In November 2025, a country song credited to an artist who did not exist reached No. 1 on a Billboard chart.
"Walk My Walk," released under the AI-created identity Breaking Rust, sold roughly 3,000 downloads in a week and topped Billboard's Country Digital Song Sales chart. The number was modest. The symbolism wasn't — a fully synthetic performer had walked onto the same scoreboard used to measure human careers.
The industry spent months arguing about what should happen next. On July 29, 2026, some of its largest companies answered.
Sony Music, Universal Music Group and Warner Music Group joined Believe, BMG, Concord, Dirty Hit, Glassnote Records, HYBE, Mom+Pop Music and Partisan Records in proposing global eligibility principles for recordings made with generative AI. Their position is more interesting than a ban: music involving AI could stay chart-eligible, provided it is substantially human-made, legally produced, transparently labeled and free from manipulation concerns.
One important caveat before anything else. As of July 30, no major chart compiler has publicly adopted this as a binding rule. It's an industry proposal, not chart law. But it's worth reading closely anyway, because the six criteria show exactly where distribution policy is heading — and where the awkward questions still sit.
What the proposed rules actually say
Under the framework, a recording involving generative AI should be excluded from an official chart where there's reason to believe it fails any of six conditions:
- The generative-AI service used to develop the recording is properly authorized and lawful.
- The recording is substantially human-made.
- It doesn't raise concerns about streaming or chart manipulation.
- It complies with copyright, related rights, personality rights and other applicable laws.
- Releasing it doesn't violate the terms of the AI service used.
- Its use of generative AI is appropriately disclosed to listeners through streaming services or other downstream platforms.
What the coalition is trying to separate is clear enough: an artist using AI inside a legitimate creative process, versus someone generating thousands of anonymous tracks with an unlicensed model, attaching fictional identities and feeding the results into a streaming farm. That distinction is sensible. Making it operational is where it gets hard.
"Substantially human-made" is carrying most of the weight
Picture a singer who writes the melody and lyrics, performs the lead vocal and records a guitarist, but uses generative AI to build a string arrangement under the chorus. Almost everyone would call that human-led work.
Now change the session. The singer writes the prompt and edits the lyrics; a generator produces the melody, the lead vocal, the instrumentation and the final arrangement. A human chose the result — possibly after hundreds of attempts. Is that substantial human creativity?
The principles don't establish a percentage, a scoring method or an evidence test, and they don't say which musical elements should count for more. That leaves a whole field of border cases:
- an AI-generated lead vocal with human-written lyrics;
- a human performance over entirely generated instrumentation;
- generated backing vocals replacing hired singers;
- an AI arrangement later reconstructed by human musicians;
- a synthetic demo that gets edited, mixed and partially re-recorded;
- an instrumental built from generated stems plus human overdubs.
The industry has at least started building vocabulary for this. On July 10, IFPI, RIAA, A2IM, WIN, IMPALA, the Recording Academy, SAG-AFTRA and others announced two voluntary track-level labels:
- AI-Generated — generative AI created the entirety or the primary portion of the recording's expressive elements. Published examples include an AI lead vocal, a key AI-generated instrumental performance, or entirely prompt-generated music.
- AI-Assisted — the recording is created substantially by humans, with generative AI used for some expressive elements while humans perform the lead vocal and primary instruments.
Useful as a starting point, but not yet a chart test — and there's a gap worth knowing about: the labeling system doesn't currently cover AI use in composition, lyrics, cover artwork or music videos. So a producer could legitimately have an "AI-assisted" sound recording built around AI-generated lyrics. The label describes the master, not every creative decision behind the song.
"Authorized tool" is not a checkbox
The coalition also wants chart-eligible recordings to use properly authorized and lawful AI services. Good principle — genuinely difficult to answer on a delivery form.
An AI company might license one label's catalog without securing equivalent rights from every other rightsholder represented in its training data. A model can be lawful in one jurisdiction while facing unresolved litigation in another. Old and new versions of the same service may have been trained under completely different arrangements. And no independent artist is going to audit a training dataset.
What you can do is read what the provider claims:
- Does it identify the source of its training material?
- Does it claim licensed, public-domain or user-authorized data?
- Do the commercial terms grant you the release rights you need?
- Does the license cover your subscription tier specifically?
- Are generated vocals restricted from imitating identifiable people?
- Can the output be distributed, monetized, and registered with automated rights-management systems?
- Could the provider change your rights after you cancel?
- Is there an indemnity — or do you carry all the legal risk?
