Music streaming fraud – how does it work, who's making millions, and why does it need so much AI music?

0:04Hello there, welcome to the price of music brought to you by Music Ally. If you want an easy guide to what's going on in the music industry and how it all works, well then here we are looking at the finances of the music industry, whilst also bringing you stories and comments from the pop biz in the past week. He's Stuart Dredge. I'm Joe Sparrow. And this week, we're not starting with a big number. We just have lots of big numbers throughout the show. And we're going to have a sort of one of those irregular deep dives that we do aren't we, Stu. And it's going to be into what? It's going to be into streaming fraud.

0:35Streaming fraud. Now, we've spoken...

0:37Stu:How to do it, how much money you can make. No, hang on, wait. We're not doing that. No, only joking, folks. We've spoken a lot recently on recent episodes about AI uploads and how this shocking fact that AI-generated music is now accounting for over 50 % of all music uploaded. And when we explain why that's the case, the answer is always, well, actually, it seems to be mainly because people are trying to commit fraud and AI music allows you to do that. So we wanted to figure out a bit more why. And that was based on a couple of emails we had from listeners saying, well, how does this work? And why, how does AI music equal fraud?

1:15So we wanted to sort of talk about that and discuss what streaming fraud is anyway, and how does it work? And we're going to have a good old route around in that topic. and we think you might be surprised at how it works, who's doing it, and how much money can be made. But first of all, Stu, when we say streaming fraud, what do we mean?

1:33Stu:What is it? Well, this is one of the first misunderstandings people have, I think, because I think what we all first assume is if someone is trying to do fake streams, they're an artist and they just want to get up the charts and they want to boost their own performance music, which, I mean, that is a thing that can happen. It's been mentioned in a few court cases that it might be a thing, people, labels employing bots to kind of boost things. But it's really not the reality of streaming fraud. Streaming fraud is about organized crime. So it's about criminals saying, well, we want to kind of get loads of money and hacking people's PayPal accounts by phishing is becoming harder.

2:13Stu:But hey, this streaming music thing seems to be a lot of money there. How can we do that? So it's very much organized crime, seeing stream music as a way to make money. But it doesn't mean it's not a victimless crime, because every dollar that fraudsters earn from music streaming is sucked away from a human artist who's making real music and putting it out there. And that is the kind of... That's why people are bothered about it, I think. It's not just that we're worried about Spotify losing some money to criminals. It's about money being sort of siphoned away from human artists to people who are not professional musicians and not even trying to make art.

2:55Yeah. Now, you and I have spoken to a lot of people in the sort of streaming fraud world. And the thing that clarified it for me was the idea that these are organized criminals and they're looking to launder money. So they're looking for ways to turn dirty money into clean money by sort of putting it, cascading it through several systems, and then clean money comes out at the end. And the way that they find music streaming to be kind of attractive is because, as we all know, we can all trot off this figure now, I think, which is if you get a million streams on a music streaming platform, you get about four to five thousand pounds worth of money as royalties for that.

3:29And it costs significantly less than four to five thousand pounds to buy a million streams. So if it costs you, I'm making these numbers up, but if it costs you even two thousand pounds to buy a million streams, you're going to make two to three thousand pounds of profit and your money is now clean. And that is the attractive feature of it. We've got some sort of quite stark numbers here to sort of put a frame around it. Well, I've got figures from 2024 here, but there was, imagine a big number basically. Here's your big number of the week. There was 4.8 trillion on-demand audio streams in 2024.

4:04A bigger number in 2025, I imagine. 5.1 trillion. 5.1, there we go. 0.3 up on that. Thank you very much. Now, there's a tech company called BeatDap, and they are an anti-streaming fraud company. So they work with labels and catalog holders and streaming platforms. and what they do is they're given all their data and they go through the data and they use their magic system to identify any streams that look like they are fraudulent. BeatDap estimates that between 7 % and 10 % of all of those music streams might be fraudulent streaming, which would equate to roughly 500 billion fake streams. Now, we can take it for granted that obviously a company whose job is to identify fake streams would say it's quite a big number because it's sort of in their interest to say that.

4:59But at the same time, Deezer has released information saying that when they looked into their fraud, they sort of found about the same figures, didn't they? Six to seven percent.

5:07Stu:Yeah. Although some other DSPs, some other streaming services argue the opposite. So I think Apple Music says, I think it's about 0.3 % of its streams are fake. And I think Spotify was... There was a French study a few years ago that showed that I think 1.14 % of Spotify streams are in relation. Now, to some extent, depends on how kind of conspirator minded you are. You might be like, well, that's the ones they know about. What about the ones they don't know about? Is it higher? So it's anywhere between 0.3 % of streams and 7 to 10 % of streams, depending on who you believe. But I think you can gauge the fact that this is an issue by the fact that the music industry is taking extremely seriously, the streaming service is taking it.

