Music streaming debates 2025 roundup: wrap-up for the streaming services as we know them? – Part 1

As 2025 draws to a close, we can start looking back on a particularly turbulent year for music streaming. One that might come to be understood as a tilting point for streaming and the music sector at large, although it is still unclear how several coin flips will play out. An eventful year either way.

There was the copyright litigation settlement between one of the three major labels, Universal Music Group (“UMG”), and the second most used generative AI (“GenAI”) service for music, Udio. The settlement resulted, to some surprise, in their joint announcement of a new interactive AI streaming service. Then there was the controversy around AI weapon technology investments of streaming profits by Spotify CEO and co-founder Daniel Ek. Boycotts of the platform by indie artists ensued over the summer. In late September, Ek swapped the role of CEO for one that is more behind-the-scenes. Two weeks later, Spotify, too, announced partnership plans for developing “responsible AI products” together with UMG, the two other major labels Sony Music Group and Warner Music Group, Merlin, a music licensing partner for independent record labels, and French record company Believe. So, 2025 brought all that, as well as music streaming-related European national rulings on royalty rates, AI-memorised copyrighted song lyrics (GEMA vs. OpenAI), and streaming fraud, plus more litigation-to-partnership-settlements between major labels and GenAI platforms.

This is the first of a series of posts that tackle the copyright-relevant discussions currently taking place concerning the music streaming sector. This post will give a broad overview of (i) the challenges for fair remuneration and ownership, (ii) its implications for AI-powered streaming services, and (iii) transparency of streaming services. In subsequent posts, these three topics will be scrutinised more comprehensively, including in light of possible legal remedies.

Okay payer?

First, the seemingly everlasting monetary discussions. Those most directly revolve around to which people and entities the money flows, of course. But there are equally important repercussions for music diversity, access to culture and the creation of new musical works, as we will see in later parts of this series.   

Many parts of the music sector’s business side are marked by non-disclosure agreements. This is known to prevent extensive inquiry into remuneration and other legal arrangements in the music sector. The private settlement between UMG and Udio in late October is a prime example. There is a compensatory element to the settlement, but what percentage of the compensation that UMG receives will reach artists remains undisclosed. As a result of limited empirical insight, discussions often resort to framing, cherry-picked statistics and lopsided narratives (Osborne 2025; Nowak & Morgan 2021). In this post, the goal is to not take anything at face value, in order to arrive at a truthful and balanced overview of the streaming field. Still, there may be a conscious bias that, as value in music originates from its creators, its proceeds should sufficiently flow back to them. This bias is, however, firmly rooted in Article 18 of the DSM Directive, which states that Member States have to ensure that where authors and performers license or transfer their exclusive rights for the exploitation of their works, they are entitled to receive appropriate and proportionate remuneration.

Now, what is sufficient remuneration for artists? An evident first step is looking at the distribution of remuneration for neighbouring rights and copyright resulting from streaming between all beneficiaries, which totalled €18.6 billion in 2024, representing 69% of total recorded music revenues. Recently available data from 2020 to 2024 show that, with neighbouring rights and copyright taken together, around 30% of proceeds remains with streaming services, 42% goes to music labels, 5% to publishers, while authors and performers pocket 10% and 13%, respectively (see here and here). So, about one tenth for artists. Is that sufficient? Many say it is not, including 85% of the European Parliament’s members in 2024.

The European Parliament adopted a resolution aimed at fairer distribution of revenues from the EU music streaming market. One of the targeted streaming sector practices is payola schemes, referring to offers where authors can accept lower or no revenue in exchange for greater visibility. Spotify’s Discovery Mode is an infamous example, where they do just that: coverage of an artist’s song in a weekly playlist format, in exchange for lower royalty payout obligations for Spotify (Buccafusco & García 2021; Gambato & Sandrini 2025). Since the 1950s, in the context of radio stations, such pay-for-play has been made illegal, yet this has not been effectively applied to streaming.

