How Spotify Sees the Next Era of Music Streaming w/ Sam Duboff
67m 15s
In this podcast episode, host Kristen Robinson interviews Sam Duboff, Spotify’s Global Head of Marketing and Policy for the music business, about the company’s recent changes. Duboff discusses Spotify’s strengthened AI policies, which target spam, impersonation, and deception, and include a self-disclosure system for AI use. He explains that AI tools have increased the scale and sophistication of spam, leading Spotify to remove 75 million spammy tracks and implement a music spam filter. Spotify is also collaborating with industry partners to develop responsible AI music products based on four principles: upfront agreements, artist opt-in consent, proper monetization, and deepening artist-fan connections. Duboff emphasizes that these tools are designed to complement, not replace, human artistry by enabling fans to connect more deeply with their favorite artists. Additionally, Spotify is rolling out new features like direct messaging, Mix of Spotify, music videos in more countries, and high-fidelity listening, evolving from a static streaming platform into an interactive space for artist engagement. The company is also working with DDEX and distributors to standardize metadata and ensure transparent AI disclosure, aiming to maintain listener trust. Overall, Spotify is navigating the challenges and opportunities of AI in music while prioritizing artist control and fan connection.
There's been a lot going on at Spotify in the last few months. You might have heard that Daniel Eck, the company's founder and CEO, has stepped down from his post, triggering a shake-up in the company's top brass. And you might have also heard that Spotify updated its policies on AI. This is very eagerly awaited by the music business, given the company has come under fire for not having any AI-specific policies in place yet to date. And to be more specific, Spotify strengthened its rules against spam on the platform, impersonation of artists, and created a system for self-disclosure of AI use. And while somewhere very grateful for this, others felt that the strength and policies did not actually go far enough. To be fair, it's hard to know the right way to treat AI music on streaming services. It's something that all of them are grappling with. If you demonetize AI music, for example, are you creating an incentive for a streaming service like Spotify to plug AI songs in their playlist recommendations, given they're all now free? Or do you ban them from the platform? But what if AI songs become part of popular music culture? And also there are significant dangers of doing nothing about it too. Like, what if someone uses AI to copy your voice in a song and tags it to your page, confusing all of your fans? To top it off, Spotify has been rolling out a series of new features as well. Like direct messaging, something called Mix of Spotify, which allows you to create DJ-level transitions in between songs, music videos in more countries, and finally, high fidelity listening. So today, let's unpack that. Hello and welcome to On the Record, a new music business podcast from Billboard and Sickbird Productions. I'm your host, Kristen Robinson, and today we are in New York City talking about the future of Spotify with one of the company's top executives, Sam Duboff, global head of marketing and policy. A little over a decade into streaming's dominance, it seems clear that Spotify is starting to evolve its service in meaningful ways. So I'm grateful that Sam Duboff, who has played a leading role in many of these changes, was down to come on the show and talked to me about all of them. Without further ado, let's bring on Sam. Sam Duboff, thank you so much for coming to On the Record. So happy to be here. I've been loving the show so far. Thank you. Thank you. I wanted to first ask you, so Spotify has really long titles for employees. So can you tell me what your title is and what it means? We've got some crazy titles on the global head of marketing and policy for the music business, which is a lot of words to just say. My team kind of works on all the services and products we offer to the music industry. Spotify for artists is our flagship artist product, but then policies around royalties, AI, artificial streaming, and education efforts like our masterclass series, LatinClear, our royalty report, anything that we're providing to artists, songwriters, producers, labels, managers, publishers, the whole industry side of the equation. Wow. Okay, that's a lot of stuff. Do you have a lot of people under you? I would imagine. Yeah, I've got a great team that helps do all the work. It's a very passionate team of people who love the music industry. So what is one of your favorite Spotify playlists? Great question. Right now I've been really into Laura. Classic. Yeah, take it wrong. And yeah, the Laura team just started a substack too. I think it's our first kind of playlist newsletter. It's a monthly newsletter and amazing. They're doing such a good job with it. And yeah, always love Laura. I wrote into the Laura substack. They were like asking for people's like favorite new artists or something and I did make I did make this a stack. What would you say? I don't even remember at this point. It was like a couple months ago, but Lizzie Sabo, who does that playlist is amazing. Yeah, she's amazing. I think my favorite Spotify playlist right now is CPH plus. Have you heard this one? No, what is it? It's like the Copenhagen scene of indie music. It's very cool. Lizzie also does that one. Love it. Yeah, shout out Lizzie. Hopefully she'll see this, but I'll send her a slack later. Yeah. Well, okay. So I've so much to ask you about. I feel like Spotify has been changing really quickly. There's a lot of announcements. Industry is changing quickly too. Yeah, working for trying to keep up. Yes, I feel like every week I'm always chatting with someone at Spotify about like a new announcement that you guys have coming up. So I want to get to like the new features that you guys have rolling out. I also want to get to these new AI policies that you've rolled out. But first I wanted to hear a little bit more about the fact that you're developing generative AI with the consent of the music industry. I just want to hear everything about that. Like what's going on there? Yeah, we have been hearing from artists and their teams for a few years now that a lot of the emerging music AI tech products don't feel like they're built for them, not built to power. They're businesses, they're careers, they're existing fan bases. So we recently announced we're going to be collaborating with some of our top industry partners across major labels and Indies to collaboratively develop artists first responsible AI music products. Like what could that look like? We want to do this in consultation with the industry. Okay. We'll be able to talk to artists about it, songwriters about it, industry partners about it, and it feels like a lot of the principles about AI music and what these tools should look like. It's kind of happening in real time. So we didn't want to wait until we have a product ready for a big launch to start talking about how we're going to build AI products. We wanted to talk now while we see lots of other folks in the industry are investing in the space to be clear about our principles and how we're going to work with the industry for any product we build. So we're looking at four key principles. We outlined first always upfront agreements with the music industry, not using tons of music and you ask for forgiveness later. Second, we want to make sure artists, songwriters, rights holders have agency and choice about how their music does or doesn't participate in these tools. They should have control and choice around how fans can or can't interact with their music using AI. So we'll always have that baked in. Okay. So like an opt-in kind of thing? Yeah, exactly. We think it should be up to artists, songwriters, producers, their teams for whether for their own projects. They decide to use AI or not and how and for how fans can interact with their music. So, yeah, opt-in consent based approach. Third, we'll always have proper monetization and compensation built in. So artists, songwriters, rights holders always compensated for all uses of their work, properly credited transparently. It will have an eye towards building new revenue streams for the music industry. Okay. So not just splitting up the existing royalty pool in new ways, but finding new incremental growth, incremental revenue streams for the industry, which we think could be really important for powering this next stage of what the industry looks like. Interesting. And fourth and really important to us when I think about our role right now in music is we want to build AI music products that deepen existing artists fan connections. You