The last few years have seen a proliferation in startups looking to use artificial intelligence (“AI”) to disrupt a number of different areas of the music industry. The emergence of AI capable of writing new musical compositions has attracted particular attention. Much of the debate focusses on the potential impact of this technology on traditional musicians and musicianship.
From a legal perspective, AI music-creation technology raises an interesting – and at present untested – question about ownership of the relevant musical compositions. Under UK law, an “original” composition is protected by copyright. The first owner is the author, defined as the person who creates the composition, with the qualification that if the composition is made by an employee in the course of their employment, the first owner is their employer.
However, this raises the question of who “creates” a composition which is ostensibly written by AI. UK copyright law ascribes authorship in “computer generated works” to a human being. In Express Newspapers plc v Liverpool Daily Post & Echo  the judge suggested that ascribing authorship to a computer was akin to ascribing authorship to a pen. The more recent case of Nova Productions Ltd v Mazooma Games Ltd & Ors  suggested that authorship in a computer-generated work would be ascribed to whichever of the programmer and the end user of the technology met the traditional UK test of labour, skill and judgment.
However, in the case of AI music-creation, because the technology composes the music with minimal human creative input, there is arguably no human author who meets this test. Could AI be capable of legal authorship in these circumstances? In the short term, we expect that the companies developing this technology will look to deal with ownership contractually – firstly by ensuring that contracts with developers contain an express assignment by the developer to the company of all rights that they might acquire in both the technology and any compositions generated by it and secondly by making clear in their terms and conditions with end users how any compositions generated through the user’s use of the technology are owned and licensed between them. However, all of this assumes that one of the developer, company and end user is the first owner in law. As yet, this assumption is untested. If the AI music creation industry continues to grow and it becomes clear that AI generated music holds real commercial value, we think it is inevitable that someone will look to settle this question in the courts.
Finally, given that much AI music creation technically learns to compose by “feeding” on existing compositions, it seems inevitable that this technology will also give rise to claims from existing rightsholders both that the very act of feeding pre-existing compositions to the AI constitutes “copying” for copyright purposes and also that individual new compositions generated by the AI themselves infringe pre-existing works on which the AI has been