5:51Stu:No one is saying this is a tiny thing and we don't need to worry about it. Even the companies that say this is tiny are saying, we're working really hard to stop this becoming a less tiny thing. Apple, for example, when it said barely any of our students, it was in the middle of a talk about how much work it was doing to identify this stuff and kick it off. So they are all really treating this seriously and acknowledging that it is definitely a thing that people are trying to do. it's a big problem and even if we scale it down even if we say it's one percent of all streams that are fraudulent that is still 50 billion fake streams a year right which is which is a lot and when we say fake we're talking about manipulated in the sense that um these are not intentional streams from a human being these are ones that are specifically activated to uh ultimately generate money for the fraudsters so before we get on to the ai things let's sort of talk about how this how this actually functionally works um and the thing to understand as a key point if you're still struggling to picture it is that what they don't do is like upload one song and then point the bots at it and say right go and stream this 100 million times and we'll take all that money that would obviously be flagged and and wouldn't work but they use this quite subtle strategy which is they release a whole bunch of tracks oh they distribute them onto these dsps they claim the ownership of them.

7:15So when those tracks are played, they create royalties that go to them. Of course, no one's going to play them because they're not real artists, right? But then they use bots. And I guess artificial intelligence has made this really helpful in terms of getting the bots out there. So they use bots then to go and play these songs, bot accounts. And then across all of those songs, they play them enough times to not be detected, but to make money. that goes back to them. You can sort of see now how artificial intelligence is really helpful here because you can really quickly spin up lots and lots of tracks that evade detection systems so that they can be played in the first place.

7:59It's a scale play, isn't it?

8:01Stu:It's not one track getting a million streams. It's 100 ,000 tracks getting 10 streams each. And you literally, even as a committed fraudster, you can't make 100 ,000 tracks as a human every day. but you can use AI to start to build up those kind of scales. So it's not that AI music is fraudulent of itself, but it's the only thing that can help you have the sheer scale of a catalog to upload that makes it work, really. So that's the association between the two. An important thing to point out here is that what they have to do is make sure that the bots are streaming this music using real accounts with real subscriptions, right?

8:45So they can't just make fake user accounts because they won't get the money. So what they do is apparently there's like, there's many levels to this scheme. So you've got, they will go out and buy a bunch of login details on the dark web of, you know, normal accounts like yours and mine that they know have an active subscription and will generate money when they're streaming. And what happens is they figure out what time zone you're in, and then they will log into your account automatically when you're asleep, stream songs for six hours, and then stop again so you don't notice it as a user. Which is, I mean, I would say is incredibly smart.

9:20Stu:Well, some people have spotted it when they get to like Christmas time and Spotify raps and go, well, who is that? It's like a most streamed artist. Why did I stream them all night for six months? It's a weird thing, isn't it? because this is why it's so hard. I think one thing people might think is this should be easy to spot once you know there's a bot. And so for one thing, these bots are not just streaming the fake music, they might be streaming playlists that include real artists, which could mean a blowback for those artists if you would say, well, everything that bot streamed is obviously fake, we'll get rid of it.

9:54Stu:But also, there are some real human behaviors that mirror this. So if you think about k-pop stands where there's a real culture of the band's album came out at midnight let's all stream it all night long to get it to number one in the charts yeah that's not fake streaming that's fandom streaming but spotting the difference between that and bots is very very nuanced and difficult so well there's a really interesting point there isn't there which is like that fandom streaming which is intentional and air quotes real it is with the intention to game the system, isn't it? They want to push that track as high up the charts as possible.

10:30They're not listening to it necessarily for enjoyment over and over and over again. They're doing it to sort of for a third reason, which is not fraud, but it is a very fine line between those two types of behavior, isn't it? And then the other thing you mentioned there, which is really fascinating, is that the other clever thing that these fraudsters do is they have to hide the plays of the fake tracks in and amongst playing real tracks by real artists so that it sort of disguises that pattern of behavior, which means that even for real artists, potentially 10 % of their streams is being done by bots to disguise.

11:07Yeah. And it's not their fault. So it would be really hard to go to them and say now, oh, by the way, we figured out that like 7 % of your plays are by bots. So That would be like a really terrible lose-lose, wouldn't it?

11:20Stu:The other thing is there's a time pressure, isn't it, because obviously they pay out every month, I believe. So they've got to spot this, analyze it, be sure that it's productivity, and then remove the streams and not pay the royalties for them within a few weeks, which is no mean feat. That's really difficult. And also, one other interesting point about this is it partially explains this controversial thing that most streaming platforms have done, which is say, we won't pay any money to you as an artist if you get less than a thousand streams a year. Explain that to us, Stu. Well, this came out of a while when Spotify did it and said, right, okay, from now on, if you haven't earned a thousand streams in the last year, we're not going to pay you.

