create significant market power. And because the blanket licenses provide for a single fee the right to play many separately owned works, they eliminate the competition that would otherwise exist among those works if they were being licensed directly.
And well, maybe at some point in the past, different PROs acted as substitutes, I don't think we can say that anymore. And for many users, like streaming platforms, the PROs are complements. You need to have everyone to have the coverage to operate in the market. So DOJ ultimately settled using consent decrees, which is basically a court-enforced contract that was meant or is meant to prevent the anticompetitive exercise of market power while, at the same time, preserving the efficiencies of the blanket license.
And I think people have talked about some of those important provisions. One, they can only do performance rights, can't expand into other things, so they can't leverage their market power. Two, only blanket licenses, you can't do single licenses or mixtures of licenses, so you can't leverage the market power from that. The licenses are nonexclusive to the PROs, super important.
If you look at the licensing guidelines, we really want to see that direct licensing is available outside of the pool to check the power of the pool. Also, you have to license everyone who wants a license, nondiscrimination. Similarly, situated licensees get the same rate. And of course, everything happens in the-- excuse me-- in the shadow of the rate court.
So the way we've consumed music has changed over time. We talked about how the formats have changed. And with that shift, we've seen that the math for the various stakeholders has changed, too.
In the 1930s, we saw copyright owners were concerned that broadcast radio was depressing their sales of physical copies of music. But by 2014, we were deep into the transition into streaming, and we were hearing very similar concerns. So the music publishers thought that they would be able to obtain higher rates from the streaming services if they negotiated direct licenses outside of the PRO.
So they petitioned the DOJ to open up a review of the consent decrees with the intent that we would allow partial withdrawal of the rights so they could pull the streaming rights, independently negotiate that, but leave the blanket licenses for brick and mortars in place for efficiency reasons. So the overall purpose was to try and increase the licensing rates through direct licensing. We can talk about this more, but that consent decree review process ultimately left the decrees in place without modifications.
And our closing statements concluded that, although stakeholders on all sides of the conversation have raised concerns with the status quo, our investigation confirmed that the current system has well-served music creators and music users for decades and should remain intact. In 2019, we opened another degree review. And at that time, the administration was undertaking a massive project to review almost 1,300 legacy consent decrees.
So they came up again. And in 2021, we left it-- well, I'd left the agency at this point, so I was not involved. But DOJ left the consent decrees in place, again, noting that there is significant reliance on the decrees within the licensing community. Throughout the division's investigation, many licensees expressed that their view that the decrees were largely working.
So why have these lasted so long? I think because our market power concerns still remain and the consent decrees continue to offer important safeguards to prevent the exercise of that market power. So I look forward to continuing our conversation today about how licensing can be used in a procompetitive manner to save AI.
[LAUGHS]
Just kidding.
Thank you.
[APPLAUSE]
MODERATOR: OK, next we have Johan Axhamm, who's going to--
JOHAN: Thank you very much, an thank you very much for the invitation. I'm very honored to be here. I was a visiting scholar a few years ago. And it has also been a privilege to listen to the very long discussions that we have heard before lunch and also after lunch.
I am a professor at Lund University. Lund is a minor-- it's a small city in Sweden with a long history. The University was established in 1666.
I also have a background as a special government advisor to the Minister of Justice in Sweden. So I have, for example, negotiated in Brussels EU directives, including one directive which I will refer to in my presentation, which is the EU Directive on Collective Rights Management. I also have experience from representing the Swedish government in WIPO negotiations, most recently the diplomatic conference in May on genetic resources related to patent applications.
This is the outline of my presentation. I will begin talking a little bit about, say, machine learning from a copyright perspective. And it concerns the scope of the right of reproduction, when do we at all have reproduction, and then talk about-- which might be considered the main part of my presentation-- the so-called extended collective licensing model.
It's a type of licensing arrangement that was established in Sweden and the other Nordic countries in the 1950s. And it has been used and is still in use in many areas of so-called mass use. That is to say, where there is a large amount of works and a large amount of right holders, it might be difficult to foresee beforehand exactly what will be used and to what extent.
And then I'll also talk a little bit about antitrust concerns and this directive that I mentioned. This directive is, to a large, extent based on EU case law on antitrust related to collective management organizations. So that directive is, one could say, a codification of EU case law on abuse of dominant position when it comes to collective management organizations.
The right of reproduction, it's not entirely clear the scope, when do we have a reproduction. In one of the major EU directives on copyright is called Infosoc Directive from 2001, includes a definition of reproduction, which includes not only permanent reproductions but also temporary. And this means that, to the extent that machine learning entails a temporary reproduction, then it's something that the rightholders as a starting point can control.
And this is not clear, or you cannot-- this does not follow from the Berne Convention, and it's not entirely clear either from the WIPO Copyright Treaty. This is actually one of the areas where I think it was a vote during the diplomatic conference in '96 on the agreed statement to Article 1(4) in the WIPO Copyright Treaty concerning the scope of right of reproduction in the digital environment, where the United States and the then European communities had different views.
And the European community wanted to make it clear that the right of reproduction at international level covers also temporary forms of reproduction. As far as I understand, the US did not want to have this explicit reference. So if you read the agreed statement or Article 1(4), you will see that it's not entirely clear at international level that the right of reproduction covers temporary forms of reproduction. But within the EU, we have settled this with the Infosoc Directive.
Some years ago, the European Commission initiated an update to the EU copyright framework to, say, make it more fit for the digital environment. And for example, the European Commission proposed what was than to be the Article III in the adopted version of the directive, an exception or limitation for text and data mining for the purposes of scientific research.
However, during the negotiations, there was a strong push from certain stakeholders to introduce a more general exceptional limitation for text and data mining. And that's the one that you find in Article IV. And what's the difference between these two? One difference is that Article 3 is specific, so it only covers specific users, that is to say, for in the scientific area, whereas article 4 covers all types of text and data mining. Or it's not limited to specific users.
