As the AI bubble swells, Spotify’s ‘artist-first’ AI music product announcement might just be code for maximising shareholder value

It turns out that Spotify has even bigger plans than its recent integration within ChatGPT: its own rollout of ‘'artist-first' AI music products’. The AI products will be developed in collaboration with Sony Music Group, Universal Music Group and Warner Music Group, indie music representative Merlin, and independent digital music company Believe, with additional rightsholders and distributors involved in the future. Spotify contends that industry backing will ensure that ‘artists and songwriters’ are prioritised through voluntary ‘upfront agreements’ that create ‘new revenue agreements for rightsholders, artists, and songwriters’ focused on fair compensation and attribution. There will be ‘thoughtful AI guardrails’ that catapult artists and fans into a new AI-powered world of music connection.

Photo by Anh Tuan To on Unsplash
This Kat is keeping an eye
on the big shareholder pie

The timing of this announcement is interesting, to say the least. Over the last few months, Spotify has announced new features and policies designed to place it as a key player in the trajectory of digital music distribution and much, much more. The story is reminiscent of its somewhat piracy-related origins. One that the Gustav Söderström, the Co-President and Chief Product Officer, nudges along in their latest press release. Spotify is scrubbing its image squeaky clean from allegations of genAI slop filling some playlists and inequitable compensation schemes, to remind us that they were the original piracy saviour solution. An industry, that they forged, today comprising more than 500 million paying listeners across all music streaming services.

Spotify is taking great pains to emphasize that they fundamentally believe in copyright, and that without them steering the ship, ‘AI-powered innovation will happen elsewhere, without rights, consent, or compensation’. In the coming months, Spotify explain that they will work closely with artists, songwriters and industry to shape these products.

We are incredibly fortunate, it seems, that Spotify is navigating these turbulent AI generated waters. But how fortunate? Or are we really facing a washout of creative autonomy?

This is about artists (but, really users)

In a recent interview, Spotify’s global head of marketing and policy, Sam Duboff, had a different yarn to spin. These headlines allegedly form part of a broader decision to transform Spotify into a platform for users to ‘rabbithole’ music. Building from previous sharing-based tools like blends and jams, or the newer direct messaging feature, these new AI products are more akin to ‘AI-powered’ remixes for users, than songs prompted into existence (e.g. Suno). One where ‘different mixing tools [...] enable users to “flex their DJ skills with different transition styles to create the ultimate mix”. Distinguishing itself from the likes of TikTok, Spotify plan to facilitate music-related content sharing that prioritises ‘quality and intention’.

No longer a site of static consumption, the relationship between artist and fan is, apparently, paramount. This seems to be part of a longer strategic vision than some may realize. While Spotify was recently granted a patent to generate ‘song mashups’, it builds upon a previous patent, filed in 2018, that relates to a method for training a neural network system to separate vocals from instruments in mixed audio tracks. Known as U-Net, the architecture was initially used in biomedical imaging but now forms the basis of a potential feature that generates personalized mashups. It also exists alongside an older patent relating to metadata from users’ voices and background noise that determines personal information to indicate an emotional state.

Spotify users may never have to actively discover new music nor leave the platform, as these tools potentially mean that your musical tastes (and mood?) will simply be re-affirmed and potentially resold to you through personalised advertising.

Pricing

So, how much is this all worth? Spotify’s narrative has always been to ‘grow the pie’ to increase royalty payments. Part of the delay on the lossless audio feature for premium users allegedly links to an industry-wide decision that is a freemium feature, one that Spotify had perhaps wanted to charge for.

The same rumour has existed in relation to user-focused ‘superfan tiers’ with ‘higher-priced’ subscriptions. Again, ‘growing the pie’ for artists, and offering new fandom-based tools that are designed to bring listeners to them that stay as fan. Duboff mentions that the provision of these tools needs to be done within an ecosystem that preserves artists’ trust. Yet, following the audiobook bundling fiasco, it is perhaps telling that Duboff cannot generally confirm whether these new initiatives may bring artist royalty payments back to pre-bundling levels.

Comment

This Kat is a little sceptical of these ‘wholly new revenue streams’, and is particularly concerned over the fate of artists that do not opt-in on the platform. It would seem that Spotify has no intention of demonetising generative AI content, unlike its policy on noise recordings. The reception to these products has been mixed, with the most cautiously optimistic reflecting that the ‘devil will be in the detail’. Others simply point to Spotify’s ‘pollut[ion] of creative ecosystems’ particularly the fact that generative AI songs have ‘diluted the already limited share of revenue that artists receive from streaming royalties’.

That being said, in-house generative AI tools introduce new dimensions to cultural production on platforms – something this Kat explores in her forthcoming book, Creative Reuse. Duboff is quick to rattle off recent artist forays into generative AI, from Brenda Lee’s ‘Rockin’ Around the Christmas Tree’ being 'translated' into Spanish using an AI voice clone of her vocals, to the songwriter for KPop Demon Hunters using generative AI to brainstorm, to 50cent actually using AI to remix his music. Part of Spotify’s wider policy with DDEX also reflects these trends that challenge a ‘binary’ perspective of AI music. One where a new industry standard for AI disclosures in music credits will prompt consistent transparency over its use from ‘generated vocals, instrumentation, or production’.

Spotify’s interference within the generative AI tool landscape is the first of its kind and it will likely shape the future of user-generated content and linked remuneration initiatives. Whether we are superfans or not, Spotify’s move begins to fill in some of the missing pieces left from the EU’s AI Act through nudging user norms, shaping terms and conditions, and outlining principles to guide licensing agreements for the purpose of training AI models.

There are of course bigger problems to overcome than creativity, that artists themselves call out. What of the sheer energy (and pollution) generative AI involves, particularly for marginalized communities? Duboff jokingly reflects that storage, if this scales, could test software engineers at Spotify. Or perhaps more striking and something difficult to forget; Daniel Ek, the founder and recent CEO of Spotify (now Chairman) investing €600 million in AI military company, Helsing. This wider context lays bare the investment-focused reasoning, perhaps amplified by the AI bubble, that sits at the heart of this new announcement. One that makes the far-out case of Spotify simply buying out Suno to offer user-prompt generated songs more likely.

Spotify becoming a responsible AI provider was definitely not on my bingo card for 2025. And while I would like to yell bingo at some point (because it is really fun), we should really keep our eyes focused on this big pie they are baking for their shareholders.