Spotify’s “Fake Artist” Issue and Other Problems at Scale

Spotify just can’t seem to catch a break in the artist community.  A story broke on Vulture evidently based on a Music Business Worldwide post alleging (and I’m paraphrasing) that (1) Spotify commissions artists to cover hits of the day and (2) there’s a lot of sketchy material on Spotify that trades on confusing misspellings, “tributes” and other ways of tricking users into listening to at least 30 seconds of a recording.  Which means that Spotify isn’t that different than the rest of the Internet.  (Thank you DARPA, the people who gave you the Internet.  And Agent Orange.  The real one.)

Spotify of course has issued a denial that I find to be Nixonian in its parsing.  Let’s not go crazy on this, but here’s the first part, according to Billboard:

“We do not and have never created ‘fake’ artists and put them on Spotify playlists. Categorically untrue, full stop,” a Spotify spokesperson wrote in an email.

Nobody said Spotify “creates” “‘fake artists,'” and the accusation was that the fake artists were on the service AND on some playlists, not just playlists.  The allegation is that Spotify commissions recordings.

“We pay royalties — sound and publishing — for all tracks on Spotify, and for everything we playlist. [If Spotify commissioned the fake tracks, they would also “pay royalties”.]  We do not own rights, we’re not a label, all our music is licensed from rightsholders and we pay them — we don’t pay ourselves.”

Notice the switch to “rights holders” which would include either publishing or sound recordings.  If Spotify commissioned fake artists they would not need to “own rights” and they could easily have “licensed” the fake artists recordings.  Cover songs would require an…ahem…NOI for the compulsory license.  And the commission payment could go to the artist as a buyout so Spotify would “pay them”.  If the object was to increase traffic for their ad supported service, commissioning recordings would both increase traffic AND reduce the prorata share of advertising revenue by making the denominator larger for everyone with a  revenue share during that accounting period.  I don’t want to go too far down that rabbit hole, but there are some odd loose ends.

Leave the holes in Spotify’s denial to the side.  The core problem identified by the Vulture post is the same for Spotify as it is for Google, YouTube, Facebook, all the other Internet companies that require “scale” to succeed, and which are, one way or another, hell bent on being monopolists.  The second part of Spotify’s denial in Billboard could apply to this lack of monitoring:

“As we grow there will always be people who try to game the system. We have a team in place to constantly monitor the service to flag any activity that could be seen as fraudulent or misleading to our users.”

Maybe that “team” could have a role in “monitoring the service” for tracks before the recordings get on the service rather than after.  Noah built the Ark before the rain.

It must be said that it sounds a bit implausible that Spotify would commission this type of recording to avoid paying artist royalties on the fake tracks.  Such an affirmative act would require a commercially tortured logic because the royalty offset on those specific tracks would be so tiny that the cost of the commissioned recordings would have to be very, very low.  One guy with Garageband in Mom’s basement kind of low.  How much the prorata revenue share would be reduced is hard to know from the outside.

But even if Spotify doesn’t hire studio musicians to perform “fake hits”, it appears that they are allowing a lot of sketchy recordings onto the service.  One might ask how those recordings get there in the first place.  I would bet that the explanation is pretty much that nobody bothers to check before the recordings are posted (or “ingested” in the vernacular, if you can stand that word).

So while there is a major difference in degree of harm, there isn’t a great deal of difference between what seems to be happening on Spotify with sketchy recordings and the links to illegal materials that the Canadian Supreme Court just blocked on Google Search, promoting the sale of illegal drugs for which Google paid a $500,000,000 fine and narrowly avoided prison, ISIS recruiting for which Google lost a chunk of market cap (at least for a while), human trafficking on Craig’s List and fake news on Facebook.  Each of these services operate at scale and they seem to have the same problem:  No one is minding the store and there are no or poorly enforced standards and practices that are only enforced after the harm has occurred.

The other trait that all these companies have in common to one degree or another is that they are all at least dominant if not monopolies in their markets.

Remember–on May 12, 2014, Spotify’s director of economics Will Page gave a presentation at the Music Biz Conference in Nashville.  As reported by Billboard, Will Page gave the audience a good deal of evidence of Spotify’s domination of the online music market:

Spotify claims to have represented one out of every ten dollars record labels earned in the first quarter….Page’s claim shows the speed at which subscription services are gaining share of the U.S. market. According to IFPI data, all subscription services accounted for 10.2 percent of U.S. recorded music revenue in 2014. If Spotify had a 10-percent share in the first quarter, it’s safe to say the overall subscription share is well above the 10.2 percent registered last year.