know, we believe, especially with 700 million monthly listeners coming to Spotify already to listen to their favorite artists. We can play this really unique role where we build tools that help fans go deeper with their favorite artists, kind of connect with their favorite artists in new ways, and make sure AI tools aren't there to kind of compete with artists or to try to replace human artistry. Got it. So those are the kind of four principles we wanted to lay out, and we were excited to have our partners on board to work with us in that way. So we're starting there, trying to make it clear that's what our approach is going to be. That's what principles we think are important. And then the coming months, we're going to be working closely with artists, songwriters, industry to figure out exactly what these products are going to look like. Yeah, I feel like that's a huge move. And I'm wondering, I know that it's still very preliminary, and we're not totally sure I was going to take shape yet, but you talked a bit about how this will increase the connection between fans and artists. Tell me if I'm off base, but it kind of sounds more like yellow leaning towards maybe like AI-powered remixing of current songs rather than a, you know, audio model of like generating a new song from scratch, right? Yes, yeah, I think we see our role as, you know, the biggest streaming home for professional artists today, right? We know who fans love. We facilitate those connections between artists and fans through their music already. So when we think about where we're best positioned to help have AI kind of power this next stage of the industry in some ways, it's just in that space of existing artists fan connections, building on artist catalog with their consent. And yeah, not tools that are built to compete or kind of siphon off from parts of the industry. Got it. It does feel like this kind of signals a shift for Spotify because Spotify has always been seen as the big destination for music consumption. And now this is like a much more playful, interactive, almost like a creation focused tool. Do you guys see yourself expanding from just being a place for static streaming? Yeah, over the past two years, I think we've been evolving Spotify from a place that's just about the music to giving artists all these tools to share the world around their music. So three, four or five years ago on Spotify, you'd get an artist profile with some pictures. We had Canvas, which are the looping three to eight second visuals you see in the back of songs. That was kind of it. And mostly it was just about the music. And then you'd have to go to social media or elsewhere to experience the artist's broader world. Or we've been focusing is bringing in artist clips so that artists can share 30 second videos, sharing the meaning of the song and what their vision was in their creative process, music videos and live performance videos. We've launched in over a hundred markets outside the US. We're working to bring that to the US countdown pages. So when you're. I love the countdown pages. Yeah, they're really taking off. And you know, in that build up to your album release, you can share kind of teasers and clips. You can do a big track list debut. You can sell your merch in advance. We're seeing artists use that in really creative ways. So we've already been on this journey of baking Spotify, a place where fans can experience the whole world of an artist, deepen that artist's fan connection. Like we're talking about. about, and fans already can interact with their favorite artists and with, you know, make their own playlists and all different unique ways or talk to our AI DJ interactively. So we've been building tools like that and, you know, these AI music principles we see as a continuation of that sort of philosophy. Interesting. And also recently Spotify has kind of strengthened its AI policies or its policies around AI music on the platform. For those who might have missed that announcement, can you break down what those policies are? Yeah, absolutely. As I mentioned, we're talking to artists and songwriters and the industry all the time about their hopes and fears about AI and there's equal parts of both. You know, what we hear is at its best, artists are excited about AI, hopefully unlocking and enabling really cool artistic expression they couldn't do before. Our fans get into connect with their music in new ways. At its worst, we know AI will enable bad actors to try to game the system more aggressively with more sophistication. And as Spotify, when we think about our role, you know, we don't police the creative process of artists. We want to leave up to artists and producers and songwriters. If or how they want to incorporate AI tools, we leave that choice up to them. But what we do think is important for us to do is to aggressively protect against the bad birds because if we don't aggressively protect events, some of those bad outcomes of AI, it probably isn't even possible to enable some of the good parts because there'll be so much kind of spam and slopp on streaming services, crowding out authentic artists building their career. So we announced some of our first steps to combat spam, impersonation and deception in the age of AI a few weeks ago. And we know we have more work to do as AI tools develop. But you know, no matter where AI tools had, we had, we know those three areas are places we need to act now aggressively, decisively to stamp out. Yeah, well, what I'm really curious about, I think for some people, they might really not realize how often AI music is used in kind of bad actor type scenarios, whether it's like streaming fraud or just trying to game the system. That sure really hurts artists who are trying to be on the platform and trying to be providing their authentic selves there. So can you talk about how AI music can be used in like nefarious ways? You know, artificial streaming, we almost think of as a separate thing. Okay. So certainly predates modern gen AI tools and it's like we take really seriously, we've invested a ton, especially the last four or five years to build a big technology department that works to identify, mitigate and remove the impact of any artificial streams. AI tools, whether it's AI generated music that you then artificially stream or kind of AI scripts or tools to help you do artificial streaming on non AI music. Certainly, that's going to create some new challenges for our artificial streaming teams and we're kind of working aggressively on that front. Every time we catch artificial streams, royalties aren't paid out, we remove the impact from public metrics like monthly listeners or all-time track streams. We alert labels and distributors so they know about all the artificial streaming happening. So AI definitely adds some complications, but we see that as just a continuation of our efforts there. Yeah. What we do is when we kind of talk about spam and these gaming the system behaviors where it might not be artificial streaming, but kind of creates an unfair playing field. So we think about things like mass uploads, how easily now with AI could you upload a more trove of songs that hope a few of them take off or SEO hacking? Can you use AI to automate changing the metadata and your track over and over until you see what starts to take off on streaming services? Really short tracks where if a bunch of 35 second tracks and you use AI to promote it maybe authentically, but you're artificially cutting the track short to maximize your royalty bearing streams. A lot of these tactics existed before AI. Yeah, that's what it's going to say. I feel like it does feel like just exacerbating all the issues you've already been working on I imagine. That's right. What we're seeing is a different level of scale and sophistication. So our recent announcement was that we're rolling up this new music spam filter. So we're using machine learning and our data models to identify all these common forms of spam and completely remove them from our recommendations across Spotify. You know, these aren't kind of edge case scenarios where it might be an artist just being a little over eager and how they're promoting themselves. We're looking for content farm behaviors. People trying to manipulate streaming services, not necessarily through artificial streaming, but through these supply side tactics like mass uploads or excessive duplicates or artificially short tracks. Like what is a mass upload? Is it like hundreds of tracks in a day or is it thousands like what kind of scale do you see in these scams? It can be over the last 12 months alone. We've removed 75 million spammy tracks from Spotify. Okay. A pretty big number. Yeah. And so, you know, that's the level of severity. We're trying to take with the issue, you know, there can be smaller scale where our spam filter will catch it, remove it from