ElevenLabs ships section-level song editing, Adobe brings AI music and SFX to Firefly, streaming platforms begin labeling and excluding AI-generated tracks, and Round Hill Music files billion-dollar lawsuits against Suno and Anthropic.
ElevenLabs Ships Composer: Edit Songs Section by Section
ElevenLabs launched Composer, a section-by-section editor inside ElevenMusic that lets you rewrite a verse, swap a chorus, or extend a bridge without touching the rest of the track. You can generate multiple takes per section and rearrange song structure by dragging sections around. It's built on Music v2, trained exclusively on licensed data from Merlin and Kobalt — every output cleared for commercial use.
This matters if you use AI generation for scratch tracks, demos, or background music. The ability to iterate on individual sections rather than regenerating an entire song is a genuine workflow improvement. Whether the musical quality justifies replacing a human session musician is still debatable, but the editing paradigm is heading in the right direction.
Adobe Firefly Audio Goes GA: Music, Speech, and Sound Effects in One Platform
Adobe made three audio generation tools generally available inside Firefly: Generate Music (original tracks tuned to video length and mood), Generate Speech (script-to-voiceover with ElevenLabs integration), and Generate Sound Effects (custom sounds matched to video action and timing). All models trained on licensed Adobe Stock and public domain content, with enterprise IP indemnification available.
If you work in video or content production, this is notable — not because any single capability is best-in-class, but because having music, VO, and SFX generation inside a platform you likely already use removes friction. Free tier with daily credits; paid plans from $9.99/month. Sound effects consume premium credits.
Suno Studio 2.0 Adds MIDI, Wavetable Synth, and an AI Plugin Builder
Suno released Studio 2.0, adding MIDI recording, editing, and import — its most-requested feature since Studio launched last September. MIDI clips can now serve as prompts for new AI audio generations. The update also includes a built-in wavetable synth, automation curves, sidechain compression, convolution reverb, updated stem separation, and unlimited 32-bit/48kHz multitrack export.
The most interesting addition is the AI chat bar (beta) that can build its own plugins on the fly. Suno is clearly trying to become a full DAW rather than just a generation toy. Whether working producers will switch from Ableton or Logic for this remains unlikely, but for creators who start in Suno and want more control, it closes real gaps. Premier subscribers only ($24/month).
Charts and Platforms Draw the Line on AI Music
Three major moves in two weeks. Australia's ARIA banned fully AI-generated tracks from official charts after an AI variation of Madonna's "Like a Prayer" spent 16 weeks in the top 20. Apple Music confirmed "Made With AI" labels are coming before year-end — distributors must tag tracks where "a material portion" was AI-created. And Spotify introduced "AI Persona" badges that will exclude flagged artists from editorial and algorithmic recommendations, rolling out mid-September.
The pattern is clear: platforms aren't banning AI music, they're making it visible and limiting its algorithmic reach. For human artists, this is protective. For AI-assisted producers working with real instruments and vocals, the key question is where the "substantially human-made" line gets drawn. ARIA's and Apple's definitions are vague enough to create uncertainty. The IFPI's global framework, now adopted by 20+ chart programs, tries to set a floor, but enforcement will vary by market.
40% of Music Released in July Had AI Involvement
SubmitHub's AI detector analyzed over one million releases from July 2026 and found 23.2% were fully AI-generated, with another 15.3% containing AI-generated audio modified by humans — 38.5% total. Separately, Deezer confirmed AI-generated tracks now exceed 50% of daily uploads (roughly 90,000 tracks/day), and Apple Music's chief said about a third of new uploads are AI-generated.
The important caveat: volume doesn't equal consumption. AI tracks account for just 1–3% of streams on Deezer and 0.5% on Apple Music. The problem isn't that listeners want AI music — it's that the flood creates noise, dilutes discovery, and enables streaming fraud. Deezer found 85% of streams on fully AI-generated tracks were fraudulent in 2025. The platform labeling moves above are a direct response to these numbers.
Round Hill Music Files Billion-Dollar Suits Against Suno and Anthropic
Independent publisher Round Hill Music (managing a $1.1B portfolio including "Iris," "Total Eclipse of the Heart," and "Lola") filed separate copyright suits against Suno and Anthropic, alleging unauthorized training on 500+ songs with plans to expand to 10,000+. CEO Josh Gruss said the company will go to trial and won't accept settlements that undercompensate songwriters. Potential damages exceed $1 billion per suit.
The UMG/Sony v. Suno case also advanced: a judge allowed a DMCA anti-circumvention claim alleging Suno used stream ripping to scrape YouTube for training data, though the court denied expanding the case from 560 to 61,000+ recordings. Fact discovery closes September 30. Combined with the GEMA v. Suno ruling in Munich — the first European AI music judgment finding Suno liable — the legal pressure is compounding from multiple jurisdictions simultaneously.
Suno-BMG Deal Signals the Licensed AI Model Era
BMG and Suno announced a global strategic alliance covering both recorded music and publishing repertoire. The deal retroactively settles prior use of BMG's catalog and feeds Suno's upcoming launch of its first model trained in partnership with the music industry. BMG joins Warner Music Group, which signed with Suno last November. Artists and songwriters must opt in.
Separately, Spotify signed Kobalt (the world's largest independent publisher) for its upcoming AI covers and remixes tool — joining UMG and Merlin. The tool will be a paid Premium add-on built on consent, credit, and compensation. The industry is bifurcating: one track of licensed, compensated AI music tools, another of unlicensed platforms facing billion-dollar lawsuits. If you're building with AI audio, which side of that line your tools sit on increasingly matters.
Stability AI released a beta DAW plugin that generates audio directly on a track inside Ableton Live and Logic Pro. The plugin uses the Stable Audio 3 Large model via cloud, sharing your web account's credit pool. Strongest for instrumentals, textures, loops, beds, and sound design — no vocals or lyrics. Variable-length generation up to 6 minutes 20 seconds of stereo audio.
The real draw is the open-weight release: three models (Small SFX, Small, Medium) are free on Hugging Face, and the Small model runs locally on a MacBook Pro with no cloud dependency. Community developers have already built alternative AU/VST wrappers (DAWalka on GitHub) for Reaper and Bitwig. If you do sound design or need procedural audio textures, this is the most accessible way to get generative audio into your existing workflow.
A chain is forming: EU AI Act Article 50 (now enforceable) requires watermarks in AI-generated audio. Suno and OpenAI are embedding SynthID watermarks. Platforms are adding visible labels (Spotify, Apple, Deezer). Charts are excluding unlabeled or fully AI-generated tracks (ARIA, IFPI's 20+ programs). Each layer depends on the one below it. This isn't a set of independent policy decisions — it's a detection-to-exclusion pipeline being built in real time.
The question is whether it holds. Watermarks can be stripped. Detection tools produce false positives. The "substantially human-made" threshold is undefined. And producers who use AI for stems, arrangement, or mixing — but play and sing themselves — risk being caught in a system designed for fully generated slop. The next six months will determine whether this infrastructure becomes a precise filter or a blunt instrument.