In addition to that, Article 4 opens up the possibility for rightholders to opt out. But the starting point is that, to the extent that the rightholders have not opted out, they are, say, covered by this exception or limitation for text and data mining. So their works can be used for text and data mining unless they opt out.
So the article 4, say, turns the presumption around and says that, unless you opt out, your work can be used for text and data mining purposes. But it's not entirely clear if you compare the definition of text and data mining with, say-- and this was also touched upon before lunch. Is text and data mining the same as machine learning?
It's not entirely clear. You have a definition there of text and data mining in Article 2(2). And this was also mentioned before lunch that, in the recently adopted EU regulation on artificial intelligence, there are provisions on transparency, meaning that the one who carries out text and data mining or operations in relation to copyright protected works must be transparent but not on a work-by-work basis but on a more general basis.
But this of course, gives the right holders the possibility to know whether their work has been used as or part of this text and data mining process. These transparency provisions in the AI regulation has caused some concerns for rights holders because it implies, more or less, that text and data mining is the same as machine learning, so some concerns from the perspective of rights holders when it comes to that, even if it's, say, positive that we have a transparency provision.
When it comes to extended collective licensing, sometimes, say, copyright is discussed in a quite, say, binary perspective. So we have exclusive rights. And then we have exceptions and limitations. But at least in countries with a long tradition of collective rights management, we deem collective rights management to be, say, something in between, so a fair balance between individual management, say, individual licensing, and exceptions and limitations.
Exceptions and limitations, as far as I understand, it would include provisions on compulsory licensing and statutory licensing. So rightholders holders have no control over, say, the permission or giving a license. They must provide-- or accept use, whereas collective management, there is normally an element of control, at least for the collective management organization.
When have we introduced provisions on extended collective licensing? In Sweden, we have had it since the 1950s, and this is the same for the other Nordic countries, Finland, Denmark, Norway, and Iceland. And I know that other countries in the European Union have similar, say, arrangements for the management of rights, sometimes referred to something else or something similar.
But say, the legal effect of an extended collective licensing is that the agreement between the collective management organization by provision in law is extended to cover all right holders who are not members of the organization or who has not given their mandate to be represented by the organization. The provisions, the statutory provisions, they also give rightholders the-- so the ones who are covered by this extended effect-- the possibility to opt out. So I don't want to be part of this extended effect.
To the effect that right holders are covered by the extended effect, they have a right to equal treatment when it comes to, say, remuneration. And according to Article 12 in this EU directive from 2019, there is also an obligation to inform rightholders that there is this license with an extended effect.
This article 12, say, legitimizes the extended collective licensing model. So in EU law, it's now an accepted form of what one could say an arrangement concerning the management of rights. So the extended collective licensing model is not an exception or a limitation but rather an arrangement concerning the management of rights.
But you can only use this model in situations where it's practically impossible to give an individual license or individual permission. So one can say that we use extended collective licensing in these mass-use situations.
And examples of that could be prime-- so user works for primary broadcasting, use of works for educational purposes in, say, universities and schools. It could be use of works in archives and libraries. And it could also be, say, for reprography and similar situations. These are situations where I gather that the provision of fair use in US copyright law would cover some of these uses but maybe not all of them.
Yes, two minutes. This is actually one of the hot topics when it comes to copyright in the European Union at the moment. Rightholders are not, say, too comfortable with this article 4 on the exception limitation for text and data mining in this directive from 2019. So there is a discussion on, say, can we standardize-- and so they opt out of that exception-- and limitation and then meet the market demand with an extended collective license.
And what would be the purpose of that? The purpose is, of course, that then the right holders can control the extent of the use and also be remunerated for the use because, otherwise, article 4 being an exceptional limitation, then its use with no remuneration to the rightholders. But this, of course, is preconditioned in practice that there are collective management organizations that are ready and available to provide a license and that this is also something that is accepted by the market.
And I will touch upon this on the antitrust concerns. We have a provision in Article 102 in the Treaty of the Functioning of the European Union on prohibition, on abuse of a dominant position, that is. And we have quite a lot of case law related to that article as applied to collective management organizations concerning licensing conditions, so the, say, the relationship between the CMO and users on, say, how tariffs should be calculated, et cetera.
There are also case law concerning the scope of the mandate that the organization can require, or when it comes to rightholders, should it be all territories and all users and all works? Or should it be more, say, limited in scope? And we also have case law related to the relationship between collective management organizations.
In general, and EU law makes a distinction between, say, the existence and the exercise, so the existence of the ECL is not put into question but rather the exercise. And so when an organization is exercising an extended collective license, it must be in line with this case law on abuse of a dominant position.
And it's now, say, simplified in practice because this EU directive on collective rights management codifies this case law. And I will provide some examples of this.
But first of all, so to make it clear, I got a question from the organizers. So the directive includes a definition of what is meant by a collective management organization. And that definition makes clear that a collective management, it doesn't matter whether the mandate to represent rightholders is based on, say, an explicit mandate or whether it's something that is provided by law. So this reference to by law here includes extended collective licensing, the extended effect, as I mentioned.
And this is the last slide. One example, licensing conditions, so this is where-- so the directive codifies this case law. It says that when collective management organizations and users negotiate, they should conduct these negotiations in good faith.
Licensing terms shall be based on objective and nondiscriminatory criteria. And tariffs for exclusive rights and rights to remuneration shall be reasonable in relation to, for example, the economic value of the use of the rights in trade, taking into account the nature and scope of the use of the work and other subject matter. This is almost word by word from case law in this area.
To sum up, could we have an extended collective license for machine learning? Yes. But we must have a simplified mechanism for writers to opt out from this article 4 on text and data mining. And there must be market conditions when it comes to the structure and, say, the acceptance of collective rights management. Thank you.
[APPLAUSE]
MODERATOR: Do you want to speak from here, or?
JUDGE STRICKLER: No. I'm going to go up there. I thought you were going to go up there.
MODERATOR: No, no, no. I'm just going to turn off the slide.
JUDGE STRICKLER: OK. Good afternoon. Thank you for the privilege of having me speak here today. It's been enjoyable to listen to the other panelists, and I'm glad to make my contribution.