recommendations, try to remove the incentive to engage in that spam all the way through, you know, removing the tracks from Spotify. And, you know, I don't think many people are streaming these tracks. They definitely aren't. But, you know, we want to take on cleaning up the platform, making sure that a greedy just spam tracks aren't there, kind of crowding out the experience in any way. Yeah. So, when you say 75 million spammy tracks, I actually saw quite a few different outlets post about this incorrectly where they thought it was AI tracks. And I was like, actually, I think they specifically use the word spammy because it's not just AI. That's right. Okay. So, can you explain what that means? Yeah. We've definitely seen the width, kind of the advent of modern Gen AI tools. We're seeing the scale increase. And so, certainly AI played a role in kind of the scale of, you know, not so long ago, there weren't even 75 million tracks on streaming services. Now, we're removing that many. But yeah, we're working to identify spam regardless of whether AI is part of the creative process or not. Certainly, if you think of like, agentic AI where you could start to ask AI to make you songs in a traditional DAW, but you're using spammy tactics to promote it. If you have music, you make yourself without the help of AI and you're engaging in these tactics, we still take that seriously. So the reason we're stamping up our investments is because AI is increasing scale and sophistication of this kind of spam. But fundamentally, you know, just because you aren't using AI to make the music doesn't mean we're okay with this spam. Those spammy tracks, I'm sure, were a mix of, you know, fully AI generated all the way through to no AI incorporated at all. And then as part of this announcement, you were also talking about how there's like a partnership with DDEX, which is kind of like the standard bearer for metadata. I don't know. There might be someone who's listening who's like, what is metadata and what is DDEX. Can you explain like what this all means for Spotify? Yeah, DDEX is a consortium of streaming services, retailers, labels, distributors, and we all agree on industry standards for metadata. So how everything from like how should a feature and a track be disclosed and a track through to all the different roles and credits that you can put in the song credits and everything in between. So yeah, we talk about deception and the age of AI and what AI transparency means. You know, we thought a lot about it. And when the most important principles is holding hands as an industry and doing this together. When it comes to, you know, listeners be able to see the role of AI and the songs they're listening to, it doesn't feel great for long-term listener trust in music. If you go to one streaming service and see its AI, another service that kind of has a more nuanced take of some AI, some not. Maybe another service doesn't mention anything at all. Maybe some labels deliver that information. Maybe some distributors don't. That kind of inconsistency feels like what's going to harm listener trust in music long-term. And Spotify, along with other streaming services and labels and other industry partners, are all going to be self-reporting when AI is used in any part of the track. That's right. We have about 20 distributors who have already signed on and public-emitted to work with us on this. We're going to be displaying the credits on Spotify as soon as we start receiving it. And yeah, you can think of it as creating like a shared language for how we'll disclose as an industry the role of AI. So you could get pretty granular with it. It's a pretty flexible standard. You can say, "I wrote the lyrics. It's my vocals. My friend played the drums. Here's their aim. But I generated the guitar on AI, where I just did the mixing and mastering using AI or any other combination therewithin. We think it really starts to recognize the way AI is starting to play a role in music-beaking, where it's on the spectrum of artists deciding which elements they might decide to use AI for. And I think a few years ago, there was a sense that maybe there'd be this binary of categorically AI music versus not at all AI music. And we're already seeing tons of examples of things across that spectrum where traditional dolls are adding AI plugins and AI features. Some of the AI music generation software companies are adding human creation features where you can upload your lyrics or export to a doll to finish it or just generate one stem and export that one. You think this standard, if we get critical mass of labels, distributors, streaming services to adopt it, and set up a world where a listener would go to listen on any streaming service and see the same information and understand the role of AI in the same way, which feels really important. I think we have the time as an industry to do this the right way. Yeah. It definitely feels like a step in the right direction to try to get everyone on board and to create a standard because, like you were saying before, I hear this all the time. refer to things as AI music that have like just a touch of AI in it. And I'm like, okay, there's a
a very big difference between generating a song at the click of a button and a person writing the entire song, but then it's like one guitar line that they used from Suno or something. I do think also if I'm a bad actor who wants to game Spotify, why would I self disclose? I probably wouldn't. - We see this as a first step. No matter what the long term solution's gonna be of like the system of incentives and deterrence that would get people to disclose, the starting point has to be shared language through the existing supply chain of music about how we'll deliver the information and what the formatting of that will be. I think we're already seeing labels and distributors start to differentiate from each other and kind of how they think about how to support AI music or not. So I could picture once the standard set up, labels and distributors made set up different rules, streaming services will set different incentives for kind of rewarding AI transparency. We certainly would have some planned, but I think you do see already a lot of artists, songwriters, producers starting to talk about how they're using AI. - It is getting more normal. - Yeah, and so you see the K-pop demon hunters, songwriter who talked about brainstorming with chat GBT when he wrote soda pop through to like Brenda Lee, - Oh yeah. - Used AI to translate rockin' around the Christmas tree into Spanish, but still her voice. - Yeah, not crazy. - So cool. And it'd be kind of confusing for Spanish listeners if they thought, you know, Brenda Lee or any artist, a spokeal language they don't speak. But it's really cool for them to know transparently, oh, this is actually their voice. You know, AI made it into language. I can understand the lyrics in, but certainly the artist I've seen use that technology have been transparent about it. Boy, the Beatles used AI to kind of touch up John Lennon's vocals in here and now. - Yeah. - Great, but I thought 50 cents the other day shared. It was some like AI riffs on unreleased tracks. - Yeah. - Different genres. So you see these artists transparently sharing it on social media. I don't see any reason why they wouldn't be happy to share that on streaming services when they deliver the music. - Yeah. - You're definitely right. There's other systems that would be needed for people trying to evade or, you know, bad actors trying to deceive. But fundamentally, I think a great starting point is, you know, how do we create an ecosystem where artists, songwriters, producers feel comfortable disclosing so that listeners can preserve that trust. We think this is probably the right starting point. - I've thought long and hard about what the right way to treat AI music on a streaming service is, just like generally. And I find it to be really challenging because like there are some common suggestions and I think they all fall into certain pitfalls. So like for example, I think if you were to say, I'm a streaming service and I've decided I'm totally demonetizing anything that's AI generated. Well, let's say fully AI generated. Just to keep it like black and white. - Sure, the prompt generated. - Prob generated music. If you were to demonetize those tracks, then maybe that streaming service would have an incentive to pack their playlist with stuff they don't have to pay for. Or like, you know, they're all these like weird side effects that could happen. - Yeah, or the uploader would add a few human elements to it. And then it's not 100% AI. And, you know, they could have to kill. Just move against that ceiling. - Yeah, exactly. Like I'm sure as you've seen in some of your work, like people will always find ways to try to get around like whatever the rule is. When y'all came out with these