The music industry just moved from arguing about AI music to labeling it. In July 2026, a coalition of eight major music organizations — IFPI, the RIAA, A2IM, WIN, IMPALA, the Recording Academy, SAG-AFTRA, and the Human Artistry Campaign — proposed a shared system for labeling AI-generated music on streaming services.
If you release music and you use AI tools anywhere in your process — Suno, Udio, AI vocals, AI stems, AI mastering — this proposal describes the world your next releases will land in. Here is what it actually says, what it deliberately leaves out, and what you can do about it today.
The two labels, in plain terms
The proposal defines two voluntary labels for sound recordings.
AI-Generated applies when generative AI produced the entirety or the primary portion of a recording's creative elements. The examples given are an AI-generated lead vocal, an AI-generated key instrumental, or music generated entirely from prompts. If the core of what listeners hear came out of a model, the recording is AI-Generated.
AI-Assisted applies to recordings created substantially by humans — where humans perform the lead vocal and the primary instruments — but where generative AI was used for some expressive elements along the way.
Two details matter more than the headlines suggest. First, the labels cover sound recordings only. AI-written lyrics, AI-assisted composition, AI cover art, and AI music videos are explicitly out of scope for now. A fully human recording of an AI-written song would not carry either label under the current proposal.
Second, the line between the two labels runs through the lead vocal and primary instruments. That is a stricter test than most artists assume. If your voice is a model's voice, you are on the AI-Generated side of the line no matter how much human work surrounds it.
Who has to follow this?
Nobody, yet. The labels are voluntary and proposed, not law. The organizations behind them say they will work with streaming services, distributors, aggregators and standards bodies on implementation.
The early signals, though, point in one direction. DiMA — the association representing Spotify, Apple Music, Amazon Music and YouTube — responded that it looks forward to receiving more detailed and accurate AI metadata, and pointed to DDEX, the industry's metadata standards body, as part of the path. Spotify has already committed to supporting the DDEX standard for AI disclosures in music credits and has tested AI tags in song credits where artists disclose through their label or distributor. Deezer has been detecting and tagging AI music at platform level since 2025. TIDAL and Qobuz run their own tagging policies. Apple Music launched a declaration-based tagging system in March.
Different platforms, different mechanisms — but the same trajectory: the question "was AI used in this recording?" is becoming a standard field in music delivery, the way explicit-content flags became standard years ago.
Why now? The volume made it unavoidable
The scale of AI music is no longer a prediction. Deezer has reported receiving close to 75,000 fully AI-generated tracks per day — more than 44% of all new music delivered to the platform. Apple Music has said more than a third of tracks uploaded to it are entirely AI-made.
At that volume, "we'll deal with it case by case" stops being an option for platforms. Labels, in both senses of the word, become infrastructure. Even Suno — currently in litigation with major labels — responded to the proposal by saying transparency is important and pointing to its own investment in watermarking, fingerprinting and disclosure tools.
Whatever you think of any individual player in that story, the direction is set by everyone at once: creators will increasingly be expected to say, on the record, how AI was used in a recording.
The gap in the proposal — and why it matters to you
Here is the part the announcements do not dwell on: a label on a streaming service is the end of a chain. Someone at the start of that chain has to make the declaration — and the declaration is only as good as the record behind it.
DiMA said it plainly: this information flows best when it travels the entire path from creator to fan. That path starts with you, at release time, stating what was AI-generated, what was AI-assisted, and what was entirely yours — in a form that survives distributor handoffs, platform ingestion, and time.
That raises practical questions the labeling proposal leaves open:
- When is the declaration made? A statement recorded at release time, before any dispute exists, carries different weight than one reconstructed months later when a platform or distributor asks.
- What exactly was declared? "AI was used" is not the same as a structured record of which tools, in which parts of the process — vocals, composition, stems, mastering.
- Can anyone check what you declared and when? A declaration that lives in an email thread is hard to point to. A declaration on a dated, independently timestamped record is something you can reference in one link.
What you can do today
You do not need to wait for platforms to finalize anything to get ahead of this:
- Decide honestly which side of the line each track is on. Use the proposal's own test: who performs the lead vocal and primary instruments? If the answer is a model, that is AI-Generated under the proposed definitions — regardless of how much human curation went into it.
- Keep your AI usage documented per track, not per artist. Which tools, which elements, which versions. Your future self — and your distributor's future compliance form — will need specifics.
- Record your declaration when you release, not when you're asked. A dated disclosure made before a track takes off is the cleanest position you can be in when tagging systems, distributors, or listeners start asking questions.
- Watch your distributor's metadata requirements. With DDEX AI-disclosure fields already supported by major platforms, expect AI-usage questions to appear in upload forms the way explicit-content checkboxes did.
Transparency is moving from a personal choice to an industry expectation. The artists who will find that transition painless are the ones whose declarations already exist — structured, dated, and easy to point to.
Audiverify records AI-involvement declarations as part of a timestamped documentation certificate — including the distinction between AI-generated and AI-assisted workflows, the tools used, and the areas of the track they touched. The declaration is recorded as submitted; Audiverify documents the record, it does not adjudicate it.
Related reading
Audiverify
Cryptographic fingerprinting, AI disclosure documentation, and dispute-ready evidence workflows for professional music releases.