Guest Post: Follow the Money: YouTube’s Failure to Pay Retroactively Gives “Conversion Rate” a Whole New Meaning

[Cross-posted from MusicTechPolicy]

by Chris Castle


A performance metric one hears from the digerati is the term “conversion rate.”   “Conversion rate” for a streaming service usually means the rate at which users of an ad-supported free service are “converted” to paying users.  That motivation is usually because they are so fed up with the advertising they are willing to pay.  (This was one of the many failed pitches from Spotify before people stopped trying to justify hanging on until the IPO riches flowed in.)

YouTube, of course, has never been too terribly interested in anything that moves users away from advertising.  That resistance (and potential internal competition between the massive ad sales team and the ever changing YouTube managers), may explain the many failed efforts at launching a YouTube subscription service by a company that knows more about user behavior than anyone in history.  They just couldn’t seem to get it right for the longest time.  You don’t suppose that YouTube’s apparent lack of interest in getting large numbers of users to substitute away from free to subscription was because YouTube made a lot more money from the ads than they ever would from the subscriptions?

One of the ways that YouTube (and Google) makes money from advertising is by taking money that is not theirs to take (sometimes called “monetizing” content).  The civil law calls that act a claim of “conversion” and   the criminal law calls it the crime of “theft”.  Conversion and theft are two sides of the same coin and often one implies the other, albeit with different burdens of proof.


YouTube’s Content ID tool is a way for copyright owners to block or permit advertising on user-generated content that includes their copyrights, often music.  Users of Content ID will tell you that it works just well enough that Google can say it is an effective tool, but even with Content ID music still gets through (and is often monetized by YouTube) for a variety of reasons.  This requires time consuming and costly manual searches.  Companies like AdRev make it a bit easier, but are essentially third party Content ID users.  These companies are compensated with a commission on infringing works they find on YouTube that they convert–there’s that word again–from infringing to monetized, which means that YouTube now splits the advertising revenue with the copyright owners who in turn split their share with an AdRev.

But see what happened there?  If you have Content ID, you can block on the upload some of the time, or you can do a search.  If you don’t have Content ID (see Maria Schneider’s class action) then you can’t block on the upload only chase the infringements manually.  But quite rightly from an economic perspective, companies like AdRev are not that interested in doing that work on a rev share basis if there’s no rev share when you block.

Here’s the point–you have a property right in your copyright.  You have a property right to license that copyright.  Any revenue derived from exploitations of that copyright is your money.  YouTube uses its monopoly power to impose a deal to monetize your copyright (under duress, of course, due to whack a mole DMCA).  That deal involves a revenue share.  (Let’s just assume you decide to take the King’s shilling and accept Google’s deal under duress which you shouldn’t have to do and which may not even be enforceable.)

The question is, when should that revenue share attach–when they start exploiting your copyright in violation of your property rights or when you catch them doing it.  And if (1) you catch them violating your property rights and (2) agree to monetize, when should they pay you your agreed upon share of the revenue from monetizing?  Should they pay retroactively to the first exploitation?  Or only prospectively after you catch them?

The correct answer is they should pay retroactively.  But they don’t.  They just keep the money.  For millions of infringements.  And they get away with it because of their monopoly power, which leaves one choice most artists won’t make, which is to sue them like Maria has.

Remember–Content ID operates largely like any other fingerprinting tool.  (Psychoacoustic fingerprinting is old technology–remember Jonesy in “The Hunt for Red October”?  That’s fingerprinting.  A “fingerprint” is simply a mathematical rendering of the waveform of an audio file.)

There is a reference databases of recordings that are “known knowns” (which is why it is important to be included in the Content ID database as Maria Schneider correctly points out in her class action.)  The fingerprinting tool encounters a new file, takes a fingerprint, then looks for a match in the reference database and reports a result that triggers an action.  Typically, fingerprinting tools are binary:  match or no match.  What happens after the tool finds a match is entirely in the control of the operator.  (So while the tool could have a match rate of 90%, the operator could report a random number of matches or a fixed number of matches, like one every ten, or one every 1000.  That means 90% accuracy could turn into a much lesser percentage of reported matches.  It’s important to know how many matches trigger an action.)