These numbers suggest that while Spotify may have a significant share of overall U.S. recorded music revenue, Spotify is clearly dominant if not a monopoly in the global subscription market with its now 100 million plus users and probably is at least dominant if not a monopoly in the U.S. music subscription market.

So how does Apple address these problems?  If you consult the iTunes Style Guide, you’ll see that iTunes expressly prohibits the use of search terms or keywords in track title metadata (like “Rock Pop Indie Rock”) or an artist name (like “Aerosmith Draw the Line).  Audio files have to match track titles on each album delivered.  “All track titles performed by the same artist on an album must be unique, except for different versions of the same track that are differentiated by Parental Advisory tags.“  And most importantly, perhaps, “the name of the original artist must not be displayed in any artist field on the track level or the album level.”  Why these rules?  One reason might be that Tunecore has encouraged their users for years to use covers as a way of getting noticed in searches on music services (with suitable admonishments to not “trick” fans).

Let’s face it–there’s only one way to keep your service clean.  Don’t let the bad stuff on in the first place.  You may think that it should be self evident that allowing sketchy recordings, ISIS videos or human trafficking on your service is a bad thing.  You may think that it should be self evident that allowing someone to change a letter in an artist’s name to trade on their reputation is a bad thing–not that different from typo squatting.  You may think that it is self evident that promoting the sale of illegal drugs is a bad thing.  And you may think that anyone who wants to engage in commerce with the legitimate commercial community, much less the artist community, wouldn’t allow these travesties into their business.

But you would be wrong.  Probably because you don’t worship at the alter of the great god Scale.

 

 

 

Will Digital Aggregators Lead the Industry on Transparency with Spotify and Others?

The Music Managers Forum UK have criticized the “secrecy” arounds Spotify’s deals with major labels.  According to Complete Music Update:

The UK’s Music Managers Forum yesterday welcomed the news that Spotify had reached a new deal with Universal Music. However, the trade body criticised the continued secrecy that surrounds the deals made between the major record companies and the streaming services. This secrecy means that artists signed to or distributed by those labels are not allowed to know the specifics of how their music is being monetised.

The same criticism could equally be made of non-statutory, statutory, or direct agreements by digital aggregators like CD Baby, Tunecore, LyricFind, Pledge Music, the Orchard and Loudr, each of which offer varying degrees of transparency of their own books, much less the deals they’ve made with digital services on behalf of the artists, songwriters, labels and music publishers appointing them as agents for relicense of music.  (Loudr, for example, has recently started participating in the most obscure licensing process of all, the mass NOI registrations with the Copyright Office.  Read more about that on another series of MTS posts or my recent article in an American Bar Association journal.  At least with mass NOIs, songwriters know what their royalty is–zero.)

Loudr NOIs
Mass NOI Filings by PK Interactive on behalf of Loudr

It is probably fair to say that there is no disclosure of the actual terms of the direct licenses between these aggregators and the services concerned.  It may also be possible that no one has ever asked the aggregators for the terms of their deals.

That’s a real head scratcher because arguably those aggregators have an even greater obligation to disclose these terms given they cater to many artists, songwriters, music publishers and labels who are unlikely to have the means–even if they have the right–to conduct a royalty examination of any of these companies.  However big a problem anyone has with major labels, every major label artist and major publisher songwriter takes their “audit” rights for granted.

It would be very simple for aggregators to disclose the terms of their deals or to at least summarize them so that artists or songwriters who are considering who to sign with could compare payouts.  It’s fine to tell people what their royalty split, flat fee, or distribution fee might be, but the assumption is that the revenue stream being shared is identical from one aggregator to another.

Also remember that it is common for music services to pay “nonrecoupable” payments to labels–just like it was for record clubs.  This comes in the form of “breakage” or “technology payments” or other ways to keep the money from being called a royalty.  We know this very likely happens with major labels although the amounts are not disclosed–hence the MMF UK’s beef.  We have no way of knowing if it happens with digital aggregators or even what the basic terms of the deals are, which makes it difficult to conduct a desktop audit (the precursor to a full-blown field audit), much less an exhaustive royalty examination.

So let’s not limit the transparency concern to just the major labels.  The digital aggregators could easily lead the way forward by posting the terms of their deals with digital services.  Unless of course the problem lies as much with the digital services as it does with the labels.

