AI tools can now generate melodies, harmonize a tune, transcribe a recording into notation, or produce a full backing track in seconds. For musicians, that raises a question that used to only matter to lawyers: who owns what, and what are you actually allowed to do with the result?
This article is a plain-language overview, not legal advice. Copyright law is complex, varies by country, and the rules around AI specifically are still being written through court cases and regulatory guidance. If you have a real dispute or a commercial project riding on the answer, talk to an actual attorney. What follows is meant to help you understand the landscape so you can ask better questions and make more informed choices.
Can You Copyright Something an AI Made?
The US Copyright Office has been fairly consistent on this point: works generated entirely by a machine, with no meaningful human creative input, are not eligible for copyright protection. The reasoning goes back to a basic requirement in US law that a copyrighted work needs a human author. A prompt alone, even a detailed one, has generally not been treated as enough human authorship to claim the output as a protected creative work.
That does not mean AI-assisted music is automatically unprotectable. The Copyright Office has drawn a distinction between works where AI did essentially everything, and works where a human made real creative choices using AI as one tool among several. If you generate a batch of AI melodies and then select, edit, arrange, rearrange, add lyrics, adjust harmony, or otherwise shape the raw material into a finished piece, that human contribution can be copyrightable, even if the underlying AI-generated fragments are not, on their own.
In practice this means the amount and kind of human involvement matters. A single AI output used unedited is a weaker copyright claim than a track built from AI-generated raw material that a human musician substantially reworked, arranged, and produced. If protecting your rights to a piece matters to you, document your own creative decisions along the way (edits, arrangement choices, chord changes you made, lyrics you wrote) rather than relying on the AI output alone.
Training Data Lawsuits: Unsettled, Not Decided
Separate from the “can you copyright AI output” question is a different one: did the AI company have the right to train its model on copyrighted recordings and compositions in the first place?
In 2024, major record labels filed lawsuits against AI music generation services, arguing that training models on copyrighted sound recordings without permission amounts to infringement. These cases are still working through the courts as of this writing, and it would be inaccurate to say they have been resolved in either direction. Don’t treat headlines claiming a clear winner as settled law. The legal theories involved, including how fair use applies to AI training, are genuinely open questions that will likely take years and multiple rulings to sort out, and outcomes may differ across countries.
What this means practically for musicians: the copyright status of output from any given AI music generator can depend on what data trained it, and that detail is often not fully public. This is one more reason to be cautious about treating AI-generated music that closely resembles an existing artist’s style or a specific existing song as safe to publish or monetize without further thought.
Public Domain Basics: Composition vs. Recording
This distinction trips up a lot of musicians, AI tools or not, so it is worth being precise about it.
The composition (the melody, harmony, and lyrics as written) can enter the public domain based on how long ago the composer died or when it was published, and the rules vary significantly by country. A Beethoven symphony or a Mozart piano piece is public domain as a composition almost everywhere, because the composer has been dead for well over a century.
A specific recording or edition is a separate matter. Even if the underlying composition by Chopin or Bach is public domain, a particular recording of a performance, or a particular modern edition of the sheet music with its own added editorial markings, fingerings, and layout, can carry its own separate copyright held by the performer, label, or publisher. You are usually free to use the notes Beethoven wrote. You are not automatically free to copy a specific pianist’s recording of them, or scan and redistribute a specific publisher’s printed edition.
This matters for AI transcription tools specifically. If you use an audio-to-notation tool on a public domain composition, you generally have a clean path with the notes themselves, but if your source audio was someone else’s copyrighted recording of that composition, the recording itself may still carry rights separate from the underlying music.
Personal Study vs. Publishing: A Practical Line
A lot of the anxiety around AI transcription tools comes from conflating two very different activities.
Transcribing a copyrighted song for your own practice, study, or a private lesson is a low-risk, everyday activity that musicians have done for as long as sheet music has existed, whether by ear or with software help. Making a personal working copy to learn a piece is generally treated very differently from distributing that copy.
Publishing, selling, or publicly redistributing a transcription of someone else’s copyrighted composition is a different matter. The composition itself may still be protected, and a transcription is generally considered a derivative work, which means the original rights holder’s permission usually matters if you plan to sell it, post it publicly, or otherwise distribute it beyond personal use.
The safest lane, especially if you plan to publish, share publicly, or monetize anything, is one of two things: generate genuinely original material rather than reproducing a named copyrighted song, or transcribe your own recordings and compositions rather than someone else’s. Both avoid the question of whether you had the right to use somebody else’s protected work in the first place.
This is part of why NotaGen is built the way it is. Its AI sheet music generator produces original classical-style compositions, guided by period, composer style, or instrumentation, rather than reproducing or transcribing a specific named copyrighted song on request. And its audio-to-sheet-music tool is meant for turning your own recordings, practice sessions, or original compositions into notation, which sidesteps the question of rights over someone else’s recording entirely.
A Few Practical Habits
If you work with AI music tools regularly, a few habits reduce risk without requiring a law degree:
- Keep track of what you generated versus what you personally wrote, arranged, or edited afterward.
- Treat AI transcriptions of other people’s recordings as study material rather than material for republishing, unless you have checked the rights situation.
- Remember that public domain covers the composition, not necessarily the specific recording or edition you started from.
- When in doubt about a specific commercial use, a short consultation with an entertainment or IP attorney is worth the cost.
None of this needs to be intimidating. Most everyday uses (practicing, learning, writing original music with AI as a tool, transcribing your own performances) sit comfortably in low-risk territory. The gray areas mostly show up when you are reproducing someone else’s specific, still-protected work and planning to distribute it.
Frequently Asked Questions
Is music generated entirely by AI copyrighted? Under current US Copyright Office guidance, purely AI-generated output with no meaningful human creative input generally is not eligible for copyright protection, because US law requires human authorship. Human selection, arrangement, editing, and other creative contributions layered on top of AI-generated material can be protectable.
Is it legal to transcribe a copyrighted song with AI for personal use? Transcribing a song for your own study or practice is a common, low-risk activity. Publishing or distributing that transcription publicly is a different situation, since the original composition may still be under copyright, and permission from the rights holder is generally relevant if you plan to share or sell it.
Are classical pieces like Beethoven or Chopin always safe to use? The compositions themselves are typically public domain given how long ago these composers died, but a specific recording or a specific modern printed edition of that music can still carry its own separate copyright. Check the source you are actually using, not just the composer’s name.