Another aim is the revision of pre-digital royalty rates, bringing them into line with fair and modern rates. Arrangements between record labels and artists can stem from before streaming existed or before it was as ubiquitous as it is now. Back when labels had to manufacture and distribute physical vinyl, cassettes or CDs, 80/20 splits may have been more justifiable (Osborne 2025). Here, it is important to note that the majority of today’s music rights are still under (major) label control, even though the streaming model allows for more modes of independence. In Q3 of 2025, the recorded music market share of major labels was 80%. What is more, the European Commission has opened an investigation into UMG’s intended acquisition of Downtown Music Holdings, the largest independent music distributor, determining whether the acquisition is competition law proof.

The implications of such market power of a few entities over the whole music sector are manifold. For one, it enabled the three major labels to form the architecture of streaming in the early days. After heated license negotiations in the early days of Spotify in the late 00s, “the big three” leveraged advance payments, advantageous licensing deals, and, at some point, up to a joint 18% of the shares in Spotify (Pelly 2025).

Progression from A Minor to AI Major?

The major label market power now is used for leveraging the music GenAI boom that took place from 2023 onward. Start-ups developed the infrastructure and revenue models for GenAI, ignoring generally assumed payment responsibilities for the music their models were trained on, and risking lawsuits by rightsholders, which quickly followed. Before we dive into that, note that there is also much to be said about developments this year surrounding GenAI content policy on streaming services. Those developments will be addressed in a subsequent AI-themed post. For now, it seems most interesting to first focus on the recent announcement of the new major label GenAI streaming services.

When almost three quarters of non-major labels surveyed in 2023 thought it was time for a new alternative to the current streaming model, it seems unlikely this was the breakthrough they hoped for. GenAI has often been lauded as the new opportunity for democratisation of the creative sector. But how democratic can the outcomes be, when for each decision on firmer establishment of large-scale breakthroughs in the music sector, major labels hold the majority of the votes? Like with streaming services in the late 00s, now shaping the architecture of music GenAI services.

Since so little is known about these upcoming services, we can only speculate about what the infrastructures will provide. Even so, they already prompt questions, especially around the transparency of these platforms for both users and the artists whose music they rely on.

RecSys-axis: unintelligible or unheard-of?

Modern day streaming platforms are an axis about which recommender systems (“RecSys”) rotate. Or that is the case for the types of streaming platforms that are widely used, anyway. RecSys are fully or partially automated systems, that, in the context of music streaming, are used by streaming platforms in their interface to suggest or prioritise specific music and playlists to its users.

Much research has been conducted into the consequences of streaming on music listening (Aguiar et al. 2024). One (by now perhaps obvious) finding is that streaming has led to an increase of playlist listening at the expense of albums (Bonini & Magaudda 2024). Through platform curated playlists, as well as a feature such as autoplay, music listening experiences are more malleable for streaming platforms, enabling them to increase engagement, for example by targeting specific listener profiles through RecSys (Prey 2021; Aegidius 2022). Bonini and Magaudda note in their 2024 book that in today’s platform society, “[t]he transition of music consumption to commercial digital platforms has generated an unprecedented information asymmetry between music gatekeepers and listeners: it had never happened in the history of cultural industries that a bunch of multinational companies held such a vast amount of data, information, and knowledge about their customers’ tastes and behaviors […]. At the same time, it had never happened that cultural consumers had so little information and knew so few things about the processes of music selection and distribution of the contents they are proposed with.” That is true not only for consumers, but also for the artists whose music is disseminated on those platforms.

In the context of their newly announced GenAI services, UMG CEO Lucian Grainge shared in an internal memo that UMG “will NOT license any model that uses an artist’s voice or generates new songs which incorporate an artist’s existing songs without their consent”. Commentator Cooke clarifies that this clearly implies that artist and writer consent will not be sought for other uses of their music in the context of AI. Spotify promises proper compensation, and “upfront agreements” instead of “asking for forgiveness later”. It is easy to imagine such promises getting lost in the black boxes that are AI music generation models.