policies, it does feel like a start, but I heard from some people that they felt it didn't go far enough. - Yeah. - So what do you say to those who feel like it's not going far enough? - It's early days for AI tech. I know it feels like it's moving fast, but consumption of prompt generated music insanely low. We have some time for artists, songwriters, producers to take the lead and figuring out how they want to use these tools. We don't want to act like we know where AI music's headed and exactly every policy and rule we need now. That's gonna be future-proof for the next two or three years as things evolve. On the other hand, we didn't just want to wait and do nothing. There's some areas we all can agree now that we need to act now, no matter where AI tech heads is going to be necessary to have great systems in place to stamp out spam, deception, impersonation. And so that's our starting point. We try to be upfront. We see these as first critical early steps, more to come. You're gonna see more and more updates from us in this space and we'll try to do it in lockstep with how the technology is evolving, but importantly also how we hear and see artists, songwriters and producers using it. We want to follow their lead a little bit. I think I've seen a lot of you reporting on the topic. It's evolving kind of in real time. I think you start to see the dial turn of more artists and producers being willing to talk about how they use it. They're still being a lot of, it's a hard topic to talk about. In every industry talking about AI, what it means for the future of our work and artistry in our careers is a hard one. And so I know I personally respect hearing all the different opinions about it and we're trying to be decisive and authoritative in the areas where it seems clear what's gonna need to happen and then really kind of follow the lead of the creative community as we see where it starts to evolve. - I mean, it's all evolving so fast. And right now what I'm hearing a lot from songwriters and producers in my orbit is a lot of them are willing to tell me off record when we're at dinner that they're using AI tools or that they've experimented with it or they're using it on their demos. But they oftentimes still don't want to talk about it publicly or they're not quite there yet. They're not sure if they're gonna use it in finished products for songs or not yet. So I feel like although the tides are turning behind the scenes in the music industry and you're starting to see more people being open to it, we still have a ways to go before it's totally normalized and we totally know where it's heading. So I do know where you're coming from on that. - Dizor has been, Dizor is a French streaming service if anyone's unfamiliar. They're much smaller service than Spotify but they've been very proactive about kind of taking charge on what they want to do with AI music. I think of all the streaming services I've been reporting on. They've been by far the most vocal. Recently they reported that 28% of songs that are delivered to Dizor daily are fully AI generated. I'm wondering at Spotify have you guys seen those same figures 'cause that feels like a shockingly high number of songs? - Yeah, it's like, you know, AI detection tech isn't really foolproof yet. So it's hard to get the exact estimate from there. Every streaming service has pretty much identical catalog. We have no reason to disbelieve. It's a similar amount on any streaming service. That said, I think they also shared is that 0.5% of streams. - Yes. - That's those songs we're getting and we've tried a few different tactics to test out different detection tech, test out different proxies, to understand how much prompt generated music may be on Spotify and we find way lower than 0.5% and kind of stream share in total consumption. And so I know sometimes it can feel scary when you see those like upload percentages where it's like, oh my God, is 28% or, you know, any percent of music that's being listened to, is the prompt generated and that would feel really scary and unsettling or seeing is yeah, there's a lot of uploads. We're doing a lot of work to police that kind of spam where there are mass uploads that can add up to those kinds of percentages, but keeping it close to eye on the part that actually matters which is our listeners listening to it is generating royalties. - Mm-hmm. - And, you know, I think all streaming services are showing the same data right now, which is an infinitesimal amount of streaming of this music. - Yeah. - And so, I feel like we have time to do things like an industry standard and music industry can be slow. I know you know that all too well. We have some time before, you know, maybe consumption will never increase. Maybe it'll look, you know, AI music will have a totally different shape, hopefully it'll be more professional artists led as, you know, some of the use cases you're talking about where artists start to use it in really powerful, interesting new ways. - Mm-hmm. - And so yeah, consumption remains really, really, really low. And so that gives us some confidence. We have some time to figure this out. - Yeah, that's interesting. I mean, consumption being really low makes me think that like, it's kind of a burden, maybe on streaming services to hold all of this music, especially when no one's listening to it. So would you guys ever remove tracks that are just getting absolutely no traction? Like I'm talking zero streams. - I don't think so. - Okay. - You know, whether they're AI or otherwise, people upload their music to streaming services for all different reasons. I have family members that upload music to send to family and friends. That's a great thing. At Spotify, we are focused on emerging and professional artists trying to make a career in music, building a business in music. Our policies are in service of professional artists and emerging artists on their way to that. And so we take on the burden of how many songs are uploaded and certainly the overwhelming majority of songs aren't getting streamed much. I still think it's really important for there to be this open outlet where, you know, you pay a low annual fee to a distributor and you can upload music and send it to anyone in the world that would have been unthinkable 15, 20 years ago. And it's unlocked a lot of really cool expression. So we take on the burden of what to do with all these uploads but we definitely want there to be this path for anyone with music to be able to share it with the world. - Got it. Is there like any cloud storage issue? Like I don't know how big of a deal, like how big songs are to hold onto. But. - Maybe someday with AI scale. It'll be at some scale. I can't even picture. - Yeah, five years from now, we'll have you back on the pod. - Right, yeah. And now we'll dig do a flashback to me saying it was all okay and it'd be like, oh, we were wrong. Yeah. Right now it looks like, we have some great engineers who work on all the smart ways. We like store the files and things I couldn't even talk about if I wanted to. But yeah, that isn't a goal of ours. We're happy to host music for anyone who kind of wants to get it out to the world. - When I thought about like AI music starting to make its way on streaming services, One of the things that my mind went to immediately was that some streaming services and
including Spotify, I have had to deal with other waves of audio content, I guess I should say, that is being used by bad actors typically. For example, I'm thinking about how Spotify has developed policies around noise tracks. Someone doesn't really know about this noise track thing. There are some very savvy people who will upload white noise to streaming services or bird sounds, or something that doesn't have it, it's not music, it's definitely not music. They'll hopefully get some traction on that and earn some streaming royalties without really having to put my artistic effort into it. But Spotify has a policy against that, which kind of, it doesn't totally demonetize it, but it kind of gives it a lesser royalty, right? That's right. Did you think about some of these other policies that you've created, like the ones around noise when you're looking at AI music? Definitely. It's a good example where our policies focus on the output of the recording. We never know the creative process behind a recording. We get the finished output and we're focused on that output. So when it comes to our noise policies, we don't know if you're brushing your microphone. I want to attempt to do it, but you could do it analog and ASMR style. You could use digital tools to make a noise recording. You could use AI tools to generate bird jerps or whatever maybe. One thing we found is 10 years ago Spotify was paying out about a billion dollars a year to the music industry, which was something, but in the scale of music industry, it wasn't enough. People