As I think has been made known from some questions and answers previously, I am the economics judge on the US Copyright Royalty Board, and we preside over trials, which, among other things, set royalty rates for sound recordings, also the underlying musical works. Now, earlier in the panel before us, my colleague Judge Rui went through some of the licenses, all of the licenses, really, quite comprehensively with regard to those licenses for which we set royalties.
I want to get under the hood a little bit in that regard by explaining how we have addressed those issues and how we have set royalties in actual cases, which is not to say that it will always be that way in future cases. Every case is decided on its facts. But economic principles tend to have a level of constancy to them. So to the extent they apply in one case, they may well apply in another case. But the facts may show that there are new economic concepts that didn't need to be addressed or were not addressed in a previous case, which you'll see as we go forward this afternoon.
So my focus today is from a more general sense to talk about how economic regulation or price regulation in lieu of what one might call traditional antitrust relief and remedies, makes sense as an adjunct to that more traditional antitrust concern, particularly as it relates to oligopolies in particular-- and I'll explain the terms as I go through-- complementary oligopolies and the presence of potential tacit collusion as opposed to explicit collusion among oligopolists. These are things that, to the extent they may be economically deleterious, are not well suited for antitrust relief and remedies because antitrust law, as opposed to antitrust economics, if you will, is more focused on conduct and behavior rather than necessarily just the economic result.
That is to say, high prices do not mean that there's a monopoly problem or a problem for antitrust or the existence of tacit collusion can either be proven or, even if proven, is worthy of a remedy. But separate and apart from antitrust, there may be an appropriate place for economic regulation.
And this is a really old concept. This concept actually goes back at least as far as 1968, '69, when Richard Posner, before he was even a judge on the Seventh Circuit, wrote an article where he spoke of the fact that antitrust is not well suited for these concepts, and it would make more sense to either just to perhaps-- he buried this. I'll use the word bury. He certainly did not-- in a footnote, the idea that there might be price regulation as an alternative, rather than antitrust activity in that regard.
And that concept is what animates the economics of what we do. So that's my general focus, to talk about economic regulation as an adjunct to antitrust. And I want to do it through the focus or through the lens of our CRB rate regulation with regard to two particular cases, Web IV which and Web V now. All jargon, and jargon is going to be lost on people who are not familiar with the jargon, so allow me to explain.
Web means webcasting. So webcasters or Webcasting IV was the fourth iteration of the case law, the cases that developed. Webcasting V followed after it.
When you hear webcasting. Think internet radio. Think, for example, Pandora. One of the other panelists, Mr. Marks, has represented Pandora before us, pursuant to these types of rate regulations.
And sometimes I wonder why there's so much jargon when we have three different names for the same exact concept. And the statute, internet radio is noninteractive. It's also considered, as I say, internet radio. It's also sometimes called, somewhat inaccurately, lean back radio. So there's a lot of synonyms, which I think are-- I don't know if they're meant to confuse, but they have the effect, very often, of confusing people.
Anyway, so when we set the Royalty rates for internet radio, we have a statutory standard that we must apply. And that's the willing buyer, willing seller, marketplace standard. That's a paraphrase, but that's the shorthand version of it.
Now when the judges have set such a standard, particularly developed in Web IV, again, in Web V, we've set what we consider to be-- got from the evidence and the expert testimony in the case an effectively competitive standard. We give you another synonym for that-- workably competitive.
Economists have long called it workably competitive. Lawyers have called it effectively competitive. It's the same thing. It basically means meaningful price competition, and I'll get into that in a little bit as well.
And I'll do that by explaining the standards and how we applied that standard in these two cases. And let me tell you something else in case you don't know this-- and many of you may well know this, but it helps to understand the cases.
When it comes to noninteractive-- I'll use that purposely, that statutory phrase for webcasting. The way it works is, if you have a Pandora account, you seed a station, S-E-E-D, seed a station with a song or with an artist. And then the algorithm takes over, and the algorithm does the work, trying to pick out the song that it thinks would be next, most likely-- that you would most likely want to hear based on your listening habits or based on that particular music and moving forward through the songs in that order.
And there are restrictions as to what can be played. You can't repeat one artist too often. You can't repeat a song too often. These are called the sound recording performance complement is the jargon in the statute, to limit the amount of repetition that can be done.
And all that's important backdrop, not just for the sake of letting you know how that works, if you didn't. But it's important backdrop for how effective competition can be created and was created according to the judges in Web V and Web V.
So Web IV was in 2016. And I'm happy to say that the DC Circuit affirmed our decision in 2018. And Web V was decided in 2021. And the DC Circuit affirmed that in 2023.
And let me tell you, we all breathe a sigh of relief when they affirmed it, not because-- just because we're happy that they agreed that it was reasonable and not an abuse of discretion. But as Judge Rui pointed out, if they don't agree and they kick it back, it's a hell of a lot more work for us than we had anticipated, although the particular case that judge Rui I don't think mentioned specifically was a different case, the phono records license case, where I had dissented from the other two judges.
Judge Rui was not on the panel, by the way, back then. And I had dissented. And the DC Circuit essentially agreed with my dissent and kicked it back. That was quite a laborious process, proving the old adage that no good deed goes unpunished.
[LAUGHS]
So anyway, back to our story, Web IV, the evidence we relied on was twofold, and it was in the nature of benchmark rates. What we look at are representative contract rates or, arguably, representative contract rates that the parties put forward that say are analogous to the rate that you're trying to set, in this case, the noninteractive rate.
We had two types. One was the interactive rate, which is on-demand music. So on-demand music the relates to the royalty-- or to the music that's played by streaming services, where you can pick any music that you want at any time.
And the labels negotiate with the streaming services. Think Apple, Amazon, Google, Spotify, Deezer, and a host of relatively smaller ones in the United States. Those transactions are not regulated at all. Those are pure market transactions.
And if you're a laissez faire fan or a laissez Faire economist, you love that. You say that's the way the market is supposed to go. That's the way this should be done. Just let the market work. Get government out of this.
