Journal · Field guide
The state of AI in music: a field guide for film & brand work
Almost everything written about AI and music is either a sales pitch or a panic. This is neither. It is a plain map of where the technology actually sits in professional film and brand work in 2026 - what it does well, what it cannot do, what you are allowed to own, and how to decide when to reach for it. It is written by a composer who uses AI in the studio every working day and scores for broadcasters whose contracts now ask exactly where the music came from. Read it as a reference, not a manifesto - the argument behind it lives in what machines can do, and what only a composer should.
Two kinds of AI in music (and why the difference is everything)
The single most useful thing you can learn about this subject is that “AI music” describes two completely different technologies that happen to share a name. Confusing them is the source of nearly every bad decision brands and filmmakers make here.
Assistive AI helps a human make music. It prepares and organizes a session, analyzes the harmonics of a sound so it can be rebuilt as a playable instrument, cleans audio, or automates the technical grunt work that used to sit in front of every idea. Crucially, every musical decision - melody, harmony, arrangement, the final sound - still passes through a person. Assistive AI raises no ownership question, because the authorship is human.
Generative AI produces finished musical material from a text prompt. You describe a track, the model returns one. It does not compose so much as continue: trained on enormous catalogs of existing recordings, it predicts the most statistically plausible music for your request. That design is the root of both its usefulness for disposable audio and its two structural problems - the average problem (its instinct is the middle of the road) and the provenance problem (its ability is distilled from other people's work).
Hold that distinction in mind and the rest of this guide falls into place. Most of the value is in assistive AI. Most of the risk is in generative AI. And most confusion comes from treating them as one thing.
What AI can and can't do: a capability matrix
Here is the landscape in one table - the common jobs in professional music, and how suited each is to AI today. “Assisted” means human-directed tool use; “generative” means prompt-to-track.
| The job | AI-suited? | Why |
|---|---|---|
| Session prep & organization | Yes | Pure friction, zero authorship. This is where assistive AI earns its keep. |
| Sound design & building custom instruments | Yes (assisted) | The machine analyzes or generates raw material; the human still chooses and plays the result. |
| Timbre analysis & recreating a tone | Yes (assisted) | A technical, human-directed task - minutes instead of nights of trial and error. |
| Temp / scratch music for an early edit | Sometimes | Fine as a disposable placeholder to test pacing, never as the final track. |
| Functional background “wallpaper” | Sometimes | The brief is fully specifiable, so a generator can hit it - but ownership caveats apply (see below). |
| Trailer & hero campaign music | No | Needs to be distinctive and owned; the average of a genre cannot sell a moment. |
| Original score written to picture | No | Serves a feeling a director often can't put into words. There is nothing to prompt with. |
| Sonic brand identity / audio logo | No | Its entire value is exclusive, ownable recognition. Unownable audio is not a brand asset. |
| Final commercial master | No | Provenance, rights, and a person answerable for the work are the product being bought. |
The pattern is consistent: AI is strongest at friction and at functional, disposable audio, and weakest exactly where music has to be owned, distinctive, or emotionally precise. As the job moves from technical toward meaningful, the machine moves from lead to layer.
Ownership & rights: the current status
Set the artistry argument aside entirely and a colder one remains. If music is meant to work as an asset, the first question is not “is it good” but “can you own it.” Here is the state of play, in plain terms.
Reference · current as of mid-2026
Copyright of purely AI-generated music
Under current US Copyright Office guidance, a work generated purely by AI without meaningful human authorship cannot be registered for copyright. In practice that means no one owns it exclusively - not you, not your brand, not the platform - so you cannot stop a competitor from using something functionally identical.
Reference · current as of mid-2026
Training data & licensing
Major rights holders have taken the large music generators to court over the recordings used to train them. Some disputes are moving toward licensing deals; the core fair-use questions remain unsettled. Until they settle, the parentage of a generated track - and therefore your downstream risk - can be genuinely uncertain.
Reference · current as of mid-2026
Provenance requirements
The industry is not waiting for the courts. Some broadcasters and markets - including the German and Italian television markets - now ask composers to guarantee in writing how music was made. Provenance has quietly become a contract term, not a nicety.
None of this is legal advice, and the picture is moving; verify specifics for your jurisdiction and use. But the direction of travel is clear: for anything that must be owned, unauthored audio is a liability, not a shortcut.
A decision framework for filmmakers & brands
You do not need to track the litigation to make good calls. You need one honest sort and three questions.
First, sort the use. Is this piece of music disposable or durable? Temp tracks, internal drafts, and throwaway functional beds are disposable - AI is a reasonable fit. Anything an audience will associate with your story or your brand is durable - and durable music needs to be human-made, distinctive, and owned.
Then, before any AI tool touches durable, professional work, ask the three questions I proposed on the AI Filmfest stage in Monaco:
- Where was this model trained? If you don't know what it learned from, you don't know whose work you are borrowing.
- Do I legally own what it produces? If the answer is unclear, so is your client's asset.
- Could I have made something equally good without it? If not, you are not using a tool - you are outsourcing your judgment.
“AI handles the scale. A person still has to handle the meaning - and sign their name to it.”
If a project fails the sort or the questions, the answer is not “never use AI.” It is “keep AI in the assistive layer and keep a human on the decisions.” That is exactly how a modern hybrid orchestral score gets built, and why a sonic identity or a custom score over library music still has to be commissioned, not prompted.
Where this is heading
Two things will keep being true. The tools will get better at the friction - faster session prep, richer sound design, more convincing functional audio - and that is genuinely good news for anyone who makes things. And the rights and provenance framework will keep hardening, because the people paying for music need to own it. The likely equilibrium is not “AI replaces composers” or “AI gets banned,” but a clear line: machines handle the scale, humans hold the meaning and the accountability, and contracts spell out which is which.
The studios and brands that do well in that world will be the ones who learned the distinction early - assistive versus generative, disposable versus durable - and stopped treating “AI music” as a single decision. It never was.
Frequently asked questions
What is the difference between assistive and generative AI in music?
Assistive AI helps a human make music - preparing sessions, analyzing a sound to rebuild it as a playable instrument, speeding up technical work - while every musical decision stays human. Generative AI produces finished musical material from a text prompt. The distinction matters because assistive use raises no ownership questions, while generative output carries copyright and provenance risk.
Is AI-generated music copyrightable?
Under current US Copyright Office guidance, a work generated purely by AI without meaningful human authorship cannot be registered for copyright. In practice nobody owns it exclusively, which is disqualifying for any music meant to work as an ownable brand or film asset.
Can I use AI music commercially or in a film legally?
Platform terms often permit commercial use, but permission is not the same as protection. The output may not be copyrightable, the training data behind some models is still in litigation, and several broadcasters and markets now require written provenance guarantees. Commercial use is possible; commercial safety is a higher bar that depends on the specific tool and use.
When is AI appropriate for professional music, and when is it not?
AI is well suited to friction and to functional, disposable, or placeholder audio - session prep, sound design, temp tracks, internal drafts. It is a poor fit for anything that must be owned, distinctive, or emotionally precise: a film score, a trailer, a sonic brand identity, or a final commercial master.