wondered, "Would we be able to scale?" We're paying out 10 billion a year now. So when you think of that 10x and 10 years, things that used to be 0.1, 0.51% of royalties wouldn't be such a financial opportunity for bad actors to try to kind of head in that direction. As we've 10x the royalty pool and we're paying out over 10 billion a year, types of consumption that are 0.1, 0.51% of consumption are pretty material, potential earnings. So noise is an example where we started to see an over supply. It's so easy to make noise recording. A lot of listeners, if they're trying to study, focus, sleep, meditate, want to play. Noise recordings on streaming services. It's a great service to provide for people who want to listen to it. But what we didn't want to see was this flood of undifferentiated noise recordings, trying to kind of compete in SEO hack their way into a portion of this 1% of the royalty pool. So we introduced a few rules. One is noise recordings under two minutes are in eligible for royalties at all. You'd see people have a playlist of 35 second noise tracks because you play it when you're going to bed, and then it kind of maximizes royalty bearing streams. So we demonetized all noise recordings under two minutes, had to kind of reduce the financial incentive to keep flooding us with noise recordings. So we now value noise recordings out of fifth of the value of music. Try to write size. What the financial opportunity is to brush your microphone and stuff in different ways and release it with different SEO terms in it. I'm always astounded by how smart people are at doing such bad things because my brain would never. Wish the world didn't work that way. I imagine so much of your job is just trying to anticipate what someone will do if you make a rule about this. How will they get around it potentially? The fun parts are like tools for professional artists to use in awesome new ways. And then unfortunately that comes along with some stuff on the margins that aren't always as fun. But you know, they're intellectually interesting at times. How do we try to stay ahead of some of these tactics? Yeah, back in April, I did a story with Selena on the glow. I guess her her title is like global head of editorial had a playlist basically. So I talked to Selena and I asked her about whether or not she would, you know, ever forbid AI tracks from living on Spotify editorial official playlists. And she didn't really have a clear answer at that moment. It wasn't a yes, but it also wasn't a no. So I'm wondering, could you ever imagine fully AI generated tracks living on a Spotify playlist? Yeah, it's a hard question because I, you know, I think as we recognize AI music is this spectrum. And certainly like the plug-ins that people have been using in dollars for years. Or you can't differentiate from what we now call AI tech. And so I'm sure there's lots of culturally relevant great music that incorporates AI tools. I think what you're getting at is like completely prompt generated music without any human and but. Is there some world where listener behavior really changes and there's huge musical cultural relevance from, you know, music that doesn't spam to see or impersonate but somehow finds an audience, you know, could be like a viral hits, you know, sort of playlist where there's some song that's doing that well. Maybe I can't speak for their team, but fundamentally, you know, 100% of the focus of our editorial efforts are helping to identify uplift, you know, develop the careers of professional artists who are making amazing music. And so yeah, it's always hard to answer questions and absolutes, but certainly that's not the focus of any editor or Spotify or I think any streaming service. Earlier this year, I think a lot of people realized that there is a lot of AI music that lives on all streaming services when this band called the Velvet Sundown, it's an AI-generated band, started popping up. And at first as a reporter when I saw it, I was like, this is not that compelling to me as somebody who follows AI music so closely because I've seen plenty of AI songs, AI bands, whatever that have ended up on streaming services. So it didn't feel particularly unique, but I think that it really brought this idea of AI music living on streaming services alongside human artists into the zeitgeist. I think people really realized it for the first time. I'm wondering based on your new policies are bands like the Velvet Sundown, which it's all AI-generated music, AI-generated images, does that violate any of your policies or no? In general, we're just looking at signs of spam. I haven't looked at their account recently, but I don't think they were flooding platforms. I don't look at it recently either. I don't think they were flooding the platform with spam or I'd have to look in more close about what tactics they were using. They're spamming. I don't think they're impersonating other artists. The one that's interesting is when you think about deception, AI transparency, that's one where I think at first they had an artist bio on streaming services that had a fake quote from a magazine. What's Bill Bourd? From Bill Bourd. Yeah, so many people hit me up and they were like, did you write about them? Right. I was like, no. There would be people probably thought it was you. I know. I searched all over our website trying to figure out if we had written about this AI-generated band that no one's ever heard of called the Velvet Sundown. We did not for the record for the record. Yeah. At some point they changed their bio cross-spartifying everywhere else to make it. I think they're famous like a synthetic project or something of the sort. This is part of why we think about the DDEX AI credits because for a project like theirs where they're happy to disclose their synthetic project, there isn't really a way to share on streaming services that you're using AI. And I think probably a lot of the conversation would have looked really different back then. I think a lot of what makes people curious about novelties like that are, you know, is there a conspiracy? I need to research online if they're AI or if they're not. What's the nature of the project? And I'd love to build a world where there's an easy way to identify the role of AI in music. You can decide whether to listen to it or not. You know, for that project specifically, I know a ton of the listing was driven by Reddit and social media and traditional media covering it. The on-spotified the monthly listener account is about a fifth of what it was at that peak when everyone was writing about it. But I get it was kind of an interesting thought experiment for people to talk about. Yeah, I don't think it ever even would have taken off in the commentary at all if we had tools and easy ways to have them disclose clearly where they're using AI and where they aren't. I think they've started being pretty open about it on their social channels. That's kind of how we would think about it. Yeah, it was just one of those interesting moments. I think we've had a few of them where it's not necessarily that it's the first time anything has happened in that realm, but as AI is developed. It's just like whatever that moment is when everyone catches on. I feel like that band just for some reason based on this Reddit post, which kind of made it go viral, then made it into this big media thing. I definitely don't want to undersell the importance of talking about the role of AI music, but there's definitely been a few of these moments. I think of like 2023, this artist ghost writer released a Drake the weekend. Deep H song that took over TikTok for a weekend and there's tons of conversation about it's a watershed moment. There's many tons of songs like this and we need to act now to make sure that doesn't happen. And since then, I don't think we've really seen any songs like that take off. We take vocal impersonations seriously. We wrote just wrote us new policies about it, but it's not like there's been any kind of viral tracks similar to that one since then. Same here where velvet sundown seemed to capture the imagination we found it really useful to read lots of opinions and thoughts on it was a good way for us to get a sense of pulse check of how people are thinking about it, but it was a bit of a novelty moment huge spike of streams and everyone's talking about it. It's cratered since then. I think it kind of became a little bit more of a lightning rod to talk about rather than something that's culturally relevant or musically relevant. As I mentioned, consumption of prompt gender music overall at any estimation we have is incredibly incredibly low. So it's an interesting anomaly to talk about because it's kind of a thought experiment, but it hasn't signaled some huge trend of anything worse thing on spotify or I think any other services are staying either. Yeah, yeah, for sure.
