Three labels stopped suing Stability AI, and bought a stake instead

The day music copyright law got three new believers#

Stability AI closed a $76 million Series B on 25 August 2026, and for the first time in the company's history the cheque came partly from the three organisations that spend the most money on music copyright law: Universal Music Group, Sony Music Group and Warner Music Group. Electronic Arts, AMD Ventures and Pacific Alliance Ventures joined them, alongside returning backers Coatue, Greycroft, Kadmos Capital, Sean Parker and Eric Schmidt, taking the company's total raised under chief executive Prem Akkaraju, who took the role in June 2024, to $232 million, according to TechCrunch's report of the round.

None of the three majors arrived at the cap table cold. Music Business Worldwide's reporting on the round traces Universal's strategic alliance with Stability back to October 2025, struck alongside EA, with Warner following its own "responsible, artist-friendly" partnership a month later. Sony had no prior deal on record and came in as a first-time backer. Three separate bilateral partnerships, signed months apart, had just been folded into one shared equity position, and so far as the public record shows, no earlier funding round in generative AI has drawn primary cheques from all three major label groups at once.

Read the sequence as a proportional timeline rather than four flat rows and the pattern sharpens: a founding, a leadership reset, two quiet bilateral pacts eleven months apart, then one loud consolidation. Stability AI's own funding record has become the clearest public timeline of how music copyright law reached a working truce with a generative model, built deal by deal rather than in one grand settlement.

That last date matters more than the dollar figure attached to it. A funding round is a financial event; three major labels choosing the same financial event, in the same month a German court was busy punishing a rival for the opposite behaviour, is a statement about where the industry now believes music copyright law is heading. That same premise anchors folkfox's music industry marketing work: the legal posture is the marketing posture now, whether a brand wants it to be or not.

Why the labels swapped some courtrooms for a cap table#

Litigation and investment look like opposite strategies. They are not. Three days after the Stability round closed, Sony Music Publishing and Warner Chappell filed a joint suit against Anthropic, alleging what TechCrunch's coverage of the filing calls a "brazen campaign" of illegal torrenting and scraping used to train Claude on copyrighted lyrics and sheet music. Sony and Warner did not choose between suing and investing in the same fortnight. They ran both plays at once, against two different companies, and the difference between the two companies is the entire argument: one licensed the material it trains on, the other allegedly took it.

Germany supplied the sharper warning a month earlier. The Munich Regional Court ruled on 31 July 2026 that Suno's training on songs represented by the collecting society GEMA, including "Atemlos durch die Nacht", "Daddy Cool", "Rasputin" and "Mambo No. 5", went beyond the country's text-and-data-mining exception, because the works turned up memorised and reproducible inside the model rather than merely analysed, per JUVE Patent's report on the ruling.

The court ordered Suno to stop training on the works, disclose infringement revenue and pay up to €250,000 for every future breach, a second GEMA win after its earlier case against OpenAI. Suno has since struck its own licensing pact with Warner. Sue the scraper, then sign the version of it that agrees to pay.

The commercial dividing line under this version of music copyright law is not size, reputation or ambition, it is whether the training data was ever licensed at all. Stability's own announcement describes Stable Audio 3.0 as "a family of open-weight music models trained on fully licensed data", and RouteNote's analysis of the deal notes the company positions itself against Suno and Udio by using licensed catalogue from AudioSparx with creator opt-outs respected. That single design choice, licensed in rather than scraped up, is the whole reason one AI company is a co-defendant and the other is a shareholder register.

Not everyone reading the round took the vision statement at face value. Industry commentators on X spent the following weekend picking apart the timing rather than the tools.

The read is unkind but not unfair. The same three names that spent two years measuring the damage generative audio could do to their catalogue spent one August week measuring the upside of owning a piece of the tool instead, a pivot our own reporting on the majors' two-track week in AI music tracked as it happened.

What artists get paid, and what they do not#

Move up from the deal to the industry and the numbers turn less flattering. CISAC's economic study, produced with PMP Strategy from case studies and interviews across collecting societies, creators and platforms, projects that generative AI will put 24% of music creators' revenue at risk by 2028, a cumulative €10 billion gap driven by both unlicensed use and straightforward substitution. In the same window, the study expects the market for AI-generated music and audiovisual content to grow from roughly €3 billion to €64 billion. The pool is getting bigger. The people who wrote the original songs are not guaranteed a bigger share of it, licensed training data or not.

Nobody has published the rate card that would settle the argument. IFPI's Global Music Report 2026 says AI licensing deals are meant to "generate revenue opportunities for artists" without naming a figure, and researcher Dorien Herremans, writing for WIPO Magazine, argues compensation should scale with how heavily a given artist's catalogue actually shaped a generated track rather than paying every contributor an identical flat rate.

A separate technical proposal called Generative Content ID, described in a peer-reviewed paper on computational copyright for music generative AI, tries to trace a generated song back to the training data that most influenced it, precisely so a royalty model has something concrete to measure. None of that machinery is running inside the Stability deal yet. The equity is real. The royalty formula is still a research paper.

This is where ai music royalties and ai music licensing deals stop being interchangeable phrases. A licensing deal is a permission slip between a label and a platform. A royalty is money that reaches the writer, performer or session player whose work sat inside the training set, and folkfox's own tracking of Universal's Hook fan-remix deal found the same gap: six licensing pacts in ten months, and not one of them disclosed a rate the artists behind the top 10% of songs, who take 96% of the resulting output, could actually check against their own statement. Equity rounds do not close that gap. They just prove the label side of the ledger is now worth writing about.

For an ai music generator built on licensed catalogue, the honest marketing claim is narrower than "we pay artists": it is "we pay the rights holder we licensed from, on terms we have not published". Music copyright law does not yet require the second half of that sentence to be public. Trust, as distinct from law, increasingly does.

The compliance layer marketing cannot skip#

Every regulator circling this deal is asking a version of the same question a marketing team should ask before publishing a single AI-related sentence: can you actually prove the claim, and does music copyright law in the relevant jurisdiction back it up? Three answers already exist, and a fourth is still being written.

Notice what none of the four bullets contain: a rule that says equity investment satisfies a training-data obligation. It does not, anywhere. A label owning shares in an AI company is a financial fact. Whether that company's music copyright law compliance is sound is a separate, checkable fact, and a marketing team that blurs the two is writing a claim its own general counsel would not sign. Part of why folkfox's content marketing services exist is to keep that line from getting blurred in the first draft.

How music brands should talk about AI now#

The risk in this story was never purely legal. A label can be entirely within music copyright law and still lose an artist's trust with one over-confident press line. The fix is not silence, it is precision, and precision is a marketing discipline before it is a legal one.

None of this requires a legal team to write the copy. It requires a marketing team that treats music copyright law as a fact-checking exercise rather than a mood, and checks each claim against the deal that actually exists before it goes out under the brand's name. That is a smaller ask than it sounds, and it is the difference between a press release that ages well and one a lawsuit quotes back at the company a year later.

The labels chasing this story through the courts for two years have just proven they can also chase it through a term sheet. This is exactly the moment folkfox's brand strategy work is built for, when the legal posture and the public voice have to say the same true thing, in the same week, without either one overselling the other.

Frequently asked questions#