Music Awards Add Rules for Algorithmic Co-Writing

Illustrative / Launch edition: the awards body, people and quotes in this story are fictional.

The fictional Meridian Music Awards have done what many in the audio world expected someone to do sooner or later: they wrote down, in plain sentences, what counts as a human-made song. The new eligibility guidance covers records made with generative tools, and its central idea is simple to state and fiendishly hard to apply. A work must contain a meaningful human creative contribution. Everything else in the document is an attempt to say what "meaningful" looks like in a studio.

What the guidance actually says

The rules arrive in four parts. First, authorship: the credited songwriters and producers must be people, and at least one of them must have shaped the melody, the lyric, or the arrangement in a way that can be described. Second, disclosure: every entry carries a short process note listing which generative tools were used, for what task, and how much of the output was kept, rewritten or discarded. Third, voices: a synthetic vocal that imitates an identifiable real performer is ineligible unless that performer has given documented consent. Fourth, categories: the awards will not create a separate "AI song" trophy this cycle, though the committee says it is watching the question.

The wording is deliberately practical. A producer who used a model to sketch twenty chord progressions and then chose, rewrote and orchestrated one of them is in a different position from a producer who typed a mood description and released the first result. The guidance does not pretend those two sit on a clean line. It asks entrants to explain their choices and leaves judging panels to weigh the explanation.

Where the line gets blurry

Every working musician has a story that tests the rule. Take the fictional producer Delphine Okoro, who builds dance records from stems she records herself, then asks a generative tool to propose alternate drum arrangements. She keeps the ones that surprise her and replays them by hand. Is the arrangement hers? Most panels would say yes, because her taste and her performance decide what survives. Now picture a lyricist, the fictional Tomas Reyes, who feeds a model a verse he is stuck on and receives six alternatives. He takes one line from the third option and rewrites everything around it. The guidance would call that a human contribution, but a stricter reader might ask whether a single borrowed line deserves a footnote.

The committee's answer is that borrowed lines are fine and unexamined ones are not. What matters is whether a person made decisions that a listener would recognise as authorship: what to keep, what to cut, where the song turns, how it sounds in the room.

The vocal question

The voice rule is the sharpest edge in the document. A synthetic vocal that evokes a real singer, even without naming them, is ineligible unless the singer has agreed in writing. That protects performers from imitation, and it also pushes producers toward original synthetic voices or toward working with human vocalists who license their sound on clear terms. Several entrants noted that proving consent is easier than proving the absence of resemblance, so expect paperwork to become part of the release routine.

Reactions from the studio floor

Responses in the fictional producer community were mixed but mostly constructive. Delphine Okoro called the disclosure note "a liner note with teeth": it gives credit to the tools without letting them take it. A mastering engineer, Ingrid Salvatore, welcomed the clarity but worried about honest entrants being penalised while careless ones simply omit details. "A rule you can only enforce against people who tell the truth," she said, "is a rule that rewards the quiet ones."

We are not trying to catch anyone out. We are trying to make the story of a song legible to the people who vote on it. — Marcus Whitlock, fictional chair of the Meridian eligibility panel

Younger songwriters, who often start from generative sketches because they cost nothing, were relieved that the rules do not ban the practice. Some older session players worried that "meaningful" will drift downward over time as tools improve and norms relax. Both concerns are fair, and the committee has promised to review the wording after the first season of entries.

What entrants should do now

If you plan to submit, treat your process note as part of the record. A few habits make life easier:

• Keep dated project files that show how a song evolved, including rejected takes and rewrites.
• Write the disclosure in plain language: tool category, task, and what you changed afterwards.
• Secure written consent for any vocal that could be mistaken for a real performer.
• Credit human collaborators precisely, including the arrangers and editors who shaped generated material.

None of this is legal advice, and the guidance is a competition rulebook rather than a law. Still, rulebooks shape culture. If the Meridian approach holds up, it could become a template other bodies borrow: separate the question of whether tools were used from the question of who made the decisions.

The unresolved question is the category issue. Is a song made largely by a human with generative help the same kind of object as one made entirely by hand? The committee says the evidence of the first season will decide. For now, the awards have picked a modest, defensible position. Use the tools if you like, say so out loud, and make sure a person is the author of the choices that matter. That is a line you can argue with, which is exactly what a good rule should allow.