12:02Stu:And this was presented as, this is a way of putting money in the pockets of the human artists who are getting traction, who are getting more than a thousand streams. And a lot of the discussion about it, like you say, has been people worrying about, what about artists who are right at the start and are still yet to get their first thousand streams on a song? This is bad for them. And it penalizes independent levels back catalogs, and this is a valid debate. But yeah, part of this also is that that approach does mean that the AI tracks that are getting five plays are making money. It's very difficult.

12:37Stu:It's one of those things that is, in theory, slapping down the AI fraud to put money in artists' pockets, but some artists are saying, hang on a minute, it's also penalizing us if we're not at that level. But yeah, I think this is one of the key motivations behind that. You'd think, as a slight side note, that it would be fairly easy. You mentioned there the hypothetical indie label that has a huge back catalogue, and maybe some of their songs only get streamed 900 times a year, but you still want, if they've got hundreds and hundreds of those songs, they still want the cumulative payment for those streams, and they're not getting it maybe.

13:17And you'd think it'd be easy enough to flag a legitimate catalogue to be paid from stream number one, wouldn't you? But that's conversation for a different time. Stu, do you have any case studies to tell us about people who are doing this?

13:31Stu:It doesn't matter. I do. Okay, great. No, there have been two notable kind of criminal cases in recent years. One is a couple of years old. It was in Denmark. And there was a man who basically made, I think it was about 2 million Danish kroner, which was about 230 ,000 pounds at the time. Basically, he released several hundred music tracks, some of which were edited versions of other musicians' work that he's known. Right. And then had them artificially streamed. And this guy became Denmark's 46th highest earning composer for streaming between 2014 and 2017. It's a long time project, a long time ago.

14:16Stu:And it took a while to bring him to court, but eventually he was found guilty and sentenced to a year and six months in prison. Wow. But then the case at the moment is on a lot of people's minds. There's a guy called, it's a man called Michael Smith. He's from North Carolina, and he was indicted federally in September 2024. So it was a big announcement. And the accusation was he worked with an AI music company to generate hundreds of thousands of songs. So again, like 10 times scale of the guy in Denmark, and then got thousands of bot accounts to stream them on Spotify and other streaming services.

14:51Stu:And the reason I can talk about this man's name without worrying about being sued too much is that he actually pleaded guilty to this. So he pleaded out to one count of conspiracy to commit wire fraud and agreed to pay nearly $8.1 million in forfeiture. Wow. So I think this is the allegation. He'd earned more than$8 million from doing the scheme. He was due to be sentenced this week as we talk. There isn't a report this time of what that sentence was, but he could have been... He faced a maximum of up to five years in prison. Wow. But yeah, that's an example of a streaming fraud where you obtain more than$8 million in royalties, and that's just one guy doing it.

15:33It's funny, isn't it? Because it's a sense of scale. When we talk about streaming, there's so many hundreds of millions of people paying for streaming around the world now, and there's trillions of tracks. And we know that the amount of money generated by streaming is, I don't know,$25 billion or whatever it is. These are huge numbers, these are. And when you talk about$8 million, in that context, it could sound quite small. But actually, that's$8 million that didn't go to human artists and was sort of being skimmed right off the top. And this is the point, isn't it, that these people are skimming really like tiny fractions of a percentage off the top of the whole system.

16:11But when you aggregate it across everything, it becomes enormous. And this is where AI becomes really helpful because it's now this real enabler for this kind of fraud. and you said it allows scammers to produce hundreds of thousands of songs at very close to zero cost and very close to zero time and as we said at the start over 50 percent of all of the tracks uploaded now are ai generated that was from deezer but we can assume it's the same across the whole system what have deezer said about then how much of that is fraud and what happens after at that point.

16:48Stu:Well, that's interesting because I think the coverage of this has been interesting because the coverage is like, oh my god, half of new music is all AI now. It's taking over. But you have to say is anyone actually playing it? And Deezer says between 1 and 3 % of its total streams are AI music. And up to 85 % of those streams, the streams of AI tracks, are fraudulent. So that's, what's 85 %? Nearly nearly 9 in 10 streams of AI music on Deezer are fraudsters, which is something the music industry is really kind of getting between its teeth and going, right, this is why we need to figure this out.

17:29Stu:Because there are some people making AI music and trying to do a creative project with it and release it. And you may agree or disagree with that as a good thing, but they have artistic intent in mind, creativity in mind. But 85 % of the streams of this music at the moment are fraudulent. Yeah. So if the sort of maximum 3 % of Deezer streams are of AI music, only 0.4 % of all music on Deezer is AI music that is intentionally streamed. So AI music is really not wanted. That sounds like a rounding error more than intent, doesn't it? And again, I think I've said this a number of times to you on the podcast recently, I just don't understand why platforms are allowing AI music to be uploaded when you're faced with information like that.