 

Five Things Congress Can Do to Stop Tens of Millions of “Address Unknown” NOIs

Copyright reform is on the front burner again after the passing of the  Register of Copyrights Selection and Accountability Act by a vote of 378-48.   But there’s another problem the Congress needs to fix that won’t require legislation in the short run:  The mass filing of tens of millions of “address unknown” notices under the archaic compulsory license for songs.

I’m going to assume that readers know the general background on the millions of “address unknown” NOIs filed with the Copyright Office under a loophole in the Copyright Act (Sec. 115(c)(1)).   If that is Geek to you, see my recent paper on mass NOIs for more complete analysis (or previous posts on MTS for the armchair version of the story.   The first distinction to remember is that we are only concerned in this post with song copyrights and not the sound recording.  This story implicates songwriters and publishers, not artists and record companies, and it only applies to the government’s compulsory license for songs, a uniquely American invention.

In a nutshell, Amazon, Google, Pandora, Spotify and other tech companies are serving on the Copyright Office tens of millions of “address unknown” notices of intention to obtain a compulsory license to make and distribute recordings of certain types of songs.  Under what can only be called a “loophole” in this compulsory license, a service can serve these “address unknown” NOIs on the Copyright Office if the owner is not identifiable in the Copyright Office public records.  The Copyright Office stands in the shoes of the “address unknown” copyright owner to receive and preserve these notices.

On the one hand companies like Amazon, Google, Pandora and Spotify say that they can’t find these millions of song owners, while at the same time at least some of the same companies brag about how comprehensive and expensive their song databases are (like Google’s Content ID) or their agents puff up the agent’s own massively complete song databases as “the worlds largest independent database of music copyright and related business information.”  And yet, these same companies and their agents can’t seem to find songwriters whose names, repertoire and contact information are well known, or whom they already pay through their own systems or through their agent.

The Database Double Loophole Trick

Here’s the loophole.  First, the loophole requires a very narrow reading of Section 115(c)(1) of the Copyright Act, a 40 year old statute being applied to NOIs served at a scale the Congress never intended.  If the song owner isn’t found in the public records of the Copyright Office, even if the digital service or its agent has actual knowledge of the song copyright owner’s whereabouts, the digital service can say they are not required to look further.

Even if you buy into this willful blindness, these digital services may not be looking at the complete public records of the Copyright Office.  The only digitized records of the Copyright Office are from January 1, 1978 forward, and my bet is those easily searchable records are the only records the services consult.  That omits the songs of Duke Ellington, Otis Redding, The Beatles and five Eagles albums not to mention a very large chunk of American culture.

The Copyright Office records from before 1978 are available on paper, so the pre-78 songs are still in the public records (which is what the Congress contemplated in the Copyright Act).

The identifiers are just not “there” if you decide not to look for them.  However, it is not metaphysical, it is metadata that exists in physical form.  This is the “double loophole”.

The Double Triple:  New Releases

Another category of song copyrights that will never be in the public records of the Copyright Office in their initial release window are new releases with recently filed but not yet finalized copyright registrations.  The Copyright Office itself acknowledges that it can take upwards of a year to process new copyright registrations.  This allows “address unknown” filers to bootstrap a free ride on the back of Congress during that one-year period.

No Liability or Royalties Either:  Trebles All Round

Once a company serves the “address unknown” NOI on the Copyright Office, songwriters are arguably compelled by the government to permit the service to use their songs.  Filing the “address unknown” NOI arguably allows the service to avoid liability for infringement and also–adding insult to injury–to avoid paying royalties.  If the NOI is properly filed, of course.

In current practice, a mass “address unknown” NOI is usually a single notice of intention filed with a huge attachment of song titles with the required fields, such as this one Google filed for Sting’s “Fragile”, the anthem of the environmental movement (which was clearly filed incorrectly as the song was registered long ago):

sting-fragile-google-noi

The number of mass “address unknown” NOIs being posted by the Copyright Office on an almost daily basis suggests that tech companies now view mass “address unknown” NOIs as the primary way to put one over on songwriters and the Congress, too.  Companies like Amazon, Spotify, Google, Pandora and others are using this heretofore largely unused loophole on a scale that flies in the face of Chairman Goodlatte’s many hearings in the last session of Congress on updating the Copyright Act.