So let's get to some of the new features. So there's a lot of stuff happening at Spotify. Like I said, I do feel like I'm hearing about something new every week. But first of all, why is there so much happening right now? Is this just like happenstance or are we building towards something? Always building towards something. A few years ago, we sat a really clear vision for where we want to head with music, which is this kind of shift from just being a place for the songs, where you may listen to the music and go elsewhere for the context, and the world, the artist world around it, to a place where, you know, Spotify can be like the place where your music world lives. And you can see everything an artist is doing, you know, merch and tickets, videos, get lyrics, get behind the song information. And so I think with that level of clarity, we've seen the team really start shipping and delivering and it's been kind of awesome to see. Mm-hmm. Yeah, it feels like almost an answer to this, like TikTok age that we're still in, where a lot of music discovery started shifting over towards social media, where people were discovering songs as they were scrolling on TikTok in the background of the videos that they were seeing. And then they would eventually go to Spotify as the final destination. But basically what you're saying is like Spotify wants to be like the destination for all of it, period. Well, what we're hearing from artists was you'd have these, you know, far-old moments on video apps, and then maybe it could make people fans of that song, and play a list of that song on streaming services, but it didn't make them a fan of the artist, right? As soon as you hear that song in a meme, you swipe up, you're seeing a cooking video, and a celebrity video, and artists are competing with the entire world of video content. So we really heard that need from artists. I can like, you know, break viral songs on, you know, TikTok or video apps, but I can't build a career. And so the role we see Spotify playing is this kind of music-centric place, where you can rabbit hole with music and artistry. And if you're an artist, you know, instead of having to upload, you know, short video every six hours to keep up with an algorithm, you can make three really intentional clips and attach them to your single. And anytime in the next, you know, day, week, month, when you're listening to a song, you scroll down, you see the clips that they made, explains the context of the song, why they wrote it, it's attached to your catalog. You go from there to see that they're touring near you with our new concerts, near you playlist, you buy their vinyl, a bunch of markets outside the US, you can go watch the music video, watch their live performance, and you're not competing with this whole world of content outside of music. We think we can be this kind of really kind of special experience for our fans and listeners, where they just rabbit hole on the music, and therefore artists have permission to share with more quality, with more intention to be about the music, instead of having to be a personality. And so that artist feedbacks really what we've been trying to build towards. And so one of the big announcements that you guys have had recently is that you're having a new lossless audio features. At last. Yes, at last. And that is where my question is going. In 2021, early 2021, Spotify announced that it would be coming out with lossless audio or high fidelity listening, and then it never came. And a lot of the competitors started rolling out high five, but it never happened for Spotify until 2025. Why now? Lossless has been like the top comment on a lot of Spotify posts for the last few years. As with lots of things in music, licensing, so always a big part to figure out. We were trying to figure out the right way to deliver lossless with a business model that would make sense for artists, for rights holders, for Spotify. Competitive dynamics definitely did change things, and we had to pause and make sure we were doing in the right way. Feel really good about the lossless offering now. It's been rolling out. I think we might even be done with the rollout by now. I got it recently. Within a week or so. So yeah. Yeah, exactly. It's lossless meaning no change from how the quality is delivered to Spotify, to how listeners hear it. So as the artist intended, a really cool way to deepen that artist's fan connection, like we're talking about, make sure you get a high production quality experience that you can get when you're listening to your favorite artist. Yeah. So back in 2021, like, hi-fi was such a hot topic in the music industry. I feel like everyone was talking about how this could be like a way to unlock greater royalty streams for artists. Like maybe you could charge more for like a more premium streaming experience. And at first, like, title was, I think the first one that had it, and they were charging more for it. But then Apple and Amazon came in and decided to offer it for free as part of their traditional premium offering. Did the fact that high-fi went from something where you could finally up your prices and charge more and move into an extra premium tier? And then it became something that actually was just offered for free. Did that weigh in to Spotify's decision to hold? Yeah. As you know, like roughly two thirds of all the revenue we make from music goes back to artists and songwriters, rights holders. So when we look at things like lossless or any other experience, if we can find ways to charge more for it, that means more revenue for artists, for songwriters, for creators. I think lossless probably could have been a good opportunity to do that in the industry, norm, way of the different direction. But I think in general, you know, work cited about ways where our incentives are aligned with with artists and songwriters and with the industry. If we can find more premium experiences that listeners love and fans love, that's a great thing for artists and songwriters. It kind of helps power that next stage of industry growth. So yeah, I think, you know, maybe there's a different timeline where that would have played out a different way. But, you know, I'm really happy that hundreds of millions of premium scholars around the world kind of get to experience it now. Another new service from Spotify is the DM feature. But this is actually not a totally new thing. I remember DMing-- I remember it too. Yeah, in box. No one remembers this, but me. It was early, early days. I had a friend that I used to DM on Spotify when I was like in high school. And I loved doing that because he had such good music taste. And then I don't know. I guess I just like woke up one day and realized it was gone. But every time I mentioned to people like, you know, Spotify used to have DMs. They don't remember it. But I'm excited that DMs are back. I think that's fun. To me, it signals a little bit more of like a social direction about interconnecting not just you to the artist and what they want to share in their page, but also how we could connect with each other on Spotify if we wanted to or how friends can connect. Is that the intention behind it to start moving more in a social direction? Yeah, but social specific to music. Or podcasts and audiobooks too. The way we think about it is trying to build all the best ways for fans and listeners to connect with music, more deeply listen to more music. So there's a bunch of social listening music, fandoms, quite social. And so we try to build some cool ways for fans to connect with each other. So blend is a really popular feature where with up to 10 friends, you share a link, ever and joins. And we make a personalized mix that updates every day. That's a perfect blend of everyone's taste. So like I recently had a road trip upstate. There's five of us. We made a blend and then as the songs are playing, we'd be like, oh, that one's actually for me. And you know, my friend's song is like, you know, they are taste profile that played it and asked them about the artist. And then I'd open, you know, open up the artist profile and follow them myself. Some great example of how that social experience can build. We have jam, which is another kind of live listening experience. Like instead of passing the ox cord, you can build a collaborative playing cue together with friends. It's like college parties or, you know, any sort of event like that. And inbox is another iteration of this. You could send music or podcasts or audiobooks to your friends, discuss them. You know, I don't think it's a sort of place where you're going to be like texting all your details every day. I think it could be a really cool place to curate a playlist, send an artist to someone, you know, your best friend or your crush or someone who you think is going to love it. To be able to talk about the song after you listen to it. So we think it's a really cool kind of complement to things like jam and blend where you can connect with your friends over the music. Earlier this year, there were reports that Spotify was working towards, like a super premium tier or something of that like, basically something that costs more than your traditional paid service on Spotify. And through it, you would get new features. But a lot of these features that are rolling out now are things that I would have expected to be in that super premium tier. So is that still in the works? Yeah, no specific details on that. But, you know, our focus has always been, how do you make Spotify premium, like a no-brainer to subscribe to? We just rolled out a bunch of improvements to our free tier as well, because we know that premium business models, what's helped us grow from a billion dollars in royalties paid out to the industry per year or 10 years ago to 10 billion now. 60% of premium subscribers start unfree. So we're trying to improve that free experience. Then once you get to premium, we want it to be an experience you never want to leave. So, you know, lossless, quality sound, playlist, mixing features that we recently launched, the uninterrupted listing experience in general. And so that's always been our approach with premium. We've tried to be thoughtful about making sure you're getting as much value as possible from that, because it's good for us and for the industry and for listeners, if we can kind of keep people on that premium experience. Got it. So, kind of maybe TBD? Yeah, I mean, certainly, you know, we look at concert tickets that we now list on the app, where, you know, we send you out to the ticket provider to go buy tickets from your favorite artists, our merch that we let artists sell. There's lots of kind of different ways of merchandising things. You know, we don't take a kind of merch sales or anything like that, but certainly in the fan economy, there's probably lots of interesting ways that we can power artists' businesses and power ours, but fundamentally, we want premium to be kind of a no-brainer, always adding value to it, and always feel like you're, you know, you're getting your money.
