Can Sampling Survive in the Age of AI?

In the summer of 2017, when Daniel Hansford was 17 and on vacation with his family, “Josie” by Steely Dan came on the car stereo. Though he was hearing the song for the first time, Hansford believed that “Josie” was sampled on “Perfect Love,” a relatively obscure song by one of his favorite artists—the dance music DJ and producer Todd Edwards. But this wasn’t an obvious Steely Dan sample like the ones that De La Soul, Lord Tariq and Peter Gunz, and Kanye West had rapped over. It was just a few notes from a guitar solo chopped into minuscule bites. “I was just freaking out,” Hansford says now. “It made me realize I had an ear for [Edwards’s] music in that sense.”

Hansford went on his phone and pulled up WhoSampled, a website where users catalog songs and their sample sources, mapping an ever-growing and interconnected pathway of influence. He chose the username Danny Shazam and submitted Edwards’s use of “Josie.” It was approved and posted the next day.

The New Jersey–born Edwards is known for his use of micro-samples—fragments of sound that provide texture or percussive elements. On their landmark 2001 album, Discovery, Daft Punk enlisted Edwards to sing and coproduce “Face to Face,” and it was through that song that Hansford initially got into his music.

Months after Hansford’s first sample submission, he and some other WhoSampled contributors started a thread on the site’s message board to track down every sample used on “Face to Face,” many of which remained publicly unknown almost two decades after its release. In October 2021, a Discord user who goes by lobelia started a server called Sample Hunting, which helped broaden the mission. That same month, lobelia used a new piece of software called Hum to Search, which is accessible through the artificial intelligence program Google Assistant, to identify a minuscule guitar sample in “Face to Face” as the Doobie Brothers song “South City Midnight Lady.” (The name of the program is a bit of a misnomer since it can also identify a song through the recorded version or when a person sings it.) While the popular app Shazam often has to tag 10 seconds of a song to identify it, Hum to Search can do it in just a second or two, though its results are limited to what’s available on YouTube Music.

On July 20, 2022, the Discord was chatting about their techniques, and lobelia explained this approach. When a member named DJ Pasta fully realized the possibilities, it set off what became known in the Discord’s lore as “The Night of Many Samples.” Over the course of the night, three “Face to Face” samples, as well as those from dozens of other songs, were identified.

Hansford was pulling a late shift at his job as a part-time manager at an amusement park when it was all going down. “I was like, ‘Oh my God, I wish I could go home and listen to it all,’” he says. “It was insane.”

In February 2023, Tracklib (a platform where producers can legally find and clear samples) ran a story on its blog about the Sample Hunting Discord and what it had been up to. In an era when the rapidly expanding capabilities and use of artificial intelligence are generating both lofty ideas and existential fear, the article sent a tremor through the hip-hop and dance music production communities, where the use of samples isn’t always aboveboard. Legally clearing a sample can be a complicated and costly process. It also means the creator of the new song has to give up a portion of the publishing (its revenue-generating mechanism) to the sampled artist, if it’s approved at all.

“No less than seven producers texted me the [Tracklib] article,” says Thes One, best known for his work as half of the Los Angeles rap group People Under the Stairs. “There was definitely a little bit of panic.”

In the months after the article’s release, membership on the Sample Hunting Discord swelled past 2,100. Several hundred more samples were identified, but there was also conflict. “I’ll be honest, the server was really peaceful before the Tracklib and other articles came out about it,” Hansford says. “People accuse us of trying to harm the artists and whatnot for doing this. Really it just started with us having fun together and just wanting to discover the samples.”

Since sampling as we know it began in the 1970s and proliferated in the ’80s, there have been both a community of people interested in publicly identifying samples and a backlash to this so-called sample snitching. With the spread of AI technology, there is a fear not only that uncleared samples from the past will be exposed, but also that current producers will feel creatively stifled since their most ingeniously disguised samples could still get called out. Yet even as modern sample hunters adopt artificial intelligence, sampling remains far from dead. In fact, in some cases it’s AI that’s unlocked new opportunities within it.

Now living in L.A., Edwards keeps putting out new music. He explains that he initially embraced micro-samples partially to protect himself from litigious record labels as his notoriety grew in the underground, but these micro-samples continue to set him apart from other dance music artists. “You go on the Beatport Top 10, and everyone is just sampling large chunks from things that were already hits,” Edwards says. “When David Guetta is sampling ‘Blue,’ it’s like, you’re David Guetta, you’re already a pop star, why do you need another pop hit to make your pop hit?”

