Music Industry Unveils AI Labels, but Streaming Giants Haven’t Signed On

Eight music industry organizations rolled out two voluntary AI labels on Friday, and the trade group representing Spotify, Apple Music and Amazon will not say whether its members plan to use them. The International Federation of the Phonographic Industry (IFPI) and the Recording Industry Association of America (RIAA) led the announcement, alongside the Grammys, SAG-AFTRA and five other groups.

The timing lines up with a flood of synthetic tracks that streaming services can no longer ignore. Deezer alone now takes in roughly 75,000 AI-generated tracks a day, and the new tags rely on the same voluntary, self-reported approach that has so far done little to catch the fraud driving much of that flood.

What Do the ‘AI-Generated’ and ‘AI-Assisted’ Labels Mean?

The coalition proposed two track-level tags on Friday: a black square marked with a capital ‘AI’ for recordings where generative AI created the entirety or the primary portion of a song, and a white square marked with a lowercase ‘ai’ for tracks that are substantially human made but lean on AI for some expressive elements.

The ‘AI-Generated’ tag covers songs built entirely from a text prompt, or where a machine produced the lead vocal or a key instrumental part. The ‘AI-Assisted’ tag requires that humans still perform the lead vocal and the primary instruments, even if AI touched other elements. For now, neither label covers AI use in lyrics, composition, music videos or cover art, according to the coalition’s joint announcement.

The full list of groups behind the push includes:

• International Federation of the Phonographic Industry (IFPI)
• Recording Industry Association of America (RIAA)
• American Association of Independent Music (A2IM)
• Worldwide Independent Network (WIN)
• IMPALA, the European Independent Music Companies Association
• The Recording Academy, which runs the Grammys
• SAG-AFTRA, the performers’ union
• The Human Artistry Campaign

Grammys chief executive Harvey Mason Jr. said artists and fans deserve a way to communicate how and when AI is being used. SAG-AFTRA’s national executive director, Duncan Crabtree-Ireland, put it more bluntly, saying performers deserve a marketplace that recognizes, values, and protects human creativity. Nobody has set a launch date. The groups said only that the labels are meant for broad, global adoption and will keep evolving as the technology changes.

Deezer Now Counts 75,000 AI Tracks a Day

The urgency behind Friday’s announcement shows up in the numbers. Deezer, the French streaming service, said in April that AI-generated tracks now make up 44% of everything uploaded to its platform daily, up from 39% in January and just 10,000 tracks a day when it first launched detection in January 2025. That is according to a Deezer newsroom announcement detailing the surge.

Apple Music is seeing something similar. Oliver Schusser, the platform’s vice president, told Billboard’s podcast that more than a third of the music delivered to Apple Music each month is fully AI-generated, though it draws under half a percent of listeners.

Fans, meanwhile, mostly cannot tell the difference. A Deezer and Ipsos survey of 9,000 people across eight countries found that 97% could not distinguish AI-made songs from human-made ones in a blind test, yet 80% wanted fully AI tracks clearly labeled. A separate study by CISAC (the International Confederation of Societies of Authors and Composers) and PMP Strategy found nearly a quarter of music creators’ revenue could be at risk by 2028, worth as much as four billion euros, according to the CISAC and PMP Strategy AI study.

The unease is not limited to music. Fans have also had to sort real from fake in a viral AI-generated Ranbir Kapoor trailer clip and in AI-generated hoaxes claiming false deaths among New Zealand athletes and their families, part of a wider pattern of synthetic content eroding trust across media.

Streaming Platforms Are Already Split

Before Friday’s announcement, streaming services had already split into two camps. Deezer detects and tags AI content itself. Apple Music and Spotify leave disclosure to labels and distributors. That divide will decide how much the new labels actually change.

Nothing about Friday’s proposal forces any of these five to change course. The two new tags are designed to sit on top of whatever system a platform already runs.

From a Grammy Snub to a Swedish Chart Ban

The labeling push follows several years of smaller skirmishes over AI music. Apple Music pulled the viral track “Heart on My Sleeve,” which used AI-cloned vocals of Drake and The Weeknd, after a copyright complaint over the misuse of the artists’ likeness and voice.

More recently, a song called “Jag vet, du är inte min” by an artist known as Jacub topped Spotify’s Swedish Top 50 chart, then was excluded from IFPI Sweden’s official Sverigetopplistan chart under rules barring works judged primarily AI-generated.