And sometimes when I talk to people and bore them at cocktail parties, when they ask me what I do and I do this, they say, I didn't even government was involved in such a thing. And then they go look for an interesting conversation.
[LAUGHTER]
So then that was one type of benchmark that we received in the case. The other one was rather unusual, and it was a rate that was not from a different market but was from this very market, from the noninteractive market itself. There were two contracts that were presented in evidence.
One was a contract that Pandora had with an entity known as Merlin, which is a consortium of some of the larger indie records, independent record labels, sometimes referred to as the fourth major because it's so big. And also there was a contract between iHeart, a noninteractive service, and Warner, one of the bona fide, big three record labels. And I'll talk about each piece of evidence in that regard, starting first with the interactive contract rates, again, interactive meaning on-demand music.
You need all of the music in order to be able to play. So when the label-- to have a service. So when the labels negotiate, they have this tremendous power. It doesn't do you any good if you're a noninteractive service to have an agreement with Warner and Sony but not Universal, which may have, for argument's sake, 30% of the market because, if you're missing 30% of the market, nobody's going to want to listen to your service because 30% of the time those Universal artists are the ones that you're going to want to hear and you won't get it. So you have a serious problem there.
So that doesn't make it a good-- really good benchmark for the noninteractive service because you have this competition problem. The major labels do not-- because of that power, because of that structure, do not need to negotiate price because they are going to "take it or leave it" mode. They have what's essential for you. And when somebody is offering you something that's essential to you, you really don't have much bargaining power at all.
So you have a competition problem. And in the economic jargon of this, what the labels are complementary oligopolists, not just your ordinary oligopolists. But they're offering complements that are perfect complements. You must have all three. So in economic terms, they're perfect complements, and therefore, there's a tremendous amount of power.
Now, if your goal is to set an effectively competitive rate, if you think that's what the statute requires, that's not going to be-- going to do the job for you. So somehow that has to be adjusted. And by the way, more than a parenthetical but before I turn to the other evidence again, how do the record-- what did the evidence show that the record companies would do to ensure that they didn't have price competition?
Well, the evidence showed that they set most-favored nation clauses in all their contracts. They had nonsteering clauses, antisteering clauses. And they acknowledged in testimony that they did not engage in price negotiation with the streaming services in this benchmark interactive market.
So it wasn't just theoretical that that's the case. It was actually acknowledged and supported by the evidence. So there needs to be-- if you're going to use that, which we did for, ultimately, for the subscription service of Pandora and iHeart or any noninteractive service, you needed to make some-- you needed to make some sort of adjustment.
But let's turn to the other evidence, the noninteractive rates, the Pandora, Merlin rates and the iHeart Warner rates. This was unusual because it came out of that very market, which we don't usually see. Pandora and Merlin and iHeart, Warner, the rates in these contracts were lower than the existing statutory rates.
Now, at first blush, that seems irrational. Why would a licensor who could get, say, I'll just use a number at random, 20 for its royalty be willing to take 15 instead when it could get 20? Well, if that was all there was to it, they wouldn't. But what was utilized in those contracts was a concept known in antitrust circles and in this particular circles as steering.
These companies, Pandora and iHeart, said, if you-- I'll just use percentages. That may not track the case but for purposes of elucidation. Pandora would say, we'll give you-- we'll drop the rate by 12% to you, but we'll give you 15% more plays. So you will drop the price, but we'll increase the quantity.
So why wouldn't any company take that? They're going to get more revenue that way because it's a higher percentage increase in quantity than there is a decrease in price. For those of you fans keeping score economically at home, that's an elasticity concept. And that concept animated the self-interest of Merlin and in its contract with Pandora and Warner in its contract with iHeart, which was similar in economic effect.
So that was the reason why you would actually see rates that were lower. Now, how could Pandora do that? How could iHeart do that? Well, they had the technological capability because of steering to be able to do that. So that, as I said, it wasn't just to give background on how these noninteractive services work.
Once the algorithm takes over, it will rank the songs that it thinks you want to hear, and those songs will be played. But these noninteractive companies injected into their algorithm a pricing function or a pricing variable to say that if, for argument's sake, if it was the iHeart, Warner contract and the fifth song in the algorithm would have been a Universal song, the algorithm may drop that down to 15th, for argument's sake, and bring up to the fifth spot, the cheaper Warner one, so that ultimately the Warner songs would be played more often, reducing the overall royalty.
So Warner would get more plays, and it would be happy. iHeart would have to pay less revenue on those plays, so it would be happy. But to do that, it would have to substitute away from those that were insisting on the statutory rate, the 20. So those are the losers. There's no free lunch in economics.
The services benefited the steering approval of Warner and of Merlin meant that they were going to benefit as well because they got more quantity than they lost in price. But there's no free lunch. So the losers were the other labels that had to be steered against in order to make that happen because you can't increase the total number of plays.
That's a function of overall demand. It has nothing to do with the steering function. So that had to be done.
So what's interesting, curiously-- and maybe it's because I knew I was going to be giving this presentation today. When I was teaching earlier today-- I teach this stuff, for lack of a technical term. I teach this stuff both to NYU undergraduates in the Steinhardt Music Business Program and also in the law school today. This semester was the Steinhardt turn.
And I tried to-- and I asked the students. I said, well, what if you-- what should the record companies that are steered against, what should they do to? And I was met with silence, as I often am when I ask the class a question.
And they're thinking, professor, it's 8:30 in the morning. Don't do this to us on a Friday, no less. But I did something with them today. And this wasn't going to be part of my presentation, but it was so noteworthy to me because it's something I will say to people.
And I never really tried it in class. And I said to them, economic concepts immediately tend to intimidate a lot of people because they don't feel comfortable with it. They took an economics class in college.
They hated it. They hated their professor. And they don't want anything to do with it, so they're immediately frozen by it.
So I said, there's an easy way to deal with this. When you deal with it, be the person, be the person who's going to make money or lose money. Be the person who's responsible for making the label money or losing money and think that you're going to be called on the carpet by the CEO and says, we're losing money because they're steering against us.