Edwards has been turning more to stem separators, one of the recent technological leaps that AI has enabled. This software has become increasingly better at identifying a specific part within a song—for instance, vocals or a bass line—and filtering out everything else. “That shit, when I got my hands on that, I was just so excited,” Edwards says. “There have been so many times where there were amazing musical samples that I just can’t use because there might be this loud-ass snare in the middle of it.”

Many of the members of the Sample Hunting Discord are also aspiring musicians. As they continue to refine their sample-identifying techniques, they also used their production software’s AI-powered plug-ins and stem separators to isolate individual parts within songs. That way, Hum to Search or Shazam has clearer material to draw from.

Edwards often communicates with fans who approach him online. He’d been aware of Hansford’s YouTube videos that break down how he employs specific samples, and the two were in touch. “Once the [Tracklib] article on the Google AI came out, he’s like, ‘Now you know the secret of where we were finding your samples,’” Edwards says. They both even use the same stem separator, LALAL.AI, although at cross-purposes.

With all the angst and anger that the Tracklib article stirred up, I reached out to WhoSampled to see whether the website had noticed if AI tools had caused an uptick in the number of samples being submitted. They responded that while the discovery of certain types of samples by certain artists (like Daft Punk and Edwards) had increased, when it came to volume, the overall number of new submissions hadn’t changed “to any significant degree.”

At the time of this article’s publishing, WhoSampled lists 20 different songs “Face to Face” incorporates, many of them by mellow rock acts like Loggins & Messina, Poco, and Boz Scaggs. Twelve of them were submitted by Sample Hunting users within the past two years, including a crucial drum sample from an album by jazz flutist Herbie Mann that Hansford found this past May.

On July 1, the member lotusladybug found what is believed to be the final unidentified sample from “Face to Face”—an audio sliver from Rockie Robbins’s “Nothing Like Love.” The words “IT’S OVER” were emblazoned across the thumbnail of the YouTube video that lobelia posted through the Sample Hunting account, announcing the find to the broader world.

The only biographical information that lobelia publicly discloses is that their first name is Erica and they are currently 19 years old. Recalling their reaction on the day when the Robbins sample was found, lobelia writes in a Discord DM, “I had tears in my eyes. I was overjoyed because it had finally been done. It was actually the first thing I saw that day, my phone was blowing up with notifications!”

At the end of September, the Sample Hunting Discord officially closed. It had gotten too big to manage, and in a parting note, lobelia wrote that “the number of bad actors joining the community has also gone up by a lot, and simply put we can’t always keep up with every little drama that’s going on since there are so many.”

At the same time, no other project galvanized the community the same way “Face to Face” did. There aren’t any more samples in the Daft Punk discography whose origins are unknown, although there’s some debate about whether there’s maybe one in “Burnin’” or whether it’s just them playing a synth. At this point, maybe only the robots know.

Obscuring the source of your inspiration has been a part of rap for its half-century history. Though Afrika Bambaataa is now a divisive figure after a series of sexual abuse reports surfaced in 2016, he is credited for expanding the types of records that DJs could play for hip-hop crowds beyond soul and funk breaks. His sets at the dawn of hip-hop, which incorporated everything from German synthesizer sounds to Nigerian Afrobeat to New York post-punk, might sound like an average night out in Greenpoint circa 2004, but they were revolutionary at the time. An often repeated tale is that Bambaataa used to submerge his vinyl in the bathtub so that he could peel off the labels. That way, competing DJs couldn’t get a peek at what he was spinning.

On the flip side, there were bootleg vinyl series like Ultimate Breaks and Beats, Super Disco Brake’s, and Strictly Breaks, which collected popular loops and songs that had been sampled for rap hits. Though Ultimate Breaks and Beats stopped putting out new editions in 1991 after a five-year, 25-official-installment run, you could still find re-presses in hip-hop-friendly record stores more than a decade later.

In the early days, knowing where samples came from could feel like speaking a secret language, for better or for worse. Knowledge was shared by word of mouth between producers and serious diggers. Discoveries were made at events like the Roosevelt Hotel Record Convention in New York and the Pasadena City College Flea Market. But with each development in the digital world, information about samples has become more widespread and easier to access.

In the mid-1990s, one of the first sites I ever went to as I surfed the World Wide Web from my high school’s computer lab was just a text list of the dozens of samples Beastie Boys and the Dust Brothers used on Paul’s Boutique. I printed it out and used to just stare at the words. Now there is a 13-minute fan-posted explainer video on YouTube that illustrates each bit of source material and how it’s used on the album.