my needs worth. Got it. And I know that you all did make some improvements to the free tier. And I wonder, do you fear when improving the free tier that you might empower people to just stay on it? Because it gets too good that they just don't want to pay for it? You know, when I always say to that, like the reason you can believe us about, you know, trying to convert people to premium is because incentives are aligned here. Yeah, we want a free tier that brings in the most free users for the top of the funnel. We're really, really good at moving people from free to premium. As I said, 60% of premium subscribers start on the free tier. So yeah, by kind of modernizing the free tier. So I think a lot of younger listeners, especially expect to be able to pick what song they listen to. They can do it on other services. And so we still have a shuffle only experience. Yeah. So what we've seen in the data already is because we have this, you know, better free experience where you can play the songs you want to. We're getting way more users in there. We're still really, really good at upgrading them to premium. Like we talked about adding more and more premium only features that you've kind of feel like you must have. And once you have them, you don't want to give them up incentives are aligned here. We're trying to get people to premium two thirds of that premium subscription goes back to artists, songwriters and rights holders. We, you know, generate, you know, around a billion a year and music advertising on the free tier. So there's good monetization happening there. But monetization definitely has even stronger once you upgrade. So, you know, we're all on the same team. The industry is on a path, the a billion paid subscribers at this point, which is pretty wild across all streaming services. We're just kind of wild. If you think about where we were 10 years ago at the low point of the reported music industry. And so, yeah, we're definitely still on that path. You know, we think subscriptions, the right business model and premiums the right way to maximize it. To do a lot of the stuff that you all are doing. I imagine these new licensing agreements that you guys have been announcing periodically at Universal, Sony, Warner, BMG, Cobalt. I'm trying to think who else in Merlin. Yeah, a lot of music companies have renewed or up to their licensing deals with you guys. There's not too much information that I have seen on exactly what that entails. And I understand that those are private agreements. But one of the things that a lot of them note is that it improves remuneration for songwriters on the platform. That's been a big issue in the music industry. It's kind of an inside baseball thing. If you're listening to this, you're not very familiar. In like March, 2024, Spotify moved premium to being a bundle, which basically just means the same price tag. They're dividing the royalties between audiobooks and music, which meant less money for music. But this improves a lot of these deals are improving the royalty rates for songwriters from that bundling moment. Right. Can you tell it? Can you give us some more information on what that all means? Yeah. So now if you're a premium listener on our kind of core premium tier, you also got 15 hours a month of audiobook listing on Spotify. Really cool. The reason that we do these bundles that we think about music and podcasts and audiobooks together is all these things, you know, massively increase the number of people we can ultimately get on to Spotify, get to pay for a premium, and you know, continue into music. And so a lot of the data we've been able to share with our partners shows how we bring in a different type of listener for audiobooks. Maybe you sign up because you heard how great that model is for audiobooks instead of buying credits or you have to buy a whole book. You can split it up and, you know, whatever you want to listen to for 15 hours that month. Once you sign up for premium, two thirds of that revenues going to music rights solders and once you're listing audiobooks on Spotify, we're pretty good at getting you to start to listen to music too. Yeah, we follow the rules about how payouts work for bundles, the same way, you know, all our competitors do bundles as well. Apple one, YouTube premium, Amazon Prime. We all follow the same rules. We're really excited to do these direct deals with publishers because all these new experiences that we're talking about, we, you know, fundamentally think licensing's the core of the music industry. You want, you know, permission and proper compensation for all the experiences that we're building on Spotify. And so we're really happy to have more direct deals with a lot of publishers so that we can work through those dynamics of them directly, make sure they feel great about how payouts are working, make sure that we're getting all the rights we need to build fan experiences that grows the pie. And yeah, you know, we've paid over four and a half billion dollars over the past two years to publishing rights holders, double digit percentage increase from, you know, 2024 to 2025. So publishing payouts are going up, they're going up faster than other services. And yeah, definitely kind of understand where a lot of the public concern came from and we're happy to kind of have that chapter behind us and really think about how we're building new products together. Mm-hmm. The MLC is still trying to pursue a lawsuit against Spotify because of the bundle situation. I'm just wondering like what's the else take on that situation and where it stands today? Thankfully for me, I don't have to deal with the lawsuits and legal issues, but yeah, you know, I know today courts have found we were acting in kind of pretty clear accordance with the same rules everyone's following about how bundles work and what that means for payouts. And I think everyone agrees and industry bundles are a good thing for growing the pie overall. But yeah, I don't know much about, I don't know if I lawsuits. Yeah. And with these direct licensing agreements that you guys have forged with a lot of the publishers, which has improved the rate for the publishers and songwriters who are part of these agreements, does it get it back to pre-bundle royalty rates or is it still a little lower? I don't think we think about it that way. And licensing terms stay private with our partners for clear reasons. But we're always thinking about licensing as a value exchange. Everyone feels really good about where we landed in these last deals where we're able to build some of these really exciting experiences that are going to grow the pie and a lot of partners feel really good about spotify as the biggest paycheck in the music industry. And so, you know, we'll continue to work with them that way. Moving on to like spotify more broadly, I feel like there are so many changes going on as we've been talking about. What are the biggest ones being the Daniel Eck, the founder of spotify, is stepped down to CEO, but he's still going to be chairman. So, what does that mean for the future of the company now that it's under new leadership? Yeah, really excited for Daniel for this next step for him for us internally. It's not a huge change. Daniel appointed Gustav and Alex or our new co CEOs. They became co presidents two years ago, I think. And so, I think Daniel stunned this really thoughtfully where, you know, that was probably a first step where he was still CEO, but, you know, he elevated the two of them to co presidents. And now is it evolves fully to him as chairman and Gustav and Alex as co CEOs. You know, you've mentioned our pace of execution of all the stuff we've been launching and building. You know, I think that's thanks to the setup we have. And I'm sure this will kind of continue to accelerate. So, shouldn't feel like too much of a change for us. I know the titles probably feel like a bigger impact externally. But for us, I think Daniel's done a really good job of thoughtfully, you know, setting up Alex and Gustav for success. And they've both been a Spotify for an incredibly, incredibly long time. So, so they're not new basically. I think they both been there longer than 15 years. Wait, whoa. Yeah. Spotify hasn't even been in the US for more than 15 years. Right. That's right. Yeah. So they've been there since almost the beginning. Wow. That's cool. Yeah. Really excited for them. Yeah. And I mean, Spotify is a Swedish-based company. I'm just curious like, what are some cultural differences that you've noticed from going back and forth to Sweden. I'm sure you've been a few times. Yeah. Yeah. You know, I've been as Spotify seven years. And my first year was pre-COVID. I got to go to Sweden all the time. And it was like, you know, a great benefit of the job. And then ever since even as we're, you know, returned, you know, back to offices and so on, you know, now we're just used