What are we going to do? How are you going to fix this problem? What would you do?
After another short delay, a hand went up. And the student said to me, she said, she said, well, why don't we offer to steer as well? She was exactly right.
That's the point is that the threat of steering, when you know that that's out there, you're going to-- you, as a label, are going to say to Pandora or to iHeart, no, no, no. Don't steer against us. Don't steer against us.
We'll match it. We will match that lower price because otherwise we're going to lose revenue relative to where we otherwise would be if we just agreed and kept the old percentages, the old quantity percentages. Now that sounds esoteric, and steering sounds like a difficult concept.
But all that is price competition. It's just simple price competition. If you were a buyer for Macy's and you had to buy shoes that you were going to put on the floor, you play one seller off against the other, trying to get the lowest price you possibly could, and make that decision. And steering is no different. You're steering the business the way that you can.
And then I ask the students, I said, but if you wanted to do it differently, and it was somewhat nefarious, what would you do? I don't need that. I'm going to keep going.
And one student said, I would ask the other record companies to stop steering. I said, yeah, you could do that. That's the right answer.
But the problem is, if you go over and ask them and you ask them by telephone and it's being audiotaped by the Justice Department, you're going to go to jail for price fixing, which is exactly what happened in the Cargill case with regard to vitamin supplements in the agricultural field. It was actually, in that case, two oligopolists, one from Cargill I think it was, and I forget the other company.
One of them actually said on the audiotape, gee, I hope the-- I hope the FBI isn't listening in on this. And the FBI was listening in on it. So it wasn't just evidence. It was evidence of a confession.
So steering is price competition. And the judges recognize this. And we took the subscription rate, which had no steering involved, and we adjusted that down by 12% to reflect the value of steering. And in the ad-supported rate-- for the ad-supported rate, we just adopted the rates that were in those two benchmark agreements Pandora, Merlin, and iHeart, Warner.
Now, labels push back very hard on this. And they said, well, judge, judge, they can make all these abstract arguments what might happen in the hypothetical market. Anybody can make a hypothetical argument. But in the real world, this would never happen.
Well, why? They said, well, we would insist on most-favored nation clauses or antisteering clauses, or we would insist-- we would say we're going to withhold our entire catalogs unless you refuse to-- you agree not to steer. Or we'll require that you pay upfront royalties, take the historical average of plays, pay it all up front.
Then if you start steering and messing around with it, you're going to pay something. But you've already paid for everything, so you be paying twice for, in essence, for a certain percentage of plays. And our response in our decision was, yeah, we recognize you would do that.
But doing that is the use of the excess market power that we're trying to cure. So you're telling us we can stop you from making an effectively competitive market by doing things that are not effectively competitive. So that argument was rejected.
So is the last thing I want to say-- it'll be much, much briefer-- on the Web V case because it was the same economic logic, the same arguments being made, but one very important difference. And I was struck by the point Elliott made during his comments when his panel was up here, and he said that Apple is afraid of the majors. The majors are afraid of Apple. And that's exactly what animated the Web V case because, in Web V, the labels were-- the evidence clearly showed they acknowledged the tremendous power of Apple, Amazon, and Google.
And they were afraid of these tech rivals, who, if they had the markets to themselves, just the three of them, they could hold out. They could vertically integrate and start their own services. They could do a whole host of things with their tremendous power.
And as a consequence, it was important-- according to the evidence and as the decision said, it was important for the labels to keep Spotify going. They had to make sure that Spotify was there as a bulwark against these big companies. And they lowered their rate in the interactive market.
Again, this is the benchmark market. They lowered that rate for Spotify. And that took us a ways towards what otherwise would have been a 12% reduction. It did roughly half the work because it was a different power. It wasn't effective competition.
And this is where economics is a very large-- has a body of work. And if you can find the right economic analysis, you may well find the right answer. And in this case, it was a matter not of price competition but countervailing power. Spotify had a countervailing power that was thrust-- that it benefited from simply by not being part of one of these large ecosystem companies but by being an independent, sometimes some people would say pure play web caching, not so pure play anymore because they do podcasts and some other things as well.
But therefore, there was another power, another piece of economics generated both by the theory of-- economic theory of countervailing power, plus the evidence that showed, as one of the witnesses colorfully pointed out, for the record labels, big as they are, their entire business is a rounding error for Amazon, Apple, and Google. And so the countervailing power led to something lower.
And in that decision-- and I'll close with this-- we pointed out that it was this-- we had this interesting pitting of two different types of economic power, the economic power of the marketplace that the labels had as must have-- you must have the music versus the fear of what Apple and Amazon and Google would do if they used that power to squeeze Spotify to the point where it would leave the industry but Apple, Amazon, and Google could do pretty much whatever they wanted.
They could use music as a loss leader rather than an attempt to make money on this. And to make that point, and before I turn this over to Tim, we quoted specifically Tim and somebody who is 180 degrees counterpart to him, former FTC commissioner Joshua Wright. Tim had written in his book the Curse of Bigness about the distinction between market power and power derived from sheer corporate size.
And he talked about Apple. He wrote specifically in that book, just a handful of giants, Amazon, Google, and Apple, transcend the narrowly economic and that we pointed that out as a counterpoint to laissez faire economist, former FTC commissioner Joshua Wright, who criticized the emphasis on sheer corporate size as calling for nothing less than the complete dismantling of the consumer welfare standard and the consensus among antitrust practitioners, enforcers, and academics about how to promote competition.
Not so simple a position to take because the fact-- and that's the-- again, I'll close with the overall arching point. Price regulation by a regulator, particularly an administrative adjudicator like the Copyright Royalty Board or like the rate court, or also arbitrators can do the same thing on the CSAC, which is a small PRO and is not covered by the consent decrees, entered into an arbitration agreement after a private antitrust suit. And now their rates with licensees are subject to an arbitrator's decision.
So there's always a third party when you have market power. CSAC only has about 5% of the market, so it seemed to be implicitly suggested before they really don't have any particular market power. But they're a cream-skimming operation, and I mean that in the most positive sense.