Don't assume a premium plan makes the underlying model authorized. A subscription buys access under the provider's terms; it proves nothing about how the model was trained. This is one of the framework's genuinely unsolved problems, because chart compilers would need a defensible way to decide which services qualify without running a copyright trial for every release.
Why charts are treating AI and fraud as one problem
Generative AI produces recordings fast and cheap. Streaming fraud manufactures the appearance of demand. Put them together and you have an efficient machine for stuffing catalogs and pointing automated plays at them.
The link shouldn't be overstated — human-made music gets fraudulently streamed too, and an AI-generated track can find real listeners. But the proposed principles put manipulation at the center of eligibility for a reason.
Deezer reported on July 21 that it was receiving an average of roughly 90,000 fully AI-generated tracks per day, exceeding 50% of new deliveries on peak days in June. Yet fully synthetic music accounted for only 1% to 3% of actual listening on the service. Deezer also says up to 85% of streams tied to fully AI-generated tracks in 2025 were fraudulent. Read those as figures from Deezer's own detection and fraud systems, not an independent measurement of the whole market.
The gap between uploads and listening is the interesting part. This is a supply event, not a demand event — enormous volumes of cheap content entering catalogs with no matching audience.
And that's precisely when charts become vulnerable, because a small amount of coordinated activity can produce a visible position. Breaking Rust's No. 1 came on a digital-sales chart with around 3,000 purchases, per Billboard's reporting. There was no public finding that those sales were fraudulent — the case simply showed how a narrow chart can become culturally significant on very little volume. Once the position exists, it becomes marketing: No. 1 song, chart-topping artist, historic breakthrough. The achievement can generate far more attention than the listening that produced it, and that feedback loop is what chart organizations are trying to protect.
Chart eligibility, distribution and royalties are three different decisions
A track excluded from a chart doesn't vanish from streaming services. A distributor may accept it, a platform may host it, listeners may stream it, and depending on the service's policy it might still earn royalties. These are separate systems, and the proposal only addresses recognition by official charts — not distribution policy, not copyright ownership, not whether platforms must monetize synthetic music.
Services are already diverging. Deezer tags detected AI music, keeps it out of editorial and algorithmic recommendations, and excludes fraudulent streams from royalty calculations. TIDAL has announced a stricter line: tracks it identifies as fully AI-generated are labeled and don't earn royalties. Others may lean more on what labels and distributors declare.
Which means the same recording can end up with several different statuses at once:
- accepted for distribution;
- labeled AI-generated or AI-assisted;
- restricted from recommendations;
- eligible or ineligible for royalties;
- accepted or rejected by a chart;
- eligible or ineligible for awards;
- fine in one territory, challenged in another.
A successful upload only answers the first one.
Don't hide your workflow
Some musicians will respond to all this by simply not disclosing. That's short-sighted.
If charts eventually adopt these principles, a questionable release can be reviewed after it's already performing. At that point you may be asked which tools were used, who performed each element, whether the model was authorized, and how the release was labeled. "We don't remember" is not a strong answer.
So for every commercially released recording involving generative AI, keep a simple production record:
- the tool and model version;
- the date it was used;
- the account and subscription tier;
- the relevant terms of service;
- what material you supplied to the system;
- which elements it generated;
- what humans wrote, performed, selected and edited;
- contributor consent;
- licenses for generated voices, samples and source material;
- the original outputs alongside the final master;
- the AI disclosure delivered through your distributor;
- the campaign and traffic sources used to promote the release.
This doesn't need to be a philosophical diary. Two accurate pages stored with the session will usually do. The test is simple: could your team reconstruct how the recording was made if a platform challenged it six months from now?
A human credit can't be decorative
There's an obvious weakness in any "substantially human-made" standard — a company can drop a human name next to a synthetic production and call that person creative director. For the standard to mean anything, human involvement has to be visible in the work and backed by evidence.
The Recording Academy already applies a comparable principle. Its current guidance says only human creators can be nominated or win, while works containing AI-generated material may qualify where there's meaningful human authorship or performance in the relevant category — in a songwriting category, for instance, the majority of the song must be written by a human.
Charts measure consumption rather than artistic merit, so they don't need to copy Grammy rules. But the comparison shows something useful: "human involvement" can be judged against the specific contribution being recognized. A producer who wrote, performed and shaped a record can describe those decisions. Adding a hi-hat to a finished prompt-generated master shouldn't convert the project into human-led music.
That's editorial judgment, not a rule any chart body has adopted. Whatever gets built will need language strong enough to stop cosmetic human credits without punishing legitimate experimentation.
What independent artists should do now
Nobody needs to redesign a campaign because a coalition published a proposal yesterday. But if you use generative AI, prepare for disclosure and provenance to become ordinary parts of music delivery.
