The Lines Blur Further: How AI-Generated Music Hits in a Litigious, Sample-Friendly World

The music industry has had to navigate the choppy waters of royalties, ownership, and inspiration well before AI-generated music became the hot new thing. And not just because of sampling.

I think, in starting this piece about sampling, we need to talk about a song that doesn’t have a single sample, but has probably done more to reshape the discussion about music and copyright in the 21st century than any other.

That song? “Blurred Lines,” a tune that Pitchfork recently described as a “harbinger of doom.”

The vibes around Robin Thicke’s collaboration with Pharrell and T.I. grew increasingly problematic over time—creating huge personal problems in Thicke’s life, and business problems for everyone involved with the track.

One of those problems came from its point of inspiration, Marvin Gaye’s “Got to Give It Up.” The fact that, musically, it borrowed so many ideas from the bass-driven, chatter-filled recording turned into a huge liability for the recording industry as a whole, as Gaye’s estate sued—and won. Now, it’s become increasingly common for pop songs to preemptively offer credit to artists any time a song seems to directly borrow from a source too aggressively—and arguably has led to a huge push by publishing companies to buy out name-brand artists.

But before “Blurred Lines” came along, sampling set the stage for famous artists (or more likely, their estates) to take a litigious stance towards new generations of musicians.

As artificial intelligence promises to bring a fresh new stage to this discussion, it’s worth discussing why sampling created this arms race in the first place. Will the record industry finally meet its match in artificially generated hip-hop? (And is that a bad thing?)

The technical roots of what (eventually) became a legal problem

If you want to figure out why the estate of Marvin Gaye was able to do such damage to modern-day songwriting, you have to go all the way back to the player piano. Before computers, before tape, even before 78s, player pianos threatened to disrupt the livelihoods of composers; simply put, manufacturers weren’t paying for the composers’ songs.

It was a big enough issue that the Supreme Court got involved, and much like many other decisions involving emerging technology and the rights of creators, the court arguably biffed it, ruling in 1908’s ‌White-Smith Music Publishing Co. v. Apollo Co. that mechanical reproductions did not represent copyright infringement.

“These perforated rolls are parts of a machine which, when duly applied and properly operated in connection with the mechanism to which they are adapted, produce musical tones in harmonious combination,” the ruling stated. “But we cannot think that they are copies within the meaning of the copyright act.”

This decision led Congress to quickly change laws around mechanical production while creating something much more sustainable—a law that not only required payments for mechanical copying, but allowed anyone to copy anyone else’s song, a strategy explicitly designed to prevent attempts to monopolize access to music catalogs. As Slate noted in 2014, this set the stage for cover songs in all their fascinating and messy forms, but also made it possible for royalty systems to gradually emerge.

Over time, these royalty systems started to get complex—in part because the technology started becoming more complicated than mere player pianos. The law got more complicated, too.

How sampling technology evolved into a creative medium of its own

Gradually, technology with more moving parts than player pianos began to reshape the way that we thought about copyright and ownership. And that helped to set the stage for other innovations that involved people playing other people’s music.

While sampling is often portrayed in the context of record scratches and computer trackers, its roots go back a little further, with the chop-up-and-put-together approach of musique concrète, which used existing recordings as raw material, generally seen as where the concept of sampling began. That tape-based technique, dating to the 1940s, found perhaps its purest artistic expression with the Dr. Who theme song, which BBC Radiophonic Workshop employee Delia Derbyshire built from mashed-together segments of analog tape.

That theme song’s later life shows how quickly the discussion around sampling had changed. In 1988, about a quarter-century after it was first recorded, a novelty pop song named “Doctorin’ the Tardis” found major success by crudely mashing it up with Gary Glitter’s “Rock and Roll (Part Two)” and Sweet’s “Blockbuster!” That song topped the British charts despite the fact that critics hated it.

Bill Drummond and Jimmy Cauty, who recorded that song under the name The Timelords (and would become far better known as The KLF after that point) even wrote a book about how they did it.

But that wasn’t sampling’s first appearance on the pop charts—far from it. In fact, while we wouldn’t consider it as such today, notable bands like The Beatles and The Moody Blues technically used a sampler on some of their best-known recordings in the form of a Mellotron, a kind of keyboard designed to control analog tape recordings. (One difference: These “samples” generally weren’t of other pop songs.)

Over time, more techniques emerged. In Jamaica, the technique of dub reggae, which relied on instrumental recordings of existing songs that gradually evolved into new types of recordings with the removal of vocals and instruments, the integration of studio effects, and the addition of echo and reverb. Some of these techniques, associated with figures such as Lee “Scratch” Perry, would soon find their way to hip-hop.

At first, hip-hop leveraged the power of the “break,” an instrumental section from a recording, as a starting point for a new type of music. Evolving from block parties in the Bronx, hip-hop’s early music took a distinct influence from popular recordings of the era—which meant that, because disco was dominant, early hip-hop sounded like disco.

“Rapper’s Delight,” for instance, was built on top of an interpolation of Chic’s “Good Times,” which nearly led to a lawsuit and the eventual addition of Nile Rodgers and Bernard Edwards to the song’s writing credits. The song did not use a sampler, but rather live instruments, because samplers were uncommon at the time.

But that would soon change. Hip-hop artists were increasingly becoming skilled in chopping up sounds from vinyl, and technology was catching up to their ambitions. In 1979, the Fairlight CMI, the first commercially available sampling synthesizer, first hit the market, and while extremely expensive, it was significantly more accessible than earlier attempts, which relied on minicomputers.