They take high-value artists, give them a better return. And those high-power artists sign signed with them. And as a consequence, even they only have 5% of the market, they're a must have, too.
I have no idea offhand where Taylor-- who represents Taylor Swift, but if you only have 5% of the market and that includes Taylor Swift-- she's probably more than 5% of the market all by herself. But if you have Taylor Swift, you're a must have for anybody who wants to have a streaming service that has all the music.
So there's a question of bigness and power in the marketplace that are in tension with each other. And these types of cases are not solved very easily by ideologies and truisms. They're solved by detailed painstaking work, digging into the facts, ferreting out the economics, and coming to a determination. Thank you.
[APPLAUSE]
MODERATOR: Yeah. Take it away.
TIM WU: Thank you. Hi, everybody. Thanks for inviting me. I want to thank Kernochan Center and also students and Pippa and everyone who does all the work behind the scenes.
So my name is Tim Wu, and I teach copyright and antitrust, which seems to be why I'm on this panel, among other reasons. And I wanted to spend my time on a slightly related but slightly different topic, which is the prospect of using copyright licensing as a remedy in antitrust cases. I was thinking this is actually a slight-- a subject that hasn't in my knowledge gotten an enormous amount of attention but seems to be important.
This is usually-- and most antitrust, intellectual property attention focuses on patent. And the subject I'm going to talk about has gotten more attention in patent, but I'll mention a few ways in which it also comes to copyright. So I'm inspired in this topic by three particular cases or situations that inform it.
The first is the 1956 AT&T consent decree, which centered on a licensing remedy requiring the licensing of all of AT&T's over 8,000 patents, including the transistor patent. So intellectual property licensing has been used before. I also want to discuss the 1909 mechanical license, known well to this crowd, which in its own way was a settlement of an antitrust dispute or controversy.
And something very top of mind that prompts this discussion is the ongoing search for a remedy in the Google antitrust case, which is the subject of a lot of discussion right now, and the question of whether copyright licensing should possibly be a part of the Google antitrust remedy. And I think that's an open question that hopefully we can discuss. So let me talk about intellectual property licensing as an antitrust remedy, first through the 1956 consent decree.
This is a well-known topic among antitrust tech historians and practitioners but I think less in the copyright crowd. So in the early '50s, the Justice Department, which was then probably at its most aggressive-- I don't know. Now it's pretty aggressive too, but the '50s were pretty aggressive, so were the 1910s.
Anyway, they sued AT&T for monopolizing telephony and other things. And they wanted to break up AT&T but didn't. And instead, on the pain of breakup-- and there's a whole story in there-- AT&T agreed to two conditions. First, AT&T agreed to stay out of computing forever, which opened a lot of room for the development of a computing industry in the United States. That was independent telecom, a different Topic
But more relevant to this, to us, AT&T agreed to license over 8,000 patents, including its most valuable patent, the 1947 or '48 transistor patent, which was probably the most valuable patent of the midcentury. So I think most people who've studied this period think this was a successful remedy, partially because of the growth of computing in the United States but also the birth of the semiconductor industry in the United States, independent of Bell Labs.
What happened during that period, a lot of people quit Bell Labs and started their own companies. You have a Fairchild. You have Shockley. Eventually Intel ends up being born during this period. And the United States takes the lead.
There's been a few studies of the effects of the licensing, and they're very positive. So that's one model for when license can be-- copyright can be a remedy for antitrust cases. Another interesting model is the 1909 mechanical license for which this is obscure at some cocktail parties but not this crowd.
So I think copyright people think of that as a copyright thing. But if you've paid more attention, if you dig into that history, it was also motivated by antitrust issues. There was a concern at the time that there would be a monopolization of the, I guess, recorded music industry through a player piano company. I don't fully know how you pronounce her name, Aeolian, which had trying to do exclusive deals with all the publishers of music.
And so that and the Apollo decision led to the 1909 decision which created the compulsory license phonorecords, $0.02 a copy. Since then, it started being adjusted. So this, as people in this room know, the mechanical license turned out to be more popular than people thought.
And I think it was in 1961, the registrar proposed the abolition in the New Act, the 76 Act. They didn't know it would be 76 then. But people seemed to like it. And so the mechanical license survived the 76 Act.
And I don't know if I would-- know how popular it is, but at least it seems to have worked. And obviously, the compulsory license has been used for other areas as well. Now, that wasn't a remedy in an antitrust case, but it was kind of a congressional settlement of an industry dispute involving market power. Compulsory license has, obviously, been used since, often to settle disputes involving market power.
The cable license, compulsory license is a good example. It was of a situation where the concern was, combined with the broadcasting industry, combined with copyright holders who were loyal to the broadcast industry by collectively, to use a antitrust-- to use the antitrust word keep out or foreclose the operation of cable and the cable copyright, cable compulsory license has a lot to do with that.
As I recall, though, I can't find any sources for this. This is my third historic point. I think in the Microsoft antitrust case, there was at least some discussion of whether licensing Microsoft code should be part of the remedy. If someone remembers that better than me-- obviously, that's not what happened.
The Justice Department proposed ultimately a breakup along vertical lines or horizontal lines, depending how you describe it. And ultimately it was settled differently, which leads us to the Google case, which I'll introduce in case you're not an antitrust aficionado. So Justice Department sued the Google company for monopolization of general search, and there was a trial last year.
The core of the claim was that Google had basically paid Apple and other companies large sums of money. They ended up being over $20 billion. I think the total number paid per year last year was $26 billion or something like that. Google had paid this money to have them keep Google as a search engine and, therefore, excluded its competitor.
Justice also suggested that Google was paying off Apple to stay out of search with a $20 billion a year. And $20 billion a year is actually isn't chump change. I think it starts to become more of the entire record labels combined, so some big money.
So Google was found liable over the summer. And currently, the search is on for a remedy. And one of the things I think is interesting to think about is whether copyright licensing might somehow form part of that remedy. And let me throw out two ideas how. I don't know if they're any good or not.