- Classify the actual use. Separate routine machine-learning tools like noise removal from generative systems creating new expressive material.
- Confirm the human contributions. Document who wrote, sang, played and arranged — and what made the work creative rather than merely functional.
- Review the AI provider. Keep its terms, its commercial-use permission and its statements about training or licensing.
- Clear identities and rights. Consent is essential whenever a synthetic performance resembles a real singer, musician or other identifiable person.
- Disclose accurately. When a distributor offers AI-generated and AI-assisted fields, answer according to the recording — not according to which label looks better in marketing.
- Keep promotion clean. Don't buy guaranteed streams, downloads, chart positions or playlist placements. A legitimate recording can lose eligibility purely because of fraudulent marketing.
- Preserve the evidence. Save agreements, stems, source performances, project files and generated outputs alongside the master.
- Watch individual chart rules. Adoption may vary between Billboard, national chart organizations and platform-specific rankings.
That last point deserves emphasis. There may never be one global moment when "the chart rules" change — different compilers could interpret the same principles differently, or decline to adopt them at all.
Where CREWPORT fits
Most of this comes down to delivering accurate information and being able to prove it later. CREWPORT validates your metadata before delivery and keeps your identifiers and release data together, so the disclosure you send matches the record you actually made — and so you can still find the paperwork when someone asks about it six months from now. Disclosure fields are appearing across distributors; answer them from the session, not from the marketing plan.
The proposal draws a boundary, not yet a map
The industry's message is getting clearer: generative AI can participate in human-led music, while purely synthetic recordings made with unauthorized tools, hidden from listeners or pushed through manipulated activity shouldn't get the same public recognition.
Plenty of artists will agree with that. Applying it fairly is the actual test, and the open questions are substantial. Who certifies an authorized model? How much human contribution counts as substantial? What happens when a track was made with several tools? Can an artist appeal a wrong classification? Does eligibility change if a model later becomes the subject of a lawsuit? Who receives the production evidence, and how is confidential session material protected?
None of that makes the proposal useless — it shows what still has to be built.
For independent artists, the safest response is wonderfully unglamorous: know how the record was made, know what the tools allowed, tell the truth in the metadata, and keep the paperwork. If the song ever reaches a chart, its history should survive the scrutiny that follows.
FAQ
Can AI-generated music chart right now?
Yes — no major chart compiler has publicly adopted the new principles as a binding rule as of July 30, 2026. Breaking Rust topped a Billboard digital-sales chart in November 2025 under existing rules. The proposal would change that for recordings that aren't substantially human-made.
What's the difference between "AI-Generated" and "AI-Assisted"?
Under the voluntary labels announced July 10, AI-Generated means generative AI created all or the primary portion of the recording's expressive elements. AI-Assisted means humans created it substantially — performing the lead vocal and primary instruments — with AI used for some expressive elements.
If a track is excluded from a chart, can it still be distributed?
Yes. Chart eligibility, distribution and royalties are separate systems. A distributor may accept the track and platforms may host it even if a chart won't count it.
Do AI-generated tracks earn royalties?
It depends on the platform. TIDAL has announced that tracks it identifies as fully AI-generated are labeled and don't earn royalties. Deezer tags AI music, removes it from recommendations, and excludes fraudulent streams from royalty calculations. Policies differ and are still changing.
Do I have to disclose AI use?
Disclosure is one of the six proposed criteria, and platform-level labeling already exists on a voluntary basis. Practically: answer distributor disclosure fields accurately based on the recording, and keep a production record you could show later.
Does using AI mean I lose copyright in my song?
The chart proposal doesn't determine copyright ownership — that's a separate legal question that varies by jurisdiction and depends on the human authorship in the work. Get advice specific to your recording if it matters commercially.
Keep your release provable
Every one of these rules eventually resolves to the same thing: accurate information attached to the recording, and evidence you can still find later.
CREWPORT validates your metadata before delivery and keeps your identifiers, credits and release data together — so what you declare matches what you made.
This article explains a developing industry proposal; it isn't legal advice. Chart rules, platform policies and copyright treatment are all still moving — verify current details before making decisions that depend on them.
Sources
- Music Business Worldwide — Music Companies Propose AI Chart Eligibility Principles, July 29, 2026
- IFPI — Music Community Introduces Generative-AI Track Labels, July 10, 2026
- Deezer — AI Music Surpasses 50% of New Uploads, July 21, 2026
- Billboard — Breaking Rust Leads Country Digital Song Sales
- The Recording Academy — Current Grammy Guidance on AI Eligibility