MBW Reacts is a series of analytical commentaries from Music Business Worldwide written in response to major recent entertainment events or news stories. MBW Reacts is supported by JKBX, a technology platform that offers consumers access to music royalties as an asset class.
Robots creating music, streamed by other robots, as criminals bank the profits.
A grimy, looping, un-Human Centipede, spewing out money towards drug gangs, people-trafficking operations, and God knows what else at the dog-end of civilization.
Is that the kind of music business you signed up for? Because it’s the one you’ve got.
The news yesterday (May 3) that tracks created on “anyone can do it!” AI music app Boomy were being deleted by Spotify for suspected streaming fraud should have surprised precisely no one.
This was the natural result of three coalescing mainstays of the modern music business:
The rise of AI music tracks that can be created at the touch of a button;
The popularity of ‘stream farm’ services, using AI bots to rinse plays of content on Spotify et al; and
The downright dumbness of the dominant ‘pro rata’ streaming royalty model, under which a share of the money you pay to play music every month is, inevitably, being sucked away by fraudsters.
Forgive me for not applauding Spotify too loudly for catching out/clamping down on the handful of ne’er-do-wells who created these tracks on Boomy (then used stream farms to rack up royalties).
This whack-a-mole approach is clearly fallible.
According to a recent French study, up to 3% of music streams on services like Spotify are known to be fraudulent. Yet this number only represents the ‘fake streams’ that services can actually detect; it doesn’t include the ones they don’t uncover.
Stream fraud monitoring firms like Beatdapp put the actual figure at “at least 10%” of overall streams.
The 7% differential there (between the 3% known figure, and the 10% cited by Beatdapp) would have been worth $1.2 billion in stolen royalties in 2022 alone. (Music streaming generated USD $17.5 billion in wholesale global industry revenues in 2022, according to IFPI.; 7% of that figure is $1.23bn.)
Beatdapp’s cited 10%+ figure is lent credibility by Spotify’s own admission that it can’t guarantee it will catch streaming fraudsters on its platform.
“We use a combination of algorithms and manual review by employees to detect artificial streams and aim to remove such artificial streams or non-bona fide user accounts created for [fraudlent] purposes and filter them out from our metrics on an ongoing basis,” reads the latest annual SEC report from Daniel Ek‘s company, issued in February.
“However, we may not be successful in detecting, removing, and addressing all artificial streams and any related user accounts.”
Nice disclaimer. It may have to do a heck of a lot of work in the months ahead.
Tick, tick, tick
The Boomy story is the thin end of the wedge. If things don’t change, and fast, Moore’s Law is going to explode the number of AI music streaming tracks, being played by AI music streaming bots.
This week, a sprinkling of Boomy’s AI-generated tracks were removed from Spotify. But exactly how many of the 14.4 million recordings users of the app have produced to date have already been used for illegitimate streaming purposes? Does anybody really know?
As for some of the biggest beneficiaries of such activity, let’s not beat around the bush: We are talking about gangsters.
In January this year, Nick Dunn, CEO of indie distributor Horus Music, told MBW that when his team previously blocked music uploads from “criminal gangs” – who were attempting to push tracks to DSPs in order to play them artificially – they were issued with death threats.
It’s obvious when you think about it: the royalty framework of bot-based streaming fraud is ready-made for digital money laundering.
So, question 1: How much of the hundreds of millions of dollars (perhaps $1bn+) lost to music streaming fraud each year – and therefore stolen from legitimate artists under the ‘pro rata’ system – has already enriched criminal organizations?
Question 2: In a world of post-Ukraine-invasion, ‘Know Your Customer’ financial scrutiny, how much liability should music DSPs bear for aiding and abetting said enrichment?
A Boomy business
Boomy, I loudly point out, has done nothing illegal itself. (A Boomy spokesperson said: “Boomy is categorically against any type of manipulation or artificial streaming.”) Its only misdemeanor in the Spotify story this week: providing suspected streaming fraudsters with a tool to easily create the music tracks they required.
MBW understands that Downtown/DashGo provided the pipes for Boomy’s users to distribute their music to DSPs.
Should Downtown, therefore, shoulder some blame for the past week’s events? Possibly. Especially if the offending tracks carried obvious hallmarks of streaming fraud fodder (being 31 seconds long, having barely any melodic variety, repeating a single audio loop etc.).
To Downtown/DashGo’s credit, it now appears to have put a pause on all uploads from Boomy, presumably while it investigates methods to lessen the likelihood of it getting involved in the chain of streaming fraud in the future.