But one little broader is to suggest that maybe Google should be asked to license its core code base as a remedy. It's a little strong. Google has about two billion lines of code at least, and they're overseen by a system called the Piper system. And it's a little bit like GitHub, owned by Google. So it's maybe a strong remedy, but this administration has been tough.
And the idea that, I guess, you could build on that to build competitors to Google in various ways. And it might end up to be a useful thing. A milder version of that would maybe include asking Google to license the search engine-related stuff. Maybe someone could build another Google. I don't know if that would work that well basically because it's a lot more than just the code that matters.
The other thing that has been discussed is the idea that the remedy in the Google case should be more forward looking as opposed to back. The competition for search using traditional technology is maybe over. And the question is, can you have a copyright panel or entry without discussing AI? That AI and artificial intelligence is more relevant.
So is there some way that Google-- I've seen proposals for the Google be forced to license or make available its AI-- related facilities. There's a couple problems with this. First, the valuable stuff for training is-- a lot of it is-- or that Google has, a lot of it is data, not-- or the valuable stuff going into making its products, some of it's data, which may be not subject to copyright.
And the degree it is subject to copyright probably belongs to somebody else other than Google. So I'm not really sure how that quite would work out. But I think it's worth discussing in a period where people are discussing remedies, whether copyright licensing of some form might form part of the antitrust remedy for the Google case.
It's a scenario that one might say needs more research, as we academics always say. Thank you again for inviting me to this panel. I look forward to your discussions and questions.
[APPLAUSE]
MODERATOR: All right. Thank you all. So we have about 12 minutes. This time, instead of my asking questions-- and I know that people didn't get a full chance to ask questions at the end of the last panel-- why don't I begin with seeing if there are any hands that want to go up and ask questions? And then I'll interject mine in the process. Yes, go ahead.
AUDIENCE: Thanks. Professor Axhamm, I hope I pronounced that correctly-- I'm asking this question based upon the case pending here in the Southern District of New York, New York Times versus Microsoft and OpenAI. I assume you're familiar with it.
So my question is, based upon the presentation you made, suppose, hypothetically speaking, that there was an AI company in Sweden. And this company, hypothetically speaking, copied the entire database of Svenska Dagbladet. What would be the result that you think what would happen in your country?
JOHAN: Yes. It would depend on the purpose of the copying. If we are talking about the new exception or limitation for text and data mining that I mentioned, then it could be that reproduction could be covered by that. But say, a complete copying, a permanent digital copying, that would constitute, well--
[INTERPOSING VOICES]
AUDIENCE: --exactly the same use as alleged in the pending lawsuit here in New York, in other words, the former large language module for the purpose of creating an artificial intelligence program.
JOHAN: Yes. But would you say that it would constitute the definition that I had on text and data mining. I'm not familiar with the details of this case that you are referring to, but it sounds as if reproduction is for a longer period of time than just for text and data mining.
AUDIENCE: Yes, it would appear to be so.
JOHAN: Yes.
MODERATOR: So it would be-- I think he's saying it would be infringement under Swedish law. But if I could just ask you a question relating it back to ECL, Johan, which is one of the things that you talked about, obviously, is the ECL having an opt out option. I'm wondering if you could tell us, as an empirical matter, how much opt out is there?
Have there been any studies that have actually indicated proclivities for opting out? Do we have a sense of what kind of categories do opt outs come in more frequently? Just anything on the empirics of how it works.
JOHAN: It's extremely rare. There are, say, anecdotal stories about Andrew Lloyd Webber opting out of the ECL for primary broadcasting. But then when he realized that then they will-- Swedish Public Radio will not play his music, he-- because they don't --
TIM WU: I'm not sure.
JOHAN: The Swedish Public Radio, they don't want to have an individual license with Andrew Lloyd Webber, so then they didn't play this music. So extremely rare. When this question is put forward to Swedish collective management organizations, they say no opt outs in practice.
So the opt out is there. One could say that it puts a pressure on the organization to negotiate licensing conditions that are acceptable also for the ones who are covered by the extended effect. So the potentiality for opt out puts pressure on the organization to when it negotiates the license. But there are this Andrew Lloyd Webber situation, but otherwise, no opt outs.
MODERATOR: Roy.
ROY: Hi. Roy Kaufman, CCC. I'm not an antitrust lawyer, but as I approach some of this, a lot of our talk here, the historical cases, what's gone on in music licensing is about a horizontal agglomeration of creators, musicians, whatever. My understanding of antitrust law at the meta-- sorry, I shouldn't say meta-- at the macro level is that it's about power imbalance.
And when I look at, so who's in AI using stuff without consent? And what is our antitrust concern? Is it that the creators are going to get together and try to negotiate something? Or is it that companies, like Microsoft, with a $3 trillion market cap have decided, hey, we're big enough that we can infringe on everything and just fight it out?
So where is the antitrust dynamic there? What should we be looking at in terms of antitrust law? Should we be worrying about the creators trying to negotiate? Or should we be worried about the incredible monopoly power of Microsoft and its semisubsidiary, almost-nonprofit OpenAI company? Where should we be actually looking at this at a policy macro level?
MODERATOR: Tim.
TIM WU: I'll speak since I don't have to speak for the administration.
[LAUGHTER]
I'll save the FTC this question. I mean, I think the Justice Department and FTC are not particularly focused on market power coming from copyright when they're concerned about the tech platforms and their monopoly power. I think they are-- like in the Google case, they're worried about monopolization of search or other markets. So I don't know if I got it right.
When I think about a OpenAI, Justice, according to reports, is investigating the Microsoft-OpenAI relationship. And maybe they're thinking that it's turning into a merger or something along those lines. So I think that's where the main concern is.
When copyright has been an antitrust concern, which is relatively rare, I think it usually centers on the idea that so a dominant monopolist is going to, like any sort of essential resource, kind of, I guess, grab a hold of all of the essential input for a certain industry.
