When we listen to a song, we do not hear a list of the tools used to create it. We hear a voice, a melody, the movement of the music. Something resonates with us — or it does not. But when the conversation turns to artificial intelligence, attention often shifts from the work itself to one short label: AI.
I want to understand how much information remains outside that label. Who wrote the music? Who found the words? Who played the parts? What decisions were made in the studio? What exactly was entrusted to the software?
The EUIPO Academy webinar included an example very close to these questions: a musician writes her own lyrics and melody, then uses AI to create the accompaniment. The presentation examines this case through the different elements of the result. For me, this is where a meaningful conversation begins.
One song, several stories
The webinar materials distinguish between the musical work, its sound recording and the performance. Alongside them is the artist’s recognisable identity: their name, artwork and visual identity. These elements can involve different rights and different people.
The writer of the melody and lyrics, the performer and the owner of the master recording have different roles, even if one person sometimes fulfils them all. A musician’s name in the credits does not, by itself, explain who owns the rights to the recording or how the income is shared. Those questions need to be clarified separately.
So the question “Who owns the song?” needs a follow-up: which element are we talking about?
This is particularly relevant to my work. There can be many stages between writing a song and arriving at its final sound. A new version may retain the music and lyrics while changing the arrangement, introducing new parts and using different forms of processing. I want the release information to help people understand what happened at each of those stages.
What remains a human author’s contribution
The presentation identifies three situations: elements created entirely by humans; elements created entirely by AI; and elements created through the involvement of both a person and technology.
Human-created elements are considered protected where there is original creative expression. Under the European approach presented at the webinar, entirely generated elements do not receive copyright protection in their own right. For mixed results, the presentation points to the person’s free and creative choices.
The practical conclusion I draw is this: using AI at one stage of the work does not, by itself, describe the origin of every other element of the song. If a person wrote the melody and lyrics, that contribution needs to be considered and described. At the same time, having written your own lyrics does not answer every question about the generated accompaniment.
Nor can creative contribution be determined simply by counting hours, revisions or prompts. What matters is the decisions themselves and how they are expressed in the result. Not every technical operation amounts to authorship, even if it required considerable effort.
I think we need to preserve that precision in both directions: speak specifically about the person’s work and explain the role of technology with the same care.
The history of the work is worth keeping
One recommendation from the webinar is to keep records of the creation process. For a musician, these are tangible things: an early recording of a melody, versions of the lyrics, recorded instrumental parts, project tracks, intermediate versions and correspondence with collaborators.
We can also keep information about what material was supplied to an AI tool, for what purpose, and what the person subsequently changed. This is my practical recommendation, building on the general advice in the presentation.
These materials help reconstruct the sequence of the work. They do not guarantee recognition of rights in every result, but they allow us to discuss specific actions rather than general statements about “a song made with AI”.
For me, there is a human meaning to this too. A project preserves traces of working together: the first phrase, a sound we discovered, a part that changed the entire arrangement. I do not want that history to disappear once the final file has been exported.
Credits should tell the truth
The presentation devotes considerable attention to metadata: writers’ names, performers, recording information, identifiers and the allocation of shares. A separate slide addresses the risk of losing accurate information when ownership of a catalogue changes.
This raises my own question about release submission forms: do they have enough room to describe modern studio work accurately?
If a recording combines a human performance with AI processing, I want to be able to identify both the performer and the nature of that processing. If a new part is produced using a model based on a musician’s material, its origin needs a separate explanation. Automatically equating that part with a new performance personally given by that musician would be inaccurate.
This is a question I am raising myself. The presentation does not prescribe a mandatory credit format for such cases, nor does it establish that any particular wording resolves all the legal questions.
Authorship and lawful use are separate questions
Keeping records of your own contribution is necessary. But another question remains: does the result use someone else’s protected elements without the necessary legal basis?
The webinar refers to GEMA’s case against Suno. The Munich I Regional Court’s judgment of 31 July 2026 concerns specific works and ways in which they were used. An official European Commission overview separately examines model training, reproduction of works within model parameters, and recognisable elements in generated outputs. At the time that overview was published, the judgment was not yet final.
For me, this is another reason to keep these questions separate. The existence of a human author’s contribution and the lawfulness of using someone else’s material require independent consideration. One does not replace the other.
After the webinar, I was left wanting to speak more precisely about my own work. To name the people, describe their contributions, explain how technology was used and preserve the history of creation. So that behind the finished sound, it is possible to see how it came into being.
Alex Kryve
Materials and scope of this article
This article combines a summary of selected points from an educational presentation with the author’s personal reflections. It is not an official EUIPO interpretation or individual legal advice. Its legal context is primarily the European Union; a particular situation may require checking the applicable national law.
- EUIPO Academy. “Music & IP: Protect Your Creativity and Shape Your Future”, 14 October 2026. Saved presentation: slides 9–14, 17–24, 33–36. Webinar page. Availability of the page may vary.
- European Commission, IP Helpdesk. German court finds Suno liable for copyright infringement over AI-generated music, 28 August 2026. The status of the judgment stated above is as reported in that overview.
- WIPO Lex. GEMA v Suno, Case No. 42 O 763/25, judgment of 31 July 2026; court of first instance.