Turn of the millennium message boards like SoulStrut created a space for vinyl enthusiasts. But as the raer grooves and secret heat that producers sampled were shouted out in posts, local music sellers realized they could get a higher price for those records from global buyers if they posted them on online marketplaces like eBay and Discogs. Today listeners don’t even need to have the means to track down physical copies; they can just check out these songs on streaming platforms, where other users share public playlists that collect samples from a particular artist or album.

Though some samples were cleared during the 1980s, many producers were testing an unformalized system to see what they could get away with. After the landmark lawsuits against De La Soul by the Turtles and Biz Markie by Gilbert O’Sullivan were settled in 1991, record labels saw the serious financial penalties that uncleared samples could cause, but most of them already knew that the potential hazards were there. Audio collagists Double Dee and Steinski first got the attention of Tommy Boy Records when they entered a remix contest for G.L.O.B.E. & Whiz Kid’s single “Play That Beat Mr. DJ.” Their winning entry became the first part in their influential “Lesson” series of songs. But when Tommy Boy considered putting it out officially, it quickly became clear how expensive and difficult an endeavor that would be. “They were like, ‘Give us a list of what’s in it,’” Steinski says. “So, OK, here’s a list of 45 or 50 different samples. I think they got three or four samples into it, and it was like, ‘Forget it. We’ll never do it.’ There was no structure at that point for dealing with samples. It was just a million billion dollars or ‘Go ahead, fuck you, whatever.’”

Instead, 12-inches of their music were eventually given away by Tommy Boy as promos or hit the open market unofficially. “It was the equivalent of [being sold out of] the trunks of people’s cars,” Steinski says.

In the early 1990s, rap producers like Pete Rock, the Bomb Squad, and DJ Premier continued to revolutionize with their approaches to sampling. But their use of multiple samples within one song meant that the pie of the publishing had to get cut into more and more slices. And as hip-hop became a greater commercial force, the people who controlled the publishing for the songs being sampled realized they could demand higher and higher figures. In response, there was a rise in synthesizer-based tracks from producers like Timbaland and Missy Elliott, Swizz Beatz, and the Neptunes. “It really changed the sound of hip-hop,” Thes One says. “It made this sort of class system where either you were P. Diddy and you could clear something ridiculous, or you were a smaller guy and you had to make a record with no samples.”

People Under the Stairs started releasing music on independent labels in the late ’90s, as this change was happening, but the group held steadfast to the sampling approach practiced by their idols. Their 2002 album, O.S.T., was the first time they tried to clear their samples. They did not get far. “When people found out that a sample was used for it, they were just completely unreasonable,” Thes says. “The only sample we got done was we cleared a record that was owned by the Brunswick bowling ball company.”

He’s now come to see the potential danger of a lawsuit for an uncleared sample as intrinsic to his approach. It’s something that differentiates his compositions from those that have all the paperwork sorted. He compares those musicians to graffiti artists who transition into gallery work. “I think that’s great, like, get your money, do your thing, but it’s not graffiti anymore, it doesn’t have the risk,” Thes says. “To homogenize the whole thing and take the risk out of it, to me at least, it was losing a lot of its original essence and a lot of what made it amazing.”

In the early 2000s, Thes became an active user on SoulStrut. He enjoyed the community, but it caused some IRL problems. Sometimes, after board members identified obscure sample sources that People Under the Stairs and other indie artists used, owners of specialty reissue labels would track down the original artists and buy the publishing in cash for a relatively small amount. Then they’d threaten the sampling producers with copyright violation lawsuits unless they paid them off.

“Originally, the existential threat was ‘Oh God, I can’t sample something on RCA because RCA’s gonna come after me,’ but it turns out the reality was RCA didn’t give a shit unless it sold a million records,” Thes says. “It wasn’t at all in their interest or time to come after a guy who released a 500-[copy] pressing of a 12-inch. But the people who wanted to blackmail and run a shady game against producers, that’s who we had to worry about.”

When a website Thes declines to name started listing People Under the Stairs samples, he spoke to them about having the page taken down, but they refused. Then he noticed that it said his group had sampled the Police, whose frontman, Sting, is notoriously litigious about sampling and has ended up with most of the publishing for hits like Diddy’s “I’ll Be Missing You” and Juice WRLD’s “Lucid Dreams.” The problem was that the listing was a misidentification and PUTS never sampled the Police. “I told [the website], we can all agree that [unauthorized] sampling is still a crime, it’s copyright violation or whatever,” Thes says. “You’re basically accusing me of a crime that I didn’t necessarily commit on a public forum, so I feel like I might have legal standing here to stop this.” This tactic got all the PUTS sample listings taken down, but over the years, they’ve started to return.