So a little bit like a US steel company getting the only iron ore mine, for example, and then being able to foreclose competitors, I guess the concern would be-- and like, back in the old days, there was only one monopolist record player company. And they somehow have exclusive deals with every single. Then you can't have other record companies starting. That's along the kind of lines. But I don't think it's a center of the administration. I think they're worried about the core platform power of the tech platforms.
MODERATOR: David.
JUDGE STRICKLER: Yes. I'd like to answer that question with reference to the--
MODERATOR: Take that mic.
JUDGE STRICKLER: I'm sorry. Yes. Here we go. I want to answer that question as it relates to potential statutory compulsory licensing, which either through an antitrust consent decree or through statutory initiative that creates a compulsory license might work. As a judge, in my role as a judge, I don't have any opinion on that.
As a-- nothing that would be within our jurisdiction now but nothing that I could say. As an adjunct professor, I teach at least one of the classes each year with regard to artificial intelligence. And in conjunction with that, I've looked at some of the comments that have been filed by licensors and licensees of music in connection with-- this is in response to Copyright Office inquiries.
And the one thing they all agree on, in no event should there be statutory compulsory licensing. Now, why do they both agree on that? The dynamics that I mentioned in my presentation would apply equally in that situation. If you own the copyrights, you have this great must-have power.
And if you have, this-must have power, why would you want compulsory licensing? You'd want to take advantage of your power. That's what businesses do. That's what they're supposed to do. That's not a criticism. That's an explanation of reality.
The AI companies say, no, we don't want compulsory licensing because, in fact, we want no licensing at all. We say there's a fair use exception. There's a transformative quality to what we do. And we shouldn't have to pay anything, so why would we license in a compulsory way or in any way if there's no property right involved here at all?
So should the answer be a consent decree in antitrust or an outright statutory conclusion that there should be compulsory licensing? Is one argument in favor of that? The fact that both sides don't want it, at least as of now, maybe, but I don't know.
AUDIENCE: Both sides don't want it because they don't know how the cases are going to come down. One side will want it very soon.
JUDGE STRICKLER: It is a game theory type of problem, no doubt.
MODERATOR: I know. I'm guessing you're not going to-- OK, all right.
[LAUGHTER]
Yes, David --
AUDIENCE: Can the panelists speak a little bit to the question of unionization? Coming back to professor Ginsburg's question at the beginning, which is, what about the authors? It's one thing to say, even Apple and Google and Amazon need protection from three record labels that have 90% market share. It's another thing to say they need protection from individual songwriters.
And so traditionally, songwriters, playwrights can't unionize because they keep their copyright, whereas film writers, television writers can unionize because they give up their copyright, which essentially makes copyright almost worthless to those that are not giving up their copyright. So just as a hypothetical, suppose there was a performing rights organization made up only of songwriters and not music publishers, does the analysis change?
JUDGE STRICKLER: You want to give it a try.
TIM WU: You go first. You go first.
MODERATOR: David, go ahead.
JUDGE STRICKLER: Thank you. I don't think-- certainly from the economic point of view, I don't think the analysis changes. The question, I suppose, is, why have songwriters given up any rights they've had or assigned any rights to music publishers? Presumably, because these are voluntary transactions, it's because the music publishers can do this better. And if they can do it better, then it's the power of the music publishers.
Now, if the songwriters decided to band together and do this, presumably songwriters don't have the skill, the commercial scale, the business skills. Their great skill is song writing, lyrics, and music. So they would have to contract out or bring something in-house and basically create the same type of structure that the music publishers have, and that would give them a power similar to what the music publishers have.
Am I understanding your hypothetical correctly? You're talking about music songwriters organizing on their own. Those are the authors you're referring to, right?
AUDIENCE: Presumably they could get together and hire a lawyer to--
JUDGE STRICKLER: A lawyer in business--
MODERATOR: : I take it David's asking a question more about the legal ramifications rather than the economics of it.
AUDIENCE: Yeah, because the question is, do individual songwriters really have market power over somebody like Apple and Google? It's one thing in the 1940s to say mom and pop movie theaters, which virtually don't exist anymore. We have five or six chains. We don't really have mom and pop movie theaters anymore. So the question is, do the music users really need protection from individual artists negotiating collectively?
JUDGE STRICKLER: Well, I think anybody who negotiates collectively creates a power for themselves. But if the parties they're negotiating against have their own power, well, that's the countervailing power I spoke of before. And those will, in some sense, neutralize one another.
So we're looking at power-- disproportionate power as a concern. If we get proportionate power, I don't think there's a particular cause for concern, which is what animated the Web V decision that I talked about earlier.
TIM WU: Can I just say a quick comment? There'a a question-- and interesting. And I guess, I think it's a difficult one for antitrust. It's the only thing-- this is a very general answer, but I think it's very important in these policy discussions to be very realistic about the economic power and where it's held in our current markets.
And most of it's with the platforms. I would say, I agree, that let's just face it. And the challenge is some of the setups slightly presume differently. But I think the reality is that whatever power rightsholders may have, the tech platform power far exceeds it by orders of magnitude. And I think we need to be a little more realistic about that.
MODERATOR: All right. We are at the end of our time, so please join me in thanking our panelists.
[APPLAUSE]
And I'm going to hand the podium back to Kate to close things up.
INTRO SPEAKER: So quickly, I just wanted to say thank you all so much for attending our annual symposium. I hope you learned something and enjoyed your time with us. We'd love to have you back. And a sincere thank you and hats off to our tremendous panelists, both here and in the audience, that shared their time and expertise with us today. Thank you.
[APPLAUSE]
And I'd also like to thank the leadership and the staff of the Columbia Journal of Law and the Arts, who you've seen throughout today. And we'll be publishing a symposium issue covering today's events. So I'd like to thank them for their work and, of course, our wonderful staff here at the Kernochan Center and Columbia Law, including Samara Weiss, who is the woman behind the curtain that made all of this happen today. So thank you, Samara.
[APPLAUSE]
And of course, our amazing facilities team, AV, and catering who just made this a wonderful day. so thank you all so much. My final CLE reminder, please sign out, and then have a wonderful weekend. Thank you again, everybody.
[APPLAUSE]