18:26Stu:It's difficult though, isn't it? Because there's 15 % of streams that aren't. I think they're all saying that we don't know yet. I know we might disagree on it, which is fine. But I think what they're saying is there might be some legitimate stuff that people want to listen to. And we don't want to throw that baby out with the bathwater. And there's a lot of bathwater right now. And so at the moment, as far as I know, we're going to talk about what they're doing later on, I think. But at the moment, they're very much focusing on we're going to kick off stuff that we can see is being manipulated, but that doesn't mean we're going to kick off all AI music, even though that's...

19:00Stu:So it's a very sort of fine tightrope they're walking, but that's their view. And that's their problem. And then, of course, we've been reporting on this for a few years, and I've always been fascinated by this as a story because it's so interesting to see how organized crime and Interpol connect with music streaming. But one of the sort of interesting developments, and we did a whole show about this a month or so ago, is how then, if you could fold in sort of a third layer of complexity around fraud, which is prediction markets. Now, we spoke about these in a previous show, and these are the platforms that are technically not gambling, but essentially you gamble on an outcome of something that's going to happen in the future.

19:41So you can say, okay, I think the most streamed song on Spotify next month is going to be Bruno Mars, this song, and you stake some money on it, and if you're right, you win more money back. Not gambling, very close to gambling um and we've had a uh we sort of identified in that in that conversation that this is kind of sort of opening the door for a new form of fraud isn't it because if you if you go and put some money on a certain song and say i think this song is gonna get over half a million streams uh next month on spotify and then you go and buy half a million and one streams and it's cheaper than the amount you're going to win that's profit isn't it no it's true

20:21Stu:and that's it's almost like a system that's guaranteed guaranteed to encourage this sort of thing yeah so we've got like if you've got this sorry it's not a big bet is it yeah if you're placing a big trade not a bet definitely not a bet thank you Steve again like you were saying though if you could make 5 ,000 pounds from streaming fraud and it only cost you 2 ,000 pounds if you're making a trade that would earn you a million dollars yeah than spending$200 ,000 to make. I mean, the weird thing about the Couchy stuff, I think, is that it's not about 100 ,000 tracks getting five streams. That has to be about making one song get lots of streams, which arguably makes it easier to spot, would you say?

21:04Yeah, I think that's where there's... It's a fascinating third layer to this fraud cake, because it offers another route. So you can say, okay, you can go the traditional fraud route of scraping fractions and fractions and fractions across thousands of songs and letting that accrue over time. Or you can take the YOLO, go big or go home approach, which is I'm going to identify one song, I'm going to use this sort of bot system to bump up the plays, and I'm going to bet on that happening. And as soon as that bet is good, you know, so you say, okay, by the end of August, this song is going to have more than half a million plays.

21:42Well, that is validated in a very short period of time. And the platform says, okay, yeah, it got this many streams, boom, you pay out, you take your money, you're out of the system and gone. And this has sort of happened recently. So there was an innocent victim in all this, the independent artist Malcolm Todd. His song Earrings, I mean, this is an example of something being spotted. His song Earrings shot to the top of the Spotify US chart. And it was completely driven by people who had staked faked money via Kalshi to try and win their not a bet, but it's a bet. And Spotify, so they'd all put money on this song going to the top of the charts.

22:25They then did things, I don't know, maybe they streamed it legitimately, maybe they didn't. They then pushed it to the top of the charts to win their money, okay? And then Spotify spotted this and was forced to delete over 500 ,000 fake streams from the song. And there's zero suggestion here, by the way, that Malcolm Todd or his label had anything to do with this. But that's an example of it happening. And that's the one they've spotted. It could be happening in smaller ways, in more complicated ways, that are a bit more obfuscated. So this is a really interesting third layer to fraud because you know, financial markets are supposedly heavily regulated to stop fraud, but these are kind of operating in a gray area and it's kind of happening very quickly.

23:08And of course, there's an actual victim here in Malcolm Todd. He's lost 500 ,000 streams. He's lost$2 ,500, let's say. And his account might have been frozen or whatever during that time.

23:22Stu:Stu, this is actually what we don't know as well as it adds in that. One thing you could argue that the streamers have to do is, if someone is caught manipulating streams, penalize them. But in this case, he hasn't done anything wrong. If Spotify's algorithm would say, right, well, there were 500 ,000 fake strings that last song, so I'm not going to promote it. I know algorithms don't have this kind of voice. But if the algorithm was like, I'm not going to promote his next single, he was the fraudster. But he wasn't the fraudster. So there's all these unknown consequences if your music as a legitimate artist that has nothing to do with doing fraud gets kind of wrapped up in something someone else is doing, then it could hurt you not just now when you have your string segment, but in the future.