Comment on “Frontiers: The Interplay of User-Generated Content, Content Industry Revenues, and Platform Regulation: Quasi-Experimental Evidence from YouTube”

Before my main caution about the article, I want to highlight an important finding: The article suggests that there’s a conflict of interest among musicians, which might make the companies poor representatives for at least some of the artists whose works are involved. YouTube helped discovery of new music but decreased revenues attributable to already-successful music (Lunney 2019); that might be a worthwhile tradeoff if we consider the declining marginal utility of wealth.¹ However, if the history of collective licensing is any guide (Band and Butler 2013), it will be the less-successful creators whose interests are sacrificed, deprioritizing discovery of music and further rewarding those who already are successful. I also note that there is no indication in the article how much of the revenues attributed to particular pieces of music reached the composers. As musicians have long bemoaned, record and publishing companies have many techniques for interrupting the revenue stream before it reaches the actual authors.

However, the article has more to say, and so it is vitally important to be clear about what regimes it is actually examining. The authors seem to believe that legal “safe harbors” are what enable YouTube to negotiate payments for less than Spotify does for music, even though YouTube has a lot of content that does not rely on major labels’ music—or any music at all—and still drives lots of engagement and revenue. Music is just less vital to YouTube’s success than to Spotify’s (although, not for nothing, Spotify’s pivot to podcasts seems related to its desire to diversify and pay musicians less too).

More fundamentally, attributing causation to safe harbors requires understanding what that phrase means. The authors state “user-generated content (UGC) platforms such as YouTube are either explicitly protected by safe harbor provisions (e.g., in the United States), or courts have ruled in favor of the platforms, stating that, for example, YouTube cannot be held legally responsible for user-uploaded content.” This description is, unfortunately, misleadingly incomplete.

In general, when lawyers talk about safe harbors, they mean rules that allow some entity to be confident that it will not be liable when it follows clear rules rather than having to balance all the relevant circumstances and decide what to do in each case. It’s the difference between having to follow the rule “drive safely” and the rule “drive under 55 miles per hour”—the latter is a safe harbor, although there may be other relevant driving rules and, similarly, copyright safe harbors have many details designed to avoid bad-faith evasion.

Therefore, what does a UGC platform have to do to get the U.S. safe harbor? A number of things, but the key is responding expeditiously to takedown notices from copyright owners. Despite a litany of complaints, the U.S. safe harbor regime generally works well; courts have not allowed true bad actors to avoid liability, and complaint services process takedowns so that infringing uploads are quickly removed (Tushnet 2020).

The European regime was, in the past, roughly similar, although its contours were less clear. In both systems, there was no safe harbor just because a user, and not the platform itself, posted infringing content on the platform, as casual readers of the paper might think. Instead, a platform has the burden of taking down identified infringing uploads, or it faces liability. Copyright owners and their agents therefore developed programs for sending notices on a large scale—Alphabet, which owns YouTube, has received billions a year. Although many large copyright owners hate this system and deride it as “whack-a-mole,” it can be quite effective for taking down infringing content before it can have much of an impact.

Nonetheless, complying with lots of takedown notices was a pain for Alphabet (Google), which developed Content ID, a system that identified certain content during upload; once that’s done, Content ID can offer copyright owners the chance to block it or allow but monetize it. During the dispute with Gesellschaft für musikalische Aufführungs und mechanische Vervielfältigungsrechte (GEMA) (Society for musical performing and mechanical reproduction rights), Google used the same mechanism to prevent GEMA-associated music from being available in Germany at all. What the authors are actually testing, then, is the difference between blocking—carried out by Content ID—and monetizing—also carried out by Content ID. That doesn’t involve a safe harbor. Indeed, the reason that YouTube was blocking music in Germany was precisely because it chose not to rely on the putative European Union (EU) safe harbor protections for user-uploaded content.

In an endnote, the authors contend that the posttreatment condition is “similar” to conditions under a safe harbor. However, their evidence simply does not allow that claim, because they do not measure the amount of music available under a safe harbor regime that includes takedowns. The posttreatment condition is licensing, which is the opposite of a safe harbor—there is no need for legal protection against liability for unlicensed use when the platform has a license for the music it hosts. What they seem to mean is that “there was a lot of music available to consumers on YouTube after treatment.” However, what they do not show—and cannot—is that there is the same amount of music available to consumers on any site that does rely on a notice and takedown or similar safe harbor. The music companies’ mass notice-sending systems make that unlikely. Of course, such a site would not be paying any money to the music companies at all—it would be relying on the safe harbor, so the payout side would be zero.

Likewise, in their rejoinder, the authors make clear that they have fundamentally mistaken what a safe harbor regime is. They clarify that they make their safe harbor analogy by assuming two things: First, if there is a licensing and filtering requirement, there will not be a lot of big label content on a platform (that is, there would not be much under “a very strict regulatory regime (e.g., platforms would have to clear rights for all UGC that is uploaded)”). Second, they assume that there will be a lot of big label music on a platform if there is only notice and takedown. However, the opposite is true on both counts: On most platforms that rely on safe harbors, there is not much infringement, and notice and takedown keep major label content scarce. Meanwhile, platforms like YouTube have cut licensing deals with major labels so that UGC containing their works is plentiful and monetized in their favor; Content ID is how YouTube is complying with Europe’s new licensing mandate.

The article has important and useful results for music licensing. What it does not do is test the effects of a legal safe harbor. There is careful qualitative research that bears on this question, with important lessons for the overall ecosystem. Importantly, notice and takedown keep costs low for UGC sites that do not have real infringement problems, which is most of them. Most UGC sites are more like Ravelry (a site for people interested in fiber arts) than they are like YouTube, and the notice and takedown system works reasonably well for the occasional infringement on those diverse sites.² Meanwhile, sites that receive many takedown notices tend to evolve into licensing-based sites like YouTube for reasons both technical and practical (Urban et al. 2016, 2017).

If we assume that laws should be made for the circumstances of Alphabet and Meta, all we will have is Alphabet and Meta. That would be bad for the flourishing of creativity in general, including for the large copyright owners who will see their negotiating leverage disappear along with everyone else’s. It would be a shame if this careful research were misused to argue for legal rules that harm the overall Internet ecosystem and its diverse participants, when it in fact shows the economic effects of one giant (YouTube) on others (major music industry participants).