to doing the virtual meetings. So I don't get to Sweden as much as I wanted anymore. I think what's fun, just like Spotify, culturally, is we've got a huge Sweden office, a huge London office, huge Newark office, Miami office, Nashville office, LA office. I focus on my team on and all those offices. So you kind of just have this like, you know, global environment where people bring different perspectives and different taste in music and different, you know, cultural touchstones. Totally. It's kind of baked into our culture because we're Swedish with such a big, kind of US office. And do you have like any recommendations of something someone should do in Stockholm if they ever find themselves there? I definitely say this is an unnoble recommendation, but the summer months are when to do it. You get the like 20-hour days. Wait, 20-hour days? Well, I haven't been to Sweden though. We're have to fact that one we're enough to fact check. Okay, okay. I'll back check if there's like a line that appears below us right now. A lot of my, that's not right. A lot of my core Stockholm memories like middle of summer, you know, like grove locks or you know, whatever, I tend fish and then like a pint of beer and it's like 11 p.m. and the sun's still out. Wow. I think that's the vibe to say go in summer. We especially in 2019, the joke used to be that everyone had a excuse for why we needed an offsite and Stockholm for some, every summer, somehow there was like an urgent reason we all needed to go to Stockholm, but in winter, there never was. Yeah, just in July. Yeah, we're literally there's a business critical reason we had to go in July, but never in the winter. Wow. I love that. Yeah, I'll have to go to Stockholm. I've been to Oslo a few times. Great city. I know if there's someone in Norway who's listening to this for some reason, I'm sorry that I just compared Sweden and Norway. I know that they have differences, but I do enjoy Scandinavia. It's a great place to end off this conversation. We've talked about so many
things, but now let's get to music. At the end of every single conversation that I have on the record, I asked people to make me a playlist and we're going to add it to this kind of compulsive. Oh, it's actually on Spotify. I love it. Plug for Spotify. You can do that to me now. Perfect. Yeah. Perfect. Yeah, so whenever I do make me a playlist, I have people pick three songs for me based on three prompts. So the first prompt is, what is a song that you can no longer gatekeep? I've been really liking Romimars lately. A Lester, Ego, amazing. I'm anxiously waiting the first album so people can get on board now before the first proper project. That's such a great pick. I wouldn't have pegged you as a Romimars fan. I'm surprising. What is a favorite throwback? Probably after I might go to karaoke song, Natalie and Bruglia Torn. I have a terrible singing voice. It's a problem. People hate doing karaoke with me, but Torn, I sing it with passion, but even though it's like she sings it amazingly, you can kind of like chant sing it. For people with my vocal challenges, it's a great selection and a great song. I once tried with a group of people that I didn't know on New Year's Eve. I tried to do before he cheats my carry-on to wood at karaoke. One of the hardest. Why would I do that? You have to properly like belt that. I know. I'm not a bad singer, but I could not do that. My vocal cords are heard even thinking about trying to sing it. It's a great a great song though. Yeah, but maybe I'm going to take up Torn. That seems like a little bit of an easier one. It is pro-tap. Okay. Last one is what is a guilty pleasure of yours? Guilty pleasure. I'm going to say Kelly Yokey from the Kelly Clarkson Show. Good line. All of her covers. You know, we all loved her from American Aid All Originally. So something just like that brings you back to childhood. It's a subject. Have her sing covers and happier than ever is probably her best one. Yes. All our covers are amazing and they put the wall on Spotify also. I didn't mean to plug Spotify with that. I'm sure they're on all streaming services. But you know, if like it's a hard day or you just need something just like truly joy in doing the Kelly Yokey playlist is really solid. She's unbelievable. That's a great pick. Well Sam, thank you so much for being here. It means a lot of her came. Of course. Really fun. Another big thank you again to Sam Duboff from Spotify for sharing his insights with us this week. Here's the hot 100 top 10 for the week of October 25th. And number 10 is actually Romantic by Taylor Swift. Number nine is the life of a showgirl. Number eight is Wood. Number seven is Wishlist. Number six is Father figure. Number five is Elizabeth Taylor. Ordinary by Alex Warren reenters the top 10 this week at number four. Number three is another reentry golden by Huntrix. Number two is Opalite by Taylor Swift. And number one for its second week in the top spot is The Fate of Ophelia by Taylor Swift. Thanks so much for tuning in to this week's episode of On the Record. And another special thank you again to this week's guest Sam Duboff. If you enjoyed today's show, please give us a reading, a follow, a thumbs up, any of those things. They all really help a new show like ours to reach new audiences. Again, I'm your host Kristen Robinson and tune in next week for even more industry insider knowledge. I'll see you then. [Music]
Podcast Summary
Key Points:
Spotify updated its AI policies to combat spam, artist impersonation, and require self-disclosure of AI use in music.
The company is collaborating with the music industry to develop generative AI tools based on four principles: upfront agreements, artist agency and opt-in consent, proper monetization and compensation, and deepening artist-fan connections.
Spotify removed 75 million spammy tracks in the past year, using machine learning to identify and remove content farm behaviors like mass uploads and artificially short tracks.
New features include direct messaging, Mix of Spotify for DJ-level transitions, music video expansion, and high-fidelity listening, signaling a shift from static streaming to interactive artist-fan engagement.
Spotify is partnering with DDEX and distributors to standardize metadata and ensure transparent disclosure of AI involvement in tracks.
Summary:
In this podcast episode, host Kristen Robinson interviews Sam Duboff, Spotify’s Global Head of Marketing and Policy for the music business, about the company’s recent changes. Duboff discusses Spotify’s strengthened AI policies, which target spam, impersonation, and deception, and include a self-disclosure system for AI use. He explains that AI tools have increased the scale and sophistication of spam, leading Spotify to remove 75 million spammy tracks and implement a music spam filter.
Spotify is also collaborating with industry partners to develop responsible AI music products based on four principles: upfront agreements, artist opt-in consent, proper monetization, and deepening artist-fan connections. Duboff emphasizes that these tools are designed to complement, not replace, human artistry by enabling fans to connect more deeply with their favorite artists. Additionally, Spotify is rolling out new features like direct messaging, Mix of Spotify, music videos in more countries, and high-fidelity listening, evolving from a static streaming platform into an interactive space for artist engagement.
The company is also working with DDEX and distributors to standardize metadata and ensure transparent AI disclosure, aiming to maintain listener trust. Overall, Spotify is navigating the challenges and opportunities of AI in music while prioritizing artist control and fan connection.
FAQs
Daniel Ek stepped down as CEO, triggering a leadership shake-up, and Spotify updated its AI policies to combat spam, impersonation, and require self-disclosure of AI use.
Spotify strengthened rules against spam and impersonation, created a system for self-disclosure of AI use, and partnered with DDEX for consistent metadata standards on AI involvement.
The principles are: upfront agreements with the music industry, artist agency and opt-in control, proper monetization and compensation, and building tools that deepen artist-fan connections.
Spotify aims to create AI tools that allow fans to go deeper with their favorite artists, such as AI-powered remixing, while ensuring artists have control over how their music is used.
Spotify removed 75 million spammy tracks in the past year, uses machine learning to filter spam from recommendations, and withholds royalties for artificial streams.
New features include direct messaging, Mix of Spotify for DJ-level transitions, music videos in more countries, and high fidelity listening.
Chat with AI
Loading...
Pro features
Go deeper with this episode
Unlock creator-grade tools that turn any transcript into show notes and subtitle files.