The technology quickly got smaller and cheaper—keyboards gradually gained sampling capabilities, and by the late ’80s, dedicated sampling devices like the Akai MPC60 became both common and easy to come by, making it possible to treat sampling as an artistic form in its own right, rather than a quick way to borrow a recording.

Artistic innovation meets legal conflict

By the end of the 1980s, sampling was having a significant impact on the artistic medium of hip-hop. Early collage-driven samplers like Double Dee and Steinski gave way to mainstream acts like The Beastie Boys and De La Soul borrowing from these same techniques.

But the legal picture complicated matters significantly. De La Soul famously fought the ’60s folk-rock band The Turtles in court, and a 1991 lawsuit against rapper Biz Markie, who borrowed heavily from the Gilbert O’Sullivan song “Alone Again (Naturally)” on his 1991 track “Alone Again,” led the industry to begin clearing samples, when possible, before release. (Markie made fun of this situation with his next album, All Samples Cleared!)

These legal complications became problems decades after the fact, especially for sample-heavy recordings produced prior to the lawsuit. De La Soul, for example, struggled to get its catalog onto streaming services in part because of issues with clearing samples, and before the catalog appeared back in March, were required to replace some of the found sounds with interpolations.

Despite this, musical innovation has largely been improved by the rise of sampling, with new types of hip-hop, such as trap music, emerging over time as new approaches became common and tastes changed.

Additionally, the rise of mixtapes, free recordings produced by hip-hop artists that often sport uncleared samples, proved a hugely effective way to allow emerging rappers to draw attention without having to go through the costly process of clearing samples. Chance the Rapper, for example, was already a mainstream pop star based on his mixtapes alone, and only released a traditional commercial album seven years after his first mixtape.

Others have gotten quite creative with their ways of working around the compensation issue. Gregg Gillis, a DJ who goes by the stage name Girl Talk, gained extreme prominence on the back of his brazen sampling of mainstream pop songs, which he distributed online as free albums through his label, Illegal Art.

In many ways, sampling reflects a conflict between two distinct artistic forms of music. One part might be built from original instruments, and another that takes those parts and builds something new. Some have become quite good at it. Kanye West, before becoming a prominent lead artist in his own right, won notices for his inventive use of sampling. 

But its roots in the creations of others mean that there’s always a risk of a lawsuit around the corner. West, for one, has been sued numerous times for samples on massive hits, with the samples sometimes being so obscure that for the source music’s creators, it was their first actual brush with fame. For example, “Bound 2” sampled the Ponderosa Twins Plus One, a group that released one album and had otherwise long been forgotten—and whose lead singer had to sue to get royalties from West’s label.

Some of these suits have gotten messy, especially when the original creators are based in other countries or have died. A family member of the Egyptian composer Baligh Hamdi attempted to sue Jay-Z and his producer Timbaland over the 2000 song “Big Pimpin’,” which used an unauthorized sample of Hamdi’s “Khosara Khosara.” (The artists thought the song was in the public domain. It wasn’t.) The family member ultimately lost in court, but not before the suit had dragged on for multiple years.

But “Big Pimpin’” also shows how samples can actually at times benefit the musicians being sampled in non-compensatory ways. The musician who actually performed the sampled track, Hossam Ramzy, worked increasingly with Western artists such as Ricky Martin and Shakira after the song further raised his global profile.

How AI music reflects the same kind of disruption as sampling

Now, nearly a quarter-century after “Big Pimpin’” became one of Jay-Z’s best-known songs, the legendary rapper finds himself in a very similar kind of situation—one where an emerging technology threatens to take away the thing he is best-known for, without his consent. In 2020, Jay-Z used YouTube copyright complaints in an attempt to remove deepfakes of his voice from the internet, a use of that system that blogger Andy Baio noted was “questionable.”

And the technology has only gotten better since then. AllttA, a production team, used a much better version of Jay’s voice on a track called “Savages,” which listeners found strangely compelling, given the source. As one YouTube commenter put it, “I think this track might be the tipping point that wakes people up to how powerful AI has become in the last few weeks.”

Kanye’s in this same spot, too, with at least one unusual example involving an AI-generated West belting out a cover of the Plain White T’s “Hey There Delilah,” in a style not far off from his massive hit “Runaway.” And Drake and The Weeknd have had to chase after a hugely popular AI-generated banger called “Heart on My Sleeve” that strangely keeps getting deleted from the internet.

Charitably, there are genuine concerns about the technology taking away an artist’s essence and handing it to a large-language model. Less charitably, one could argue that it’s amusing that these artists, who spent literal decades doing this exact same thing—at times, long before legal precedent caught up with the times—are now stuck with the same problem that their predecessors had.

Perhaps it’s for that reason that it’s not really a surprise that Universal Music is already pushing its weight around in an effort to rein in AI-generated tracks.

But it doesn’t necessarily have to be quite so bad.

One artist who seems to have found an innovative path forward to this problem is Grimes, who is allowing people to make music using her AI-generated voiceprint—then offering them half the royalties, along with distribution, an innovative model that will likely go a long way to validate AI-generated music if it’s effective.

In a way, artificial intelligence could become just as important a tool for music as the Mellotron, or the Akai MPC60, or Pro Tools. It’s disruptive and separates creators from money, and leads to situations where reaction through the courts is the only option.

But on the other hand, we’ve been trying to solve for this exact problem since the player piano. Odds are, we’ll figure it out eventually.