When volume becomes the business model - AI Slop, Market Share and the Double Royalty Problem.
On 15 September, Universal Music Group filed suit against DistroKid in the US District Court for the District of Delaware, accusing it of unlawful practices and copyright infringement. Universal Music Group names 1,000 recordings and seeks statutory damages of up to USD 150,000 per work, while calling those tracks the tip of the iceberg. DistroKid claims to distribute roughly 40% of all new music in the world, for more than four million artists.
Universal Music Group says the case is not about the distribution of AI generated music when clearly disclosed as such, but about a distributor masked as something it is not and benefiting from that false impression. That specific framing deserves to be highlighted, as it locates the damage where it actually happens, as in the existence of AI music, but as undisclosed AI music, which is not copyright protected and hence not entitled to royalty at all. Undisclosed AI music today enters and dilutes the same royalty pool as everything else.
Let’s start with what distributors are for. An artist or independent label without a direct deal with the DSPs has no realistic way to market except through an aggregator. Distributors deliver catalogue, manage metadata, report usage and collect royalties at a price a small catalogue can digest. This is genuinely valuable infrastructure, and it is why the DIY distribution sector exists at all. But as it stands today in the deals with the DSPs, the economics reward volume, and volume is indifferent to origin. A distributor who is paid per upload, or valued on catalogue growth, has no structural incentive to ask whether a release was made by an artist or by a prompt.
The consequence for legitimate rightsholders is a double robbery. The first taking is visible on every royalty statement. Streaming revenue is distributed pro rata from a finite pool. Every stream captured by undisclosed AI content reduces the market share of human repertoire, and with it the percentage of the pot that flows to the people who actually hold copyrighted rights.
The second taking is quieter. Unmatched and unclaimed royalties are commonly distributed by market share if not resolved. A distributor whose volume has been inflated by mass uploads therefore collects a larger slice of that market share and the residuals that, by definition, belong to someone else. The same conduct that shrinks legitimate rightsholders' share of the identified money enlarges the infringer's share of the unidentified money.
This raises a broader question about the business model that is emerging around distribution itself. We increasingly see distributors claiming catalogues of 500,000, one million or even several million artists. At what point does scale stop being the consequence of a successful distribution business and become the business model itself? If market share determines not only a distributor’s share of identified royalties, but potentially its allocation of unmatched and residual money, there is an obvious economic incentive to accumulate repertoire at enormous scale. The lower the cost and friction of onboarding that repertoire, the stronger that incentive becomes. Generative AI supercharges that equation: millions of tracks can now be created and uploaded at a speed and cost that simply did not exist when these distribution and residual allocation models were designed.
That leaves an uncomfortable question for the industry. Has distribution, for some actors, become less about servicing artists and more about manufacturing market share? If catalogue volume can translate into a larger claim on residual money, then mass distribution is no longer merely an infrastructure business. Scale itself becomes monetisable. And where the repertoire creating that scale is undisclosed AI content that may not attract copyright protection in the first place, we should be asking whether that content should be capable of increasing a distributor’s entitlement to residual royalties at all.
European law has begun to address the disclosure half of this problem, but the architecture deserves a careful reading. Article 50 of the AI Act, applicable since 2 August 2026, distributes its obligations by role. The provider of the AI generative system, such as Suno, must ensure that synthetic audio is marked in a machine readable format and detectable as artificially generated. The deployer, the person actually using such an AI generative system (like Suno) must disclose the artificial nature of the content, but only where that content constitutes a deep fake, meaning it resembles existing persons, works, entities or events and would falsely appear authentic. A fake track passed off as a known artist falls squarely within that duty. A million generic tracks imitating nobody in particular arguably fall outside it. And the distributor and the platform, the actors through whom the AI content reaches the royalty pool, appear nowhere in Article 50 at all. Nothing in the provision obliges them to check for the marking, preserve it, or act on it. The point of generation and creation is regulated. The point of upload is only regulated sometimes. The point of dilution is not regulated at all.
US law offers no straight transparency rule, which explains the shape of UMG's complaint. Where there is no disclosure regime to enforce, rightsholders reach for the tools that exist, in this case false designation under the Lanham Act, state deceptive trade practices law and the Copyright Act. UMG's suit against Believe and TuneCore, alleging infringement at industrial scale, settled in April 2026 with all claims dismissed with prejudice. Litigation is doing the work that regulation has not yet done, one distributor at a time.
That is not a sustainable enforcement model for a market in which one company claims to handle 40% of all new releases. If undisclosed AI content not protected by copyright and not entitled to any royalties continues to flow into a pro rata pool, the pool cannot correct itself, every actor who plays by the rules loses out to every actor who does not.
The industry knows how to solve provenance problems. It has done so before, with ISRC, with content identification, with KYC standards now proposed for distributors. What it has not yet decided is who bears the obligation when the generator, the uploader and the distributor each point at each other.