Stability AI raised $76M from all three major labels to rebuild around music under Sean Parker, the Third Circuit delivered the first federal appeals ruling against AI training as fair use, and Suno launched a spoken-word tool while crossing $300M in annual revenue.

Sean Parker Is Rebuilding Stability AI as a Music Company

The Napster co-founder is doing it again — this time with the labels' blessing. A TechCrunch report published Thursday reveals Sean Parker's vision for Stability AI: pivot the once-floundering image-generation startup into the go-to AI toolmaker for music professionals. The company announced $76 million in new funding from Sony, Warner, and Universal, who also licensed their catalogs for model training as part of the deal.

Stability has already shipped three new audio models (Stable Audio 3.0) and a DAW plugin for iterative music editing, now in public beta. The largest model generates professional-grade tracks up to six minutes long. Parker says an upcoming update will let users guide generation by humming a melody or beatboxing a drum pattern — a far more intuitive interface for working musicians than typing text prompts. For anyone who lived through the Napster era, watching Parker build a music AI company with the majors rather than against them is quite the arc.

Source: TechCrunch, Music Business Worldwide

Suno Launches Speech: Spoken Word Meets Generated Music

Suno's first product beyond song generation dropped Wednesday. Speech (beta) generates spoken-word audio and an original backing score together in a single track — type a poem, a bedtime story, or a meditation script, describe the voice and musical style, and it produces a cohesive piece without the usual stitch-two-layers-together workflow. The model generates voice and music as one integrated output, so the pacing and dynamics of the score actually respond to the speech.

Open to all Suno users on web, iOS, and Android after a month of closed testing. If you produce podcasts, guided meditations, audiobooks, or spoken-word content, this is worth trying even in beta — the integrated generation means you don't need to source and sync music separately.

Source: Bloomberg, Unite.ai, WebProNews

Third Circuit Delivers First Appeals Ruling Against AI Training as Fair Use

On September 29, the Third Circuit became the first federal appeals court to rule on whether training AI on copyrighted material qualifies as fair use — and the answer was no. In Thomson Reuters v. ROSS Intelligence, the court found that using Westlaw headnotes to train a competing legal research AI was "highly commercial and only minimally transformative." Forbes ran the headline: "AI Lost Its First Fair Use Appeal. Suno Should Be Worried."

The case wasn't about music, but every music AI lawsuit pending now has this opinion in hand. The court was careful to note its facts differ from generative AI cases, leaving the bigger question technically open. But the direction is clear: the first appellate court to weigh in sided with rights holders. Combined with Suno's loss to GEMA in Munich this summer, the legal walls are closing in on anyone still training on unlicensed content.

Source: Forbes, Courthouse News, IPWatchdog, Ballard Spahr

Suno Crosses $300M ARR and 2 Million Subscribers

Alongside the Speech launch, CEO Mikey Shulman disclosed that Suno has surpassed two million paying subscribers and $300 million in annual recurring revenue — making it one of the fastest-growing consumer AI products, not just in music. For a company that launched publicly less than two years ago, those numbers are striking.

The growth is particularly notable given the rocky v6 launch. Suno retired all older models when v6 shipped September 9, built on licensed catalogs from Warner, BMG, and Believe. User reception was mixed — some called the output "muffled" and "generic" compared to earlier versions, with several canceling subscriptions. The subscriber count kept climbing regardless, which tells you something about demand for the category even when quality is debated.

Source: Bloomberg, CryptoBriefing

Engram: An AI Sampler That Deliberately Breaks Its Own Models

While most AI music tools chase polish, Engram goes the other direction. This hardware sampler and groovebox from Thoughtful Things runs a tiny, locally hosted neural network and deliberately pushes it past its limits — think circuit bending, but for generative audio models. The result is uncanny, glitch-laden textures and hallucinated sounds that you sample and sequence in real time, entirely offline.

No cloud, no subscription, no copyrighted training data — the models were trained exclusively on CC-BY licensed datasets, and the firmware will be open-sourced so users can load their own models. Currently on Kickstarter at $770 for a first-run unit, already three-quarters funded. If you do experimental sound design or glitch music, this is one of the more genuinely interesting hardware concepts to emerge from the AI audio space.

Source: MusicRadar, The Verge
Tool of the week
Stability AI DAW Plugin (Public Beta)

With Sean Parker's music pivot making headlines, the practical product worth trying right now is Stability AI's DAW plugin for Stable Audio 3.0. It brings text-to-music generation directly into your production workflow — type a prompt, generate a clip, edit segments inline, extend tracks with causal continuation, and export. The plugin supports single-segment and multi-segment editing, meaning you can regenerate just the bridge without touching the verse.

The underlying model (Stable Audio 3.0 Large) generates up to 6 minutes and 20 seconds of coherent music and maintains structure across that length. Three smaller models are available as open weights for self-hosting if you want to run locally. The DAW plugin is free during the public beta; the web app runs on Free / $11.99/mo Pro / $29.99/mo Studio tiers. API access is $0.26 per generation.

Free during beta · DAW plugin + Web + API · platform.stability.ai
Worth watching
The Great Licensing Convergence

Look at this week's news through one lens and a pattern emerges: every major AI music company is now licensing from the labels. Stability AI just signed Sony, Warner, and Universal. Suno trained v6 on Warner, BMG, and Believe. Udio has deals with Universal, Warner, Merlin, and Kobalt. Google trained Lyria 3.5 exclusively on licensed content. The era of training on scraped audio and arguing fair use later appears to be ending — accelerated by the Third Circuit ruling and Suno's loss to GEMA in Germany.

What remains to be seen is whether licensed-only models can match the quality of ones trained on the open internet — Suno's v6 backlash suggests the answer isn't obvious — and whether indie artists and smaller labels get meaningful revenue from these deals or get left out entirely. The majors are getting paid. Everyone else is still waiting.