Having had some experience with audio fingerprints, I think you will find that once a fingerprint is in the reference database, the recognition tool (Content ID in this case) will spot the reference fingerprint a very, very high percentage of the time.  The fingerprinting tool I’m most aware of caught matches over 90% of the time.  I can’t imagine that a tool developed by the biggest technology company in commercial history would do less–unless they wanted it to.  Remember, this is not taking into account re-records unless the re-record is itself in the database, or pitch bends.  This is an exact match which is very common use of Content ID.  (See Maria’s class action complaint, and Kerry Muzzey has a great description of this in his recent Senate testimony.)

If Content ID is actually missing matches to known knowns on the upload (assuming exact matching is possible), I find it very odd that Content ID is missing much.  Maybe it’s not, but one way to find out is to force Google to reveal the inner workings through discovery in the class action case.

But if Content ID does miss exact matches, it would be interesting to know what percentage of those misses end up being monetized, and of those, what percentage end up getting caught later by a subsequent use of Content ID or a manual investigative process.  This will give an idea of the scale of the retroactive payment issue.

As Maria rightly points out, it is virtually impossible for an artist or film maker without Content ID to catch YouTube monetizing infringing works.  But I think the analysis has to go a step further–even if you have Content ID, at the moment you catch YouTube monetizing illegal versions, you are in no different position than the artist who lacks access to the Content ID tool.

Both have the same problem–YouTube is profiting from illegal copies.  If when you catch them you then elect to monetize, YouTube will pay you going forward, i.e., prospectively.  But I do not believe they will pay you retroactivelyfor the illegal use.  (There is a rumor that some music publishers do get paid retroactively under some settlement, but that needs to be confirmed.)

That means that YouTube is directly profiting from piracy for the retroactive views which could total into the hundreds of millions per day given the massive number of daily views on YouTube.  If you elect to monetize due to YouTube’s monopoly power, you are essentially releasing them from liability under duress.  If you catch them.

So YouTube takes your property, monetizes it, and refuses to pay you for how much they made before you caught them if you ever do catch them.  They dare you to sue them because you would be taking on the biggest company in commercial history that controls 90% of the access to information in the world and routinely defies governments.  Not everyone has the spine of Maria Schneider.

Failing to license at all or failing to pay retroactively means that YouTube profits from piracy by converting your property to their own.  And as Maria rightly points out, Google scrapes user data through non-display uses in the background even if YouTube is not monetizing overtly which they then use to compile user profiles in “millions of buckets” (which dribbled out before Judge Koh in the Gmail litigation (In Re: Google, Inc. Gmail Litigation,  Case No. 13-MD-02430-LHK, (U.S.D.C. N.D. California, San Jose Division, Sept. 26, 2013)).

In either case, the value of the amount converted or stolen should rightly include the value of these user profiles scraped in the background, as well as the advertising revenue.

And don’t forget that Google is controlled by Larry Page, Sergei Brin, and Eric Schmidt through their “supervoting” shares of stock.  It’s hard to believe that this YouTube policy was created without their blessing.

The simplest move for Google would be to simply pay both retroactively and (if the copyright owner elects to monetize) prospectively.  Otherwise, it seems like a huge number of crimes are going on in a very planned and organized way dreamed up by YouTube and Google employees.  “Dreamed up” is also called a conspiracy, and if there’s an actual conspiracy it’s not a theory (which came up in an interesting trade secret misappropriation RICO case against Google they managed to wriggle out of, at least for the moment).

The law has another word for organized theft at scale–we sometimes call it “racketeering.”