The legal backdrop runs deeper. Major labels have sued AI music generators Suno and Udio, accusing them of training on copyrighted recordings without permission or payment. Some of those fights have already turned into licenses instead of lawsuits, with Warner Music and Universal Music striking settlement and licensing deals with the AI firms over the past year.

Why Self-Reported Tags Won’t Catch the Fraud

Here is the gap in Friday’s plan. The labels depend on artists, labels and distributors to voluntarily declare AI use, the exact mechanism that legitimate musicians embrace and fraud operators ignore.

Deezer’s own numbers make the point. Up to 85% of streams on its fully AI-generated tracks were flagged as fraudulent, run up by bots to harvest royalty payments rather than to reach listeners. In March, a North Carolina man pleaded guilty to fraud after generating hundreds of thousands of AI songs and streaming them billions of times through bots, pocketing more than $8 million in royalty payments, according to Time’s reporting on musicians fighting AI slop. Nobody running a scheme like that is going to check a box marked ‘AI-Generated.’

Analysts who track the industry frame it as two competing philosophies rather than one shared standard, as laid out in a Water and Music analysis of AI labeling responsibility.

• Deezer’s camp argues that platform-level detection is the only model that actually catches fraud, since it does not depend on the uploader’s honesty.
• Spotify’s camp argues that a shared metadata standard, built through DDEX, scales better across millions of tracks than any single platform’s detection tool.
• DiMA, the Digital Media Association representing Spotify, Apple Music and Amazon, says it wants more detailed and accurate AI metadata before committing its members to specific tags.

RIAA chairman and chief executive Mitch Glazier defended keeping the system voluntary when he spoke with the Wall Street Journal about the announcement.

Transparency is just the best way to have it both ways.

Glazier, RIAA’s chairman and chief executive, made the remark while explaining why artists who want to use AI in their creative process should still be free to do so. Suno, the AI music generator named in the labels controversy, echoed that flexibility in its own statement Friday, saying it should be up to artists and platforms to decide how to treat these complex issues, while pointing to its own investments in watermarking and audio fingerprinting.

The Rollout Still Has No Fixed Date

Spotify declined to comment on the specific labels Friday, and Apple Music did not respond to requests for comment, according to reporting on the announcement. DiMA’s statement amounted to a cautious welcome rather than a commitment.

The coalition says it will work with distributors, aggregators and standard-setting bodies like DDEX on industry-wide implementation, but offered no timeline for when fans might actually see a black or white square next to a song title. Deezer, for its part, already licenses its detection technology to other platforms and in June launched a cross-platform tool that scans playlists for AI tracks on any streaming service, a sign of what enforcement-first labeling can already do without waiting on a voluntary standard.

Whether the new tags become mandatory depends on how the labels perform once services start testing them, and none have promised a start date yet.

Frequently Asked Questions

What is the difference between the ‘AI-Generated’ and ‘AI-Assisted’ labels?

The ‘AI-Generated’ tag is a black square with a white capital ‘AI’ for tracks where AI created the entire song or its lead vocal and key instrumentals. The ‘AI-Assisted’ tag is a white square with a lowercase ‘ai’ for tracks where humans still performed the lead vocal and primary instruments but used AI for other expressive elements.

Will Spotify and Apple Music actually display the new AI labels?

Neither has committed. Spotify declined to comment on the specific proposal Friday, and DiMA, which represents both platforms along with Amazon, said only that it wants more detailed AI metadata before its members act.

Do the labels cover AI-generated lyrics or cover art?

No. The system currently applies only to sound recordings. Generative AI used in lyrics, composition, music videos or cover art falls outside the labels for now, though the coalition says the framework is built to expand.

How much AI-generated music is already on streaming platforms?

Deezer receives about 75,000 fully AI-generated tracks daily, or 44% of everything uploaded, up from just 10,000 a day in January 2025. Apple Music has said more than a third of its monthly deliveries are fully AI-generated, though those tracks draw under half a percent of listeners.

Was an AI-generated song ever blocked from a music chart?

Yes. A track called “Jag vet, du är inte min” by Jacub topped Spotify’s Swedish Top 50 chart in early 2026 but was removed from IFPI Sweden’s official Sverigetopplistan ranking after being judged primarily AI-generated.

Why do critics say voluntary labels won’t stop AI music fraud?

Because the tags rely on self-reporting by artists, labels and distributors, the same people running streaming fraud schemes have no reason to disclose it. Deezer has found up to 85% of streams on its fully AI-generated tracks were fraudulent, and a North Carolina man pleaded guilty in March to pocketing more than $8 million through bot-driven AI song streams.