Like that Police sample misidentification, artificial intelligence programs aren’t infallible either. They can provide inaccurate or sometimes completely fake information. “Let’s not forget, every program that’s created is programmed by a human, so there’s always errors and flaws,” says Deborah Mannis-Gardner, the president and owner of DMG Clearances. Mannis-Gardner has worked in the music industry since the 1990s, and her company has become a go-to source for clearing samples, working with superstars like Drake and Kendrick Lamar, as well as helping bring previously unavailable projects by De La Soul and Mac Miller to streaming platforms. “The bottom line is we’re here to try to make sure that samples are cleared if someone has utilized them, whether it’s before the fact or after the fact, to make sure that all the copyright holders are financially compensated,” she says.

As of now, DMG Clearances has not used AI programs to identify samples. “We use our ears,” Mannis-Gardner says. She explains that the programs aren’t precise or nuanced enough for her yet and can give incorrect results. At this point, there are decades’ worth of songs that use samples, songs that interpolate other songs, and songs that sample songs that use samples or interpolations. It can get complicated.

That said, Mannis-Gardner is engaged in conversations about AI and what it could mean for her world. She notes that too often, the music industry isn’t proactive in reconciling its relationship with new technology. “When we started talking about Metaverse, Web3, and NFTs, the labels and the publishers were like, ‘Oh, we’re sitting on the sidelines,’” she says. “You know what? Stop sitting on the sidelines. You need to get involved. It’s not going to go away, because that was the position 30 years ago with the world of sampling. That was the position even before, when CDs came out.”

Now 68 and 72 years old, respectively, Double Dee and Steinski don’t make much new music anymore and said they haven’t experimented with AI-enabled music software. But as Dee turned some ideas around in his head, his eyes glinted with mischief through the Zoom screen. He pondered the possibilities of creating a drum loop that sounds like “The Crunge” by Led Zeppelin, but played by Ringo Starr. “Now you’re creating something that still can have that thing, but is not a legal problem because it never existed before, it doesn’t belong to someone,” he says. “That would be pretty incredible.”

Maybe it would be incredible, but Mannis-Gardner doesn’t think it would, in fact, be legal. I emailed her to get her thoughts on this potential usage, as well as on another hypothetical situation in which a producer uses AI to create a beat that sounds like it uses an Earth, Wind & Fire sample, but the original EWF song doesn’t actually exist. She replied that both examples “pinpoint why we need to regulate AI in music directly rather than just adapt existing licensing structures to this new technology. People will always come up with creative new ways to avoid paying artists what they are owed!”

Working off the belief that “the resulting AI-generated track would be so intrinsically tied to the source material that the original rights-holders should also hold a share of the AI-generated track,” she continued that, to execute Double Dee’s idea, the AI would need to incorporate information from the original recording of “The Crunge,” as well as multiple instances of Ringo Starr’s drumming—all of which would need to be cleared. And for the Earth, Wind & Fire beat, you’d have to license every EWF song used by the AI to emulate the group, as well as the resulting AI-generated song. “It’s better and likely cheaper for people to license their samples correctly in the first place rather than trying to take a shortcut using AI,” Mannis-Gardner wrote.

While sampling has persisted through the various strata of hip-hop this century, its popularity in the mainstream ebbs and flows. The current climate most recalls the 1997 era, when producers like the Trackmasters and Bad Boy’s Hitmen were gettin’ jiggy with huge chunks of recognizable disco and 1980s pop loops for rappers and R&B singers. The difference is that now artists like Saweetie, Jack Harlow, and Armani White are often interpolating rap singles from the early 2000s while racking up hundreds of millions of streams.

Toby Oniyitan is the young CEO of Stomp Down, a label and management team based in Houston. He says he loves songs that use samples and going through playlists to hear how old tracks have been flipped, but he has issues with the latest wave. “I feel like it’s all very on the nose,” he says.

Nevertheless, Oniyitan recognized the possibilities in this approach and realized that no one was resurrecting regional Houston smashes from the Y2K era. He had the teenage producer Merion Krazy create a beat based on Yungstar’s “Knocking Pictures off Da Wall,” then tried to convince multiple nationally known artists to rap on it. Oniyitan was sure it was a hit, but it got no takers, and the track languished. Then Merion gave the beat to Monaleo, a local artist with just a few Instagram freestyles out in the world, who turned it into “Beating Down Yo Block.” It quickly blew up around Houston, then the rest of the country.

