AI presents all kinds of challenges for music copyright, and intellectual property legality and morality in general. But this is not Music Copyright’s first rodeo.
On the one hand, one can say there’s nothing new under the sun. Minor disruptions arrive now and again, perhaps looking like a disruptive sea change at first, and then softening back into the gentle flow of evolution. Copyright has seen piano rolls, radio, vinyl, cassettes, Napster and mp3s, apple and Spotify, and with some minor adjustments, it’s been mostly a gentle flow.
On the other hand, loathed though I am to say it, since saying it never ever turns out well, but here we go anyway: this time it’s different.
Nitpicky thing #1 — Selection and Arrangement
Copyright can roll with the little things, however stupid. Selection and arrangement is one. Selection and arrangment is most often the last refuge of a plaintiff who has no claim on unprotectable ideas in Song A that they’ve observed in Song B. This is what happened in the Thinking Out Loud versus Let’s Get It On trial. The chord progression common to both works was not a protectable one original to Let’s Get It On, and the rhythm in which it was played was common and also not theirs to claim. So they asserted “Selection and arrangement,” which is a legitimate way to think of a creative work where the assets are chosen and assembled rather than created. Selection and arrangement is a tenet of copyright. But the application of it should be reined in. It’s impossible to render a chord progression in the absence of rhythm, so it’s convoluted to think of it as a pair of individual elections “arranged.” You might be able to argue “combined,” but you’re being pedantic. Selection and arrangement more readily applies to, say, a collage, be it of clay and cutlery, or of Shakespeare sonnet fragments, or of recordings of automobiles, babbling brooks, and Vitamix blenders. Musique concrete was a real thing. Rotten Tomatoes Top Ten Movies Of All Time is a real list, their creation, of movie titles that are not their creation. Not their movie titles but their selection and arrangement of movie titles. But most music is not selected and arranged unless you’re a glutton for semantics. Judge Stanton, who presided over the Sheeran case, said there was no bright line rule which said two elements selected and arranged couldn’t be sufficiently original to enjoy protection by copyright. And I’d imagine he was right about that. So there’s your problem. We wasted a lot of time and expense on a silly argument. We don’t need a bright line on the amount, we need to do away with the concept where it has only a corrupting application.
Nitpicky thing #2 — Filtering
Extrinsic filtering, the last refuge of a defendant from time to time, can go in the same bin. Just as we can’t disassemble a chord progression and the rhythm with which it’s played because the two are so intertwined, we also cannot forensically filter out all musical elements that are unprotectable and then evaluate the originality of what’s left over. What is left over is bound to be unmusical. Any melody worth litigating is going to have unoriginal fragments within it. It’s not forensically sensible to erase them from the analysis. They’re a part of the analysis even if they are unprotectable by themselves. Music is an art form where one thing happens in the context of whatever else is happening on one or another axis. By axis I mean, music happens in time, let’s say that’s the x axis, and unless the work is a solo on a monophonic instrument, there is polyphony, so that’s your y-axis, and the instrumentation might be your z-axis. All of these musical elements derive meaning from the way they relate to what happens before, after, and during. To filter those contexts away is harmful expediency. And when it’s applied as a tactic, it’s shameful. Protectable expression is what we’re looking for, not protectable elements. If we’ve anything nice to say about s election and arrangement, it’s this, the valid premise on which the mostly invalid application of Selection and Arrangement is based. Unprotectable expression is unprotectable, but the idea of filtering is confusing and obfuscating. But it’s not reasonably possible to render music that people will enjoy without unprotectable elements in the mix. It’s the interplay between elements that makes the music. Music is complex this way.
Those are little things, relatively speaking. They might even go away on their own. I’m at least reconsidering my stance on the so-called Inverse Ratio Rule. I used to rail against the Inverse Ratio Rule because it had a logical flaw in it — the ratio to which it refers is much more reasonable in one direction than in its inverse. The inverse ratio rule looked at the two prongs of infringement — access, and substantial similarity. And it reasoned that the more substantial similarity Song B exhibited, the less evidence of access we should require. The relationship makes sense in that direction. If you write a song that’s sufficiently similar to another, we can infer from that similarity that you must have had access to Song A. Fine. But the inverse? What if the two songs are not that similar, but you definitely had access to Song A? Can we infer copying because you had so much access? No, at some point that’s just silly. If the two songs are barely alike, it doesn’t matter how familiar the creator was with Song A. Song A could be Happy Birthday To You, in which case, access approaches infinity, and still doesn’t imply that Song B sounds any closer to Happy Birthday. We shouldn’t have a “rule” called the Inverse Ratio Rule. Rules oughta work more reliably than that. But still, is it nonsense? No. I was too harsh. If two songs sound a bit the same and Ed Sheeran wrote Song B the day after he saw Song A’s artist in concert, we might more readily infer that the similarity in Sheeran’s Song B is the result of copying. We’d be putting on a blindfold if we didn’t consider that forensically relevant. Again, therefore, the problem is in how the concept is applied, whether it’s used to support a frivolous case and whether the judge or jury can adequately evaluate it.
But artificial intelligence? Oh boy.
I started thinking, “this could get outta hand” when Yung Gravy impersonated Rick Astley for his Betty Get Money send-up of “Never Gonna Give You Up.” Parody is usually fair use, but people confuse parody with just plain leverage. If you’re not parodying the work you’re borrowing, then it’s your asset in another pursuit. And you should not be entitled to it. That includes Astley’s “appearance” on your song.
Astley can’t sue for copyright infringement because he doesn’t own the rights to the song, which Gravy licensed, and Astley also doesn’t own the recording, which Gravy didn’t sample. Everything in Betty Get Money that sounds like Gravy sampled it, was rerecorded to sound as though it was sampled. That’s all there is to it, copyright-wise. Copyright expressly allows that rerecording tactic, as long as you’ve licensed the underlying composition. So Astley has to sue for right of publicity which is long for “likeness.” And he probably won’t win that either, in my estimation. But nevermind that. Let’s return to the rerecording thing. Should you be able to do that?
It’s one thing for copyright law to have been written to say, “Okay Stan, you can record your own version of Hey Jude without having to worry that the owner of the recording is going to sue you for sounding too much like it, even if you use a Hofner bass on it. We’re not getting into that.” It’s another to say it should be able to appear in another work as though it’s making an appearance. There’s an awful lot of intellectual property in a recording. Should there not be any limit on how much of that you can borrow? In a world where samples need to be cleared, is that not damaging to the market for the original? Who is going to sample Never Gonna Give You Up after Yung Gravy has had his hit song? We are also, unfortunately, living through a time when interpolations are appearing in the Billboard charts more than I think ever. Are you not stealing a tremendous amount of value if you mock up a soundalike of a famous recording having licensed just the song, and duplicating the recording?
I can imagine a bunch of songs packing the top fifty with soundalike re-recordings of famous choruses that are disproportionately the springboard to the charts. Is that okay? It’s legal. But it’s weird, right?
And AI is going to be able to write songs that sound like Drake and The Weeknd collaborated to create and record a song. They are displaced as both creatives and as performers.
Sure, on some level, all music is derivative. But on another level, is that what we want? And “what we want” is copyright’s purpose, so it’s copyright’s problem to solve.