When the Recording Academy announced its latest updates to the Grammy Awards rules last month, there was a rule that seemed like it should be simple, but was anything but: It took some 202 words to explain that with the advent of AI, only music genuinely created by humans will be eligible to be nominated for or win a Grammy.

Yes, recordings that use one of the multiple types of AI technology will be eligible, but words and melodies essentially written by AI — a.k.a. generative AI — are not.

“We’re not going to be giving a nomination or an award to an AI computer or someone who just prompted AI,” Recording Academy CEO Harvey Mason Jr. clarified to Variety. “That’s the distinction that we’re trying to make. It’s the human award highlighting excellence, driven by human creativity.”

The full wording of the ruling follows: The GRAMMY Award recognizes creative excellence. Only human creators are eligible to be submitted for consideration for, nominated for, or win a GRAMMY Award. A work that contains no human authorship is not eligible in any Category.

A work that features elements of A.I. material (i.e., material generated by the use of artificial intelligence technology) is eligible in applicable Categories; however: (1) the human authorship component of the work submitted must be meaningful and more than de minimis; (2) such human authorship component must be relevant to the Category in which such work is entered (e.g., if the work is submitted in a songwriting Category, there must be meaningful and more than de minimis human authorship in respect of the music and/or lyrics; if the work is submitted in a performance Category, there must be meaningful and more than de minimis human authorship in respect of the performance); and (3) the author(s) of any A.I. material incorporated into the work are not eligible to be nominees or GRAMMY recipients insofar as their contribution to the portion of the work that consists of such A.I material is concerned. De minimis is defined as lacking significance or importance; so minor as to merit disregard.

The extremely detailed wording reflects the confusion around AI and the creative arts in general, which was thrown into bold relief when Paul McCartney revealed last month that AI technology had been used to clean up a rough recording of a late ‘70s John Lennon demo to create a “new” Beatles song.

While alarmists conjured visions of a completely AI-fabricated Lennon verse and voice — similar to the 100% AI-generated voices of Drake and the Weeknd on the song “Heart of My Sleeve,” which was found to be in violation of copyright and removed from streaming services — instead the technology was used to sonically clean up the recording, presumably eliminating hiss and background noise and improving the sound.

McCartney later clarified, “Nothing has been artificially created.”

Mason stressed that the last-minute rule revision was made to clarify the fact that those types of AI technology are fine, but generative AI — whereby an AI program ingests multiple examples of, say, Drake’s songs and then essentially recombines them to create its own Drake song — is not.

“What we intended to say was that material using AI can be submitted,” Mason tells Variety, “but the human portion of the composition, or the performance, is the only portion that can be awarded or considered for a Grammy Award. So if an AI modelling system or app built a track — ‘wrote’ lyrics and a melody — that would not be eligible for a composition award."

"But if a human writes a track and AI is used to voice-model, or create a new voice, or use somebody else’s voice, the performance would not be eligible, but the writing of the track and the lyric or top line would be absolutely eligible for an award," Mason noted.

Central to the wording is the term “di minimus,” a legal term that, according to the Cornell University Law School defines as “something that is very trifling or of little importance. Usually refers to something so small, whether in dollar terms, importance, or severity, that the law will not consider it.” The human contribution to the song must be more than “di minimus” in order to be eligible — which means that even if there’s a fully AI-generated “feature” on a song, it can still be eligible.

In fact, he confirmed that means the “new” Beatles would presumably be eligible.

“The provision states that as long as there’s more than a de minimis amount of human involvement in the portion of the creativity that is being evaluated for nomination, then it will still be considered for a nomination,” Mason says. “So if you had a rap record where there was eight bars of AI rap, but the rest of the song was human rapping and there’s a human chorus, that would still be eligible for a performance award. There would not be an award given to the AI-created piece, of course.

“I think that’s a good comparison when we talked about this Beatles record with Paul McCartney: If three or four Beatles are singing on the record, and one of the voices [has been sonically enhanced by] AI, it’s still a live human performance with a more than di minimis amount by the Beatles, therefore it would be eligible.”

In fact, Mason said he’s spoken with “Ghostwriter,” the anonymous creator of the fake Drake-Weeknd duet.

“I got the chance to speak to ‘Ghostwriter’ and hear his perspective, and it’s very interesting,” The Grammy CEO said. “He’s very forward-looking, he’s very creative. And from my perspective, this has been an exercise for him to try and establish a dialogue and create awareness around the possibilities and what will be some of the potholes. I hate to put statements in his mouth, but my feeling is that he understood exactly what he was doing — he knew this was going to be controversial.”

Earlier this year, the singer-songwriter Grimes said that she had no objection to people using AI-generated versions of her voice as long as they shared half of any resulting song’s royalties with her — she even built an app to help people do it. During the National Music Publishers Association meeting last month, Madonna manager Guy Oseary said that when he spoke with artists who were vehemently opposed to AI about a revenue-sharing AI model, they were much more interested.

“I have not spoken to Grimes,” Mason noted, “but the Grimes model seems to be a viable path forward for artists who want to allow people to use their voice and likeness. The financial model and a royalty structure around what she’s established is something that makes sense in theory — of course, there’s got to be some fine-tuning and some nuanced development to it, but seems like a very reasonable possibility. And you’ve identified the two critical issues: credit and remuneration. And if we can figure those two things out, the approval process is not unlike a normal approval process when you’re trying to utilize an artist’s song, or a picture of an artist or other uses.”

The Grammy CEO noted that he’s had preliminary meetings about AI with the U.S. Copyright Office and has interacted with Senator Majority Leader Chuck Schumer’s staff.

He also added that from the many conversations, he’s had on the subject, both within the Academy and the industry at large, is “a general sense of optimism from the majority of people, both from the tech side and from the [rights-holders’] side."

"We’re approaching it from the standpoint of what the possibilities can be, as opposed to trying to ban it or block it or prevent it, because you can’t. And this is why we have to come together and sort this out.”