Artists often informally test-market new songs by previewing snippets on their social media accounts, as Monaleo did with “Beating Down Yo Block,” or let their teams surreptitiously leak them. The problems come when the fans demand the whole thing, but the sample hasn’t been cleared yet. Sometimes it can take months, as was the case with the City Girls’ “Twerkulator.” “In the time period we’re in now, everything moves fast,” Oniyitan says. “We’re not waiting for CDs. Things are going from the internet to people’s ears, to people’s hearts, and you have to move on that moment. Sampling slows things down because you have to wait. And if you don’t wait, if you try to press your luck, [record labels] could take you for everything.”

Oniyitan says that Monaleo is now hesitant whenever he suggests new beats that incorporate samples because she saw how prolonged the process of clearing them can be. “It’s always a nasty fight,” Oniyitan says. “Very rarely have we had something sampled and it’s just smooth sailing.”

Yungstar cleared “Knocking Pictures off Da Wall” without any issues, but his song includes a sample of Kris Kross’s “Da Streets Ain’t Right,” which was produced by Jermaine Dupri and released on his label So So Def, a subsidiary of Columbia Records. Columbia is owned by Sony, so the megacorporation became involved. In turn, “Da Streets Ain’t Right” features a sample of the Notorious BIG’s “Warning,” which sampled Isaac Hayes’s “Walk On By,” which is a cover of a song first performed by Dionne Warwick. And that’s why, on “Beating Down Yo Block,” Monaleo shares songwriting credits with Burt Bacharach, among others.

Oniyitan maintains that it still makes economic sense to use a sample to make a song hot, despite the financial consequences. “Even though the artist may get less on publishing, the amount of money you make from shows, the amount of money you can make from your appearances, your brand endorsements, just the celebrity of a big record alone makes you money,” he says. “I’ve always died by the philosophy that a small percentage of something’s better than zero of nothing.”

Even hip-hop subgenres that were once defined by not using samples have started embracing them. Artist and producer Cash Cobain, one of the rising stars of New York’s drill scene, is known as “the Sample God.” Building off the sound’s skeletal framework, he fleshes out his beats with reinterpretations of J. Holiday, Plain White T’s, and Jai Paul cuts.

At 25 years old, Cobain is an artist firmly of his era. He still makes his music on FL Studio, the software he started out on as a teenager. He used to go on YouTube to watch tutorials about Metro Boomin, Zaytoven, and Young Chop productions. He admits that he’ll hit the internet to find the sample sources of songs he likes so that he can track down the original material and possibly flip it differently. “I’m on WhoSampled today,” he says. “They saved my life, for real.”

While previous producers got their fingers dusty looking for inspiration in stacks of old vinyl, Cobain often relies on AI-powered song recognition apps. “I could be in a restaurant, I pull out my phone, Shazam,” Cobain says. “I could be in the mall, any type of song, it don’t got to be R&B, when I hear it, I know I can sample this, I can make it sound sexy. I could be in an Uber or in the car and going through [radio] stations. I could be watching YouTube and an ad comes on and I like the music. I like it when it comes to me instead of trying to look for samples.”

In the 1990s, Alan Braxe emerged from the same French Touch scene that fostered Daft Punk. The producers from this community were inspired by hip-hop’s approach to sampling, except they bumped up the tempo to a disco-fied 120 beats per minute. Speaking over Zoom from his home recording studio, he motioned to a decades-old E-mu SP-1200 behind him. “For me, sampling is using this kind of machine,” he says. “It means sampling preexisting music, editing it, truncating it, splicing it, playing it in a different order, reversing it. All kinds of stuff, which makes you feel like you are splitting wood or something with your hands.”

Braxe has continued releasing new music and updating his sound through the decades, but he admits he often feels overwhelmed by how many options he now has. With a relatively small budget, he can replicate basically any instrument or equipment through software or plug-ins. “I got lost in the computer,” he says. “Your working frame is constantly evolving, so you cannot focus because you can’t learn properly. Each week you buy a new compressor or a new library. It’s too much to process in your mind.”