Such methods may, ironically, come to lean heavily on AI innovation: Denis Ladegaillerie, CEO of Believe, said the other day that Believe’s DIY distribution service, TuneCore, will be blocking the upload of 100% AI-made music to DSPs via its service in future, and that the tech required to automatically weed out non-human-made tracks already exists.
Yet as I’ve argued before, none of these measures – neither Spotify playing whack-a-mole, or Believe playing stop-a-mole – cuts streaming fraud off at the source of its oxygen: the ‘pro rata’ streaming model.
Slicing the pie
For those catching up: The ‘pro-rata’ streaming model sees the royalty-bearing portion of the subscription fee you pay Spotify/Apple Music etc. every month pooled with the same money from every other subscriber. This central pool of money is then paid out to artists and labels based on their market share of total plays across the service.
The problem: This system financially incentivizes every party to rack up as many streams as possible – whether legitimate or illegitimate – as opposed to attracting as many listeners as possible.
This results in the crazy outcome of a portion of the $9.99 per month you pay for Spotify/Apple Music etc. going to acts you haven’t streamed… including those acts who’ve only been played by stream farms.
One swift solution to kill these incentives (and instantly choke streaming fraud) is the so-called ‘user-centric’ model, rebranded as the “fan-powered” model by SoundCloud.
Under this model, the royalty-bearing portion of your streaming subscription fee is paid out only to the artists you individually listen to.
Suddenly, rinsing endless plays of 31-second tracks via one subscription account becomes far less lucrative – because you’re not absorbing money from a large central ‘pot’. You’re just absorbing it from the $9.99 per month you pay for your own account.
Some of Universal’s apparent reticence towards ‘user-centric’ is based on the fact that, for real-life human artists, it’s a zero-sum game.
Although studies show that a switch to ‘user-centric’ on Spotify tomorrow would improve the royalty earnings of over 50% of artists, that would also mean lessening the earnings of every other act.
Instead, Universal appears to be moving towards emulating the anti-fraud basis of ‘user-centric’ – disincentivizing the blunt commercial reward for simply getting loads of streams – while reducing the likelihood that artists may ultimately earn less from a transition away from ‘pro rata’.
One tactic in UMG’s “artist-centric” model, for example, might be to reduce the royalty amount paid for non-interactive streams (i.e. ‘lean-back’ algorithmically-served music), while increasing the royalty rate for streams that have been searched out and actively played by users.
This would also reduce the earnings power of so-called ‘fake artist’ production music, which can often be found filling some of Spotify’s most popular mood playlists.
In addition, Universal may call for AI-made music that falls below minimal-quality identifiers (see: 31-second, no melodic variety, repeating single audio loop) to be blocked from being uploaded to digital services.
So far, other than the knowledge that Universal is trialing various ideas with two streaming partners (Deezer and TIDAL), not a lot else is known about what “artist-centric” may become.
On Universal’s Q1 earnings call last month, Sir Lucian Grainge said: “We’re confident that the artist-centric model will work because it’s in everyone’s interests, including the platforms.
“The real issue is this flood of content, this oversupply [of music uploads] will only increase with generative AI… It’s in everyone’s interest to make sure that the platforms don’t become awash with problematic content that fans and artists don’t want.”
“We’re confident that the artist-centric model will work because it’s in everyone’s interests, including the platforms.”
Sir Lucian Grainge, Universal Music Group
Later on that Q1 call, Michael Nash, UMG’s EVP and Chief Digital Officer, indicated that “artist-centric” would also look to increase revenue flow from “superfans” of artists and/or genres, who are willing to pay extra to access additional content on DSPs.
“Our consumer research says that among [music streaming] subscribers, about 30% are superfans of one or more of our artists,” said Nash, adding: “How does that relate to artist-centric? When you start to focus on the artist-fan relationship, these high-value relationships are driving the economic model of the platform, so you [can begin] segmenting around high-intent, high-integrity, artist-fan relationships.
“You’re starting to move the platform towards a deeper level of engagement with fan bases; the opportunity to monetize, we think, flows very naturally from that.”
For an industry beginning to worry about a potential future slowdown in global music streaming revenue growth, this will all sound very appealing.
Yet at the opposite end of the industry from the sophisticated “superfans”, we see the machine-powered streaming fraud grub – that un-Human Centipede – rapidly mutating.
Soon enough, it threatens to grow fat off an uncontrollable prospect: millions of AI tracks, played millions of times by AI bots, every day.
It needs to be stomped on, and fast.Music Business Worldwide