Guest Post: The Royal Scam: Content ID and Google’s Massive Profits From Piracy and Crime

By Chris Castle

Google and YouTube have managed to create a scam that has gone both largely undetected and largely unpunished for a decade–illicit activity that can be both seen and quantified through the sale of advertising and is also unseen and unquantified through data scraping in the background.  (I leave it to you to speculate which is more valuable.)

It is rare for Google to get caught like they were with the massive multi-agency sting operation and grand jury investigation by the then-U.S. Attorney for Rhode Island that led to the $500,000,000 punishment and non prosecution agreement in 2011.  (Which led to a very expensive shareholder lawsuit against Google’s board of directors and bizarre settlement.  We’ll come back to the board of directors issue here.)

If you had to put your finger on a moment in time that Google began buying Washington in earnest, it was this sting.  It was also the closest that Larry Page ever came to going to prison with all its earthly delights.  That evidently got his attention.

Google has also faced down civil RICO claims for racketeering through the theft of intellectual property.  The last reported RICO case against Google offers a checklist for how to make a civil RICO claim stick against the Leviathan of Mountain View.  I like the YouTube case a lot better than the inventor’s case they beat back.

But most of the time Google just keeps the money when they get caught.  A prime example is YouTube’s standard practice of refusing to pay a revenue share retroactively after you catch them infringing your work using Content ID.  That unjust enrichment creates an incentive to sharply limit the number of artists or songwriters who get access to Content ID in the first place.  I think this is why Google massively overreacted to Mississippi Attorney General Jim Hood’s Civil Investigate Demand and subpoena that they never did respond to.  Maybe they were covering up the same crimes that got them prosecuted in Rhode Island and they did not want to go through that again.

And therein lies the rub and our topic today:  If Google never gets caught, Google quietly keeps all the money.   For our world, this happens because they’ve artificially limited the tools that independent creators can use to catch the massive infringements.  And even if the majors and a handful of independents get the Content ID tool, YouTube still has the incentive to make Content ID just good enough that they can say it works, but not so good as to actually stop the infringement before it starts.

The majors using Content ID have to employ still other means to catch them, sometimes manually, at great cost.  In fact, you have to wonder if net-net the total costs of administering the YouTube deals actually exceeds the minimum guarantee and royalty payable.  Those tools are simply beyond the reach of the creators, even the few who YouTube grants access to Content ID.

And of course, any user of Content ID (big or small) has to sign up to the take-it or leave-it shakedown deal that limits what you can do about it when you catch them.  Which is just another form of the protection rackets.

This criminal enterprise comes in two flavors (at least):  Ad sales for illegal products (like the drugs, counterfeit tickets and the like), and selling legitimate advertising around content that Google knows or should have known was illegal (like YouTube’s monetization of infringing works).  And, of course, Google scrapes data in the background on all these criminal activities to its great–and secret–profit.

As we saw with the drugs case, Google knew exactly what it was doing, and I’m not willing to believe their rudderless ad sales teams don’t also know exactly what they are doing (remember Google’s ad sales team gave credit terms to infringers, and the drugs sting operation also shows that they brainstormed many criminal dodges to deceive Google’s own best practices team).

What little evidence we can lay hands on in the open source demonstrates that Google must know very well that it engages in criminal behavior–why else was Eric Schmidt advised by then-counsel David Drummond to refuse to answer Senator John Cornyn’s questions regarding the drugs case when Schmidt testified before a 2011 Senate Antitrust Subcommittee hearing?  (Also known as “taking the Fifth.”)  After engaging in a weak attempt at misdirection.  Did they think this question wouldn’t come up so didn’t prepare for it?  I doubt that very much.  (If they cooked up this story without the lawyers, this might well have been a conspiracy.  Attorneys take note:  Crime/fraud execution?)

schmidt senate

Eric Schmidt Takes the Fifth on drugs case to Senator John Cornyn: ” I have been advised — unfortunately, I’m not allowed to go into any of the details and I apologize, Senator”

Now that the U.S. Senate is investigating the effectiveness of the safe harbors under DMCA, this would be a good time for the Department of Justice to investigate Google’s business practices and potential criminal activities.  Smells like RICO to me.