He contrasts this limitless possibility and almost infinite data storage capacity that musicians now enjoy to the early 1980s, when Kate Bush made her song “Babooshka.” On it, Bush used a Fairlight CMI to sample the sound of breaking glass and layered it to create a percussive element—an entirely innovative approach at the time. “This sound is part of pop music history now,” Braxe says. “What is interesting is the size of this sound in terms of megabytes. It’s not even megabytes, it’s kilobytes. It’s nothing.”

Braxe hasn’t gone deep into new artificial intelligence programs to make music yet, but he’s been exploring. What intrigues him isn’t the potential they open up, but the focus they can create. He cites an affordable program called Waves and its AI sample finder function, called Cosmos. It’ll scan your drive for audio files, then classify, map, and color-code all your sounds before putting them in a searchable database. “The problem now with the computer world is just to have access to all the files you have,” he says. “There is too much; you forget almost everything.”

The potential for using AI to generate actual songs has created plenty of speculation and hand-wringing in the industry. Producers, engineers, and songwriters are already using AI as a tool in ways that aren’t always obvious to listeners, but in terms of a purely AI-made final product, there hasn’t been a creative breakthrough yet. The Weeknd- and Drake-mimicking “Heart on My Sleeve” from earlier this year might have generated headlines, but it and other similar tracks feel more like party tricks, reminiscent of the deluge of conceptual mashups that came out after Danger Mouse’s The Grey Album.

In 2022, the long-running English electronic music duo Plaid collaborated with Emma Catnip to utilize AI to create the visual art and videos for their album Feorm Falorx. Ed Handley of the group notes that they’ve also explored some AI music-generating technology, but the results have been uneven. For every interesting notion it creates, there is far more derivative, disappointing material. “Those particular tools, some of them are just very much like a slot machine,” he says.

Even if AI-generated music doesn’t make its way up the charts, it most likely will make its way to us through other mediums. “There’s a lot of interest in the industry to generate new content,” notes Andy Sarroff, the senior research manager at the audio company Native Instruments. “Some of those reasons have to do strictly with the bottom line. If you want to use content for a reason, whether it’s a movie or a TikTok, you have royalties to pay. There are a lot of companies that are interested in being able to generate believable musical audio that is royalty-free.”

Handley believes that this outcome is inevitable, but that it’s critical to realize in this pre-Skynet era that the potential enemy isn’t AI technology itself; it’s who wields it. “The danger obviously is that the big, terrible corporations decide they want to get some cheap, free music and they just put their prompt in and they press go,” he says. “I mean, [AI is] coming for everyone. It’s coming for coders, it’s coming for journalists, it’s coming for all kinds of these jobs. We have to figure out a way of using it rather than it using us. But it’s not going to use us. It’ll be the usual people using us.”

Sarroff believes that what’s often presented as examples of artificial intelligence in music creation and identification isn’t really AI at all. It’s actually machine learning, though that phrase doesn’t have the same marketing zing that AI has right now. “We’re really just using some very interesting and complicated ways to analyze large amounts of data to extract patterns that are very difficult for humans to extract, and then making decisions based upon that,” he says, “which is a slightly different thing than responding to a dynamic, changing environment that is unpredictable and making decisions in that environment intelligently.”

Which is a pretty apt description of the challenges of trying to navigate, you know, existence.

For Braxe, it’s this discrepancy that gives him hope, not fear, about the future of music. “I’m not sure [computer-generated music] is problematic because it’s going to force us to dig deeper into what’s human, what’s almost impossible to develop as an algorithm for now,” he says.

As music and technology continue to evolve, artists dedicated to sampling (both legally and illegally) have proved themselves to be incredibly adaptable to the times. “The Music Genome Project really changed sampling, Shazam changed sampling,” Thes One says. “[AI’s] been coming for a long time. Now in terms of what the effect is going to be, it all depends on how people want to weaponize it.”

It’s been estimated that 100,000 songs are uploaded to digital service providers a day. That sheer volume makes it hard to imagine that the major record labels will dedicate the time and resources to running each one through programs to detect unauthorized samples, especially given the lengths that members of communities like the Sample Hunting Discord still sometimes have to go through to identify them. Still, inevitably, some artist will get caught out there for uncleared usage because of new AI-enabled technology, by either the legitimate rights holder or an opportunist looking for a payout. If it happens enough times, the sampling tactics will just be adjusted again. And again. And again. The cycle will keep spinning like a well-worn copy of a familiar breakbeat.

“There will be more tools to detect the sample usage, but there will also be more tools to go against the law,” Braxe says. “So who’s going to win? It’s a pure game, if you see it this way.”

Eric Ducker is a writer and editor in Los Angeles.