Eight music trade groups agreed on two AI labels for streaming: “AI-Generated” and “AI-Assisted”

The trade groups that speak for most of the world's recorded music have settled on a common way to tell you when a song was made by a machine. On July 10, IFPI, the RIAA, A2IM, WIN, IMPALA, the Grammys, SAG-AFTRA and the Human Artistry Campaign put out a joint plan for two track-level tags: AI-Generated and AI-Assisted. The icons are plain, a solid black "AI" for the first and an outlined "ai" for the second, and the idea is that one of them rides along next to a track once streaming services agree to show it.

The split between the two comes down to who actually performed the thing.

AI-Generated is for a recording where generative AI made all or the primary part of the creative elements. The examples the groups give are an AI lead vocal, an AI key instrumental take, or a track that is entirely prompt-generated. AI-Assisted is for a recording that humans wrote and performed, human lead vocal and human primary instruments, where AI was used on some expressive piece along the way. So a singer who ran a synth patch through an AI tool is assisted. A prompt that spat out a full song with a fake voice on top is generated.

The reason this is landing now is a number. In April, Deezer said AI-generated tracks made up 44% of all new music uploaded to its platform. Apple Music has put more than a third of its incoming uploads in the "100% AI" bucket. That is the wave the labels are built for, a catalog filling up faster with synthetic tracks than any human staff could ever tag by hand.

"Fans want to know whether and how generative AI has been used in the music to which they listen," IFPI CEO Vikki Oakley and RIAA chairman Mitch Glazier said in a joint statement. Harvey Mason jr., who runs the Grammys, framed it around credit: "This initiative ensures that creativity, authorship, and artistic intent remain at the center of every song." SAG-AFTRA's Duncan Crabtree-Ireland put the sharper point on it. "Transparency is essential, but it is only the beginning," he said, adding that AI "should not be used to replace, imitate, or exploit artists without consent and fair compensation."

Here is the catch, and it is a big one. The labeling is voluntary. Artists, labels and distributors flag their own tracks. Nobody is scanning the upload pipe and stamping the synthetic stuff on the way in. Think about who is least likely to raise a hand: the person pumping out prompt-generated tracks by the hundred to farm playlist royalties. The 44% of Deezer's uploads that triggered this whole conversation is exactly the material with the least reason to volunteer. An honest indie who used an AI harmony on one bridge will tag it. The synthetic-track factory will not.

The scope is narrow too, at least for now. The labels cover the sound recording only. They say nothing about AI in the lyrics, the composition, the music video or the cover art. And the services themselves have not signed on to display anything yet. The groups say they will "work with digital music services, distributors, aggregators and standard-setting bodies" on rolling it out, which is the polite way of saying the hard part has not started.

None of that makes the move pointless. A shared vocabulary is worth having, and getting IFPI, the majors, the independents, the Grammys and a performers' union to agree on two definitions and two icons is not nothing. It also lines up with where the rest of the industry is already walking. Tidal is cutting royalty payments on fully AI-generated songs starting July 15, and the NO FAKES Act, aimed at AI voice and likeness clones, cleared a Senate hurdle last month.

But a label is only as good as the honesty of whoever fills it in, or the platform willing to apply it when they do not. Right now this is an honor system pointed straight at the corner of the catalog with the least honor in it. Get Spotify, Apple and the rest to actually show the tag, and pair it with detection instead of self-reporting, and it starts to mean something. Until then it is a good idea waiting on the people who need it least to go first.