This “address unknown” practice also undermines the efforts of Chairman Goodlatte and Ranking Member Conyers to modernize the Copyright Office.  Indeed, based on the very lopsided vote on HR 1695 the Register of Copyrights Selection and Accountability Actit is clearly the desire of the overwhelming majority of Members of Congress, too.

March Spotify NOI Filings

What Can Be Done?

Congress can play a role in in providing immediate relief to songwriters by stopping the mass “address unknown” NOIs or at least requiring the Library of Congress and the Copyright Office to take steps to verify the NOIs are filed correctly.

At the moment, the government takes away property rights from the songwriters by means of the compulsory license without taking even rudimentary steps to assure the public that the “address unknown” NOI process is being implemented correctly and transparently.

Here are five steps the Congress can take to rectify this awful situation.

  1.  Stop Selling Incomplete Data:  Congress should instruct the Library of Congress to stop selling the post 1978 database until due diligence can be performed on the database to determine if it is even internally correct.  It appears that many if not all the mass “address unknown” NOI filers use the LOC database to create their NOIs.  It is also highly unlikely that this database will include new releases.  Congress can simply instruct the Librarian to stop selling the database.loc-prices-databases
  2.  Stop Accepting “Address Unknown” NOIs With Compressed File Attachments: Congress should instruct the Library of Congress and the Copyright Office to immediately cease accepting “address unknown” NOIs with compressed files as attachments for what appears to be a single NOI.  These compressed files are so large in most cases that songwriter can never uncompress them on a home computer to determine if their songs are subject to “address unknown” NOIs.  Google in particular is a major offender of filing huge compressed files.  Each compressed file contains tens of thousands of song titles.Google March NOIs
  3.  Require Accounting Compliance with Copyright Office Regulations:  Long standing regulations require that anyone relying on an NOI must file mostly and annual statements of account reflecting usage of the songs subject to the NOIs.  The tech companies serving mass NOIs are not rendering these statements and thus fail to comply with the transparency requirements of Copyright Act.  All of the “address unknown” NOIs served during 2016 are out of compliance with the regulations, and all “address unknown” NOIs served in the first quarter of 2017 are likewise out of compliance.  Congress should instruct the Copyright Office to require monthly and annual statements of account be filed with the Copyright Office for anyone who has relied on these NOIs as required by the regulations.  All statements of account should be certified in the normal course as required by the regulations, and made available to the public by posting to the Copyright Office website.
  4. Require the Library of Congress to Create a Searchable Database of NOIs:Congress should instruct the Library of Congress to create a single database maintained online that is maintained by an independent third party and is searchable by songwriters in a manner similar to a state unclaimed property office.  That database needs to be updated on a regular schedule.  Given the size of the compressed files served to date, it is essentially impossible for songwriters to determine if NOIs have been filed on their songs.  This is particularly true as the NOIs are served on an effectively random basis, so even if songwriters were able to search, they would essentially have to search all the time.
  1.  Pay Royalties Into A Permanent Trust Account:  Given that it is highly likely that the mass NOIs filed to date have a significant number of errors, it is also likely that songwriters will become entitled to payment of royalties retroactively if these errors are ever caught.  Therefore, the Congress should require that royalties should be paid to a trust account maintained at the Copyright Office and held in perpetuity like a state unclaimed property office.  Of course, it is equally likely that the song copyright owners will be entitled to terminate any purported license under 17 USC Sec. 115(c)(6).  These payments should be based on actual usage and not black box.  This is another reason why the statements for “address unknown” NOIs should be public.

What started in April 2016 as a trickle of NOIs from a handful of companies has now expanded exponentially.  Based on Rightscorp’s analysis in January 2017, some 30 million “address unknown” NOIs had been filed–and that did not include the dozens of “address unknown” NOIs filed by Spotify in March 2017 alone which themselves likely total over a million songs.

NOI Table
Top Three Services Filing NOIs

April, 2016-January 2017

Number of NOIs Per Service
Amazon Digital Services LLC 19,421,902
Google, Inc. 4,625,521
Pandora Media, Inc. 1,193,346

It is rapidly becoming standard practice for tech companies to try to pull the wool over the eyes of the Congress by leveraging an apparent loophole and they are doing it on a grand scale.

As we have seen with everything else they touch from the DMCA to royalty audits, the tech companies will continue this loophole-seeking behavior until they are forced to stop.  Since no one at the Library of Congress seems to have the appetite to right this wrong, the Congress itself must step in.

Ultimately Congress should fix the loophole through legislation, but in the meantime most of the harms can be corrected overnight by policy changes alone.