Sue the scraper, fund the licensed: the majors' two-track week in AI music

One week, two moves, one strategy#

On 25 August Universal, Sony and Warner all joined Stability AI's $76m Series B, the first time the three majors have taken equity in the same AI company in a single round, alongside Electronic Arts, AMD Ventures and Pacific Alliance Ventures. The same day, per Music Business Worldwide's court reporting, Universal and Sony filed an amended complaint against Suno in Massachusetts, adding a DMCA circumvention claim to the existing copyright case.

Read separately, the moves look contradictory: the ai music industry's biggest rights holders funding the very class of ai music companies they are suing. Read together, they are one policy with two enforcement arms. Licensed ai music companies get capital and catalogue access; unlicensed ones get process servers. The fox does not choose between the carrot patch and the snare line; it decides which field each belongs in.

Stability's chief executive Prem Akkaraju called the round "an affirmation of our vision where generative AI empowers every producer, musician, and storyteller", per TechCrunch, with the funds aimed at creative production tools, applied research and professional services. CelebrityAccess notes Universal and Warner already held partnerships with Stability before converting to equity, while Sony arrives as a new investor.

The equity track: the ai music companies the majors now own a piece of#

The investment thesis is stated almost as plainly as a term sheet. Stability's position, per MBW, is that "Artist-centric AI will only win if the product experience on a licensed platform is better than the experience on an unlicensed platform." Its Stable Audio 3.0 models generate tracks up to six minutes long, trained on licensed data, delivered through digital audio workstation plugins where working producers actually sit.

That is what ai music licensing looks like when it grows teeth: not a takedown regime but a product bet, funded by the rights holders themselves, that the legal catalogue plus better tooling beats the scraped catalogue plus a web demo. Dealroom's round record lists the returning backers, Coatue, Greycroft, Kadmos Capital, Sean Parker and Eric Schmidt, which tells you the financial market thinks the licensed lane is investable rather than charitable.

For music marketers the practical consequence is a clean rule: build campaign workflows on ai music companies whose training data is licensed and whose investors include the people who would otherwise sue them. The marketing translation is that brand-safe generative audio now has a named supply chain, and the buyers know it.

The litigation track: the rolling cipher claim#

The amended Suno complaint sharpens the biggest ai music lawsuit into something more dangerous than a fair-use fight. Per MBW, Judge F. Dennis Saylor IV granted leave on 18 August, and the labels now allege Suno "obtained recordings by bypassing YouTube's 'rolling cipher,' an encryption measure", using stream-ripping tools YT-DL and YT-DLP. That is a DMCA section 1201 circumvention claim, and circumvention does not care whether the downstream use was fair. The theory has legs beyond the majors: an earlier class action led by independent artist Tony Justice makes the same rolling-cipher allegation, so Suno is defending the acquisition route on two fronts at once.

The money at stake is structured in two tiers: up to $150,000 per infringed work, plus up to $2,500 per act of circumvention. The judge refused to let the labels bolt on 61,026 additional recordings, citing the "obvious consequences of complexity and delay", which keeps the case anchored on the original 560 works and a theoretical maximum around $84m rather than the $9bn-plus the expanded list implied.

Two dates now matter to every team building on ai music companies' tools: fact discovery closes on 30 September, and summary judgment on the fair-use question is expected next, per MBW. If the circumvention claim survives, the acquisition method becomes a liability separate from the training question, and every ai music lawsuit that follows will copy the template.

What the two-track week means for the ai music industry#

Are AI songs legal? The honest answer after this week: it depends entirely on the pipeline that made them. A track from a licensed platform, trained on cleared catalogue with the rights holders on the cap table, carries a fundamentally different risk profile from a track whose training data allegedly arrived through a bypassed cipher. Same output, opposite legal weather, and the gap between the two widened on 25 August.

The chart eligibility layer is moving the same direction: Australia's ARIA set eligibility rules for recordings made with AI from the chart dated 31 August, keeping wholly AI-generated tracks out while AI-assisted, substantially human work stays in. Provenance is becoming infrastructure, and ai music companies are sorting themselves into a documented lane and an undocumented one at every layer: training, charts, playlists, sync.

Meanwhile the platform economics underneath keep shifting: YouTube Premium's price rises roll out globally from 23 September, lifting most European tiers by roughly two euros. Subscription ARPU rising while training-data litigation tightens is the same story twice: the era of free-riding on recorded music, by listeners or by models, keeps getting more expensive.

The marketing read: pick your pipeline before it picks you#

For labels, artist teams and the brands that borrow music's glamour, the two-track week writes a procurement rule. Any campaign touching generative audio should be able to answer three questions in one email: which model made this, what was it trained on, and who licensed that catalogue. If the answers involve a shrug, the campaign inherits the litigation risk of the whole undocumented lane, and after the rolling-cipher claim that risk has a dollar figure attached per act, not just per work.

The sharper scent for marketers is on the opportunity side of the hedgerow. Licensed ai music companies now need exactly what the wider music business has always bought: positioning, differentiation and trust-building against louder unlicensed rivals whose product demos better than their depositions read. That is a brand story, a content story and an education story all at once, the ground folkfox works in music industry marketing, with brand strategy for the positioning, content marketing for the proof, and SEO and GEO so the answer engines cite the documented lane rather than the loud one. If your product lives on cleared catalogue and nobody can tell, that is fixable.

The vulpine read on the whole week: the majors stopped arguing about whether generative music should exist and started deciding who profits from it. Investment on one track, discovery deadlines on the other, and a quiet quarry in the middle: the workflow of every producer who will, within a year or two, make something with these tools without thinking twice. The companies that win that producer will be the ones whose paperwork survived this fortnight's scrutiny.

Frequently asked questions#