As independent composer (and MTP guest poster) Kerry Muzzey highlighted in his recent testimony before the United States Senate regarding Content ID:

My name is Kerry Muzzey, and I am a film and television and modern classical composer.

I am one of the very few independent artists who has access to YouTube’s Content ID system; and most of my experience with notice and takedown has been on YouTube. Content ID has become a core piece of my licensing business: it is the x-ray that reveals the theft of my music to me. This is why I am also nervous about speaking out today – because I fear retaliation by YouTube and Google. I am concerned that they may take Content ID away from me for raising my concerns publicly. The technology behind Content ID is nothing short of brilliant, and I don’t want to lose access to it.

Growing up, my mom always said: “You’re not allowed to complain unless you’re gonna do something about it.” Senators, my being here today is my “doing something about it.” Today, I have the most unique opportunity I have ever had in my lifetime. I have the opportunity to ask Members of my United States Senate to fix a broken law.

Let’s also not forget the way Google is governed (as is Facebook, Spotify and many others).   Larry Page, Sergei Brin and Eric Schmidt hold a special class of  “supervoting” shares, what SEC Commissioner Robert Jackson has called “corporate royalty”.

These insiders get 10 votes for every one share they own of a special class of supervoting stock.  This means that the insiders control over 60% of the voting stock and win all shareholder votes—including votes to appoint the board of directors.

Supervoting shares give insiders absolute control of Google–one of the most successful public companies in commercial history.  Because they control every aspect of Google’s operations, Google truly is their “alter ego.”  One purpose of Google’s lobbying spend must be to keep the corporate royalty out of prison.

These supervoting Google Class B shares are not available to the public.  The public can buy two classes of stock:  GOOGL shares are Class A (one vote per share) and GOOG shares are Class C (no votes per share).  (GOOG shares were issued in a dividend to GOOGL holders.)  GOOGL shares typically trade slightly higher than GOOG which may demonstrate that the market has priced in a lack of meaningful voting rights in GOOGL.

It should not be surprising that Google shareholder meetings are a one-way communication event. The supervoting corporate royalty tell the other shareholders how things are going to be and vote down any move by GOOGL holders to change the status quo—like converting supervoting shares into one share one vote.  As Floyd Norris reported in his New York Times “Economix” column, “Rarely has a shareholder vote been less suspenseful.”

So Google’s profit from evil is not an accident.  If Congress wants to fix the DMCA, let’s fix all of it.  And as U.S. Attorney Peter Neronha discovered ten years ago, that requires a grand jury.


Guest Post: Copyright Office Regulates the MLC: Selected Public Comments on MLC Transparency: MediaNet

By Chris Castle

The wisest among us learn to use their portents well
There’s no need to hurry, it’s all downhill to hell.

From “Don’t Stand Still“, written by Original Snake Boy, performed by Guy Forsyth

The Copyright Office has solicited comments on the transparency of The MLC and received quite a few well-thought out comments (if I say so myself).  MediaNet

has raised some very interesting questions about the NMPA’s relationship with HFA and The MLC that many have questioned both in prior comments and in the many lawsuits against HFA clients like Spotify for its various licensing failures.  (Note that I don’t really fault HFA all that much because I think it really boils down to choices made by Spotify, another Internet company that is in a rush to enrich themselves at the expense  of songwriters and artists.  If you can fault HFA for one clear choice in that cluster, it’s that they didn’t resign from the job both during and after their ownership by NMPA and SESAC.  Maybe they got stock, too.)

MediaNet raises this interesting point:

In passing the MMA, Congress recognized that the party who controls the database may enjoy an economic advantage over others.9 Although not applicable to the MLC-HFA contract, The Federal Acquisition Regulation System, codified at 48 C.F.R. § 1.000 et seq., provides guidance regarding the principle cited by Congress under the MMA. For example, under FAR 9.505 a contracting officer cannot award a federal contract to a contractor where an organizational conflict of interest (or “OCI”) cannot be avoided or mitigated.

But here’s the clincher:

Applying the principles from the FAR, the arrangement between MLC and HFA raises a number of questions regarding the potential for unfair economic advantage to HFA as a consequence of its control over the operation and administration of the MLC database, including the following:

· Who owns the database, MLC or HFA? [The answer is neither]

· If HFA is terminated by MLC, does HFA own or have a claim to any proprietary or intellectual property rights in the database?

· Will HFA have access to “Confidential” or “Highly Confidential Information” (e.g., contract terms, payments and financial information) of music publishers or other similarly situated organizations such as PROs and administration service providers?

· Will HFA have access to the reporting of usage and required payments of the administrative assessment by significant nonblanket licensees (“SNBLs”) in the notices of nonblanket activity (“NNBAs”) required under the MMA?

· Sources suggest HFA may offer [an “ethical wall”] between its work on the MLC database and other work for third parties not using the blanket digital license, and an audit right to ensure HFA complies with this separation. Can HFA effectively separate such third party work from the database it administers for the MLC?

What are the remedies for non-compliance with such measures?

MediaNet respectfully requests that the Copyright Office, as part of its regulatory and oversight authority to ensure transparency, require that the agreements between MLC and all of its vendors be made publicly available, and with respect to the MLC agreement with HFA, if the information requested above is not disclosed in such agreement, require MLC and HFA to submit answers to the forgoing questions.

It should be obvious to everyone that there is an inherent conflict of interest between NMPA and HFA.  Insufficient care was taken at the Copyright Office and at The MLC to create systems to reduce the fact of this conflict negatively affecting the operations of The MLC which presents an opportunity to leave the bad days behind.  But that didn’t happen and here we are again.

But let’s not forget that The MLC is essentially a quasi-governmental organization and must comply with the Copyright Office’s oversight role despite the intimidation tactics.  And the Copyright Office is already looking a bit ragged around the edges from even the little connection to corrosion they’ve had to date.

For example, the Copyright Office announced that “the Butler Report” was commissioned by the Copyright Office to poll ex-US CMOs about their black box practices, knowledge which likely was common to everyone on The MLC’s board.  I must have missed where this work product was put out for bid, which leads me to think it was a single source consulting contract which is what they use to pave the road to hell when good intentions have supply chain disruptions.  Nothing against Susan Butler who is very competent and engaging, but I can think of several academics who would be better suited and would have been peer reviewed.  We can disagree about that, but why not have them submit proposals?  And also deliver all the work product that the taxpayer financed?

MediaNet raises many more excellent points about the inherent conflicts in the NMPA-The MLC-The HFA relationship and The Copyright Office’s designation process that are well worth reading.  You can find the full comment here.

And keep this in mind:

MLC executive Richard Thompson said at the Copyright Office panel on unclaimed royalties last December,[1] “[A] lot of the time since July has been spent working very closely with the staff at HFA and ConsenSys, really starting to nail down how all of this is going to work at the, you know, lowest operational level, all of the things that we need to work out.”  (Referencing the July 8, 2019 designation of The MLC as the MLC.)  Of course, The MLC didn’t announce the selection of HFA and ConsenSys until November 26, 2019. [2]

If The MLC was already working with HFA in July as Mr. Thompson says, why did they give the world the impression that they had not picked a vendor until November?




[1] Transcript, United States Copyright Office Unclaimed Royalties Study Kickoff Symposium (Dec. 6, 2019) at 28 ln 15.  (my emphasis)

[2] Tatania Cirisano, Mechanical Licensing Collective Selects Leadership, Partners for Copyright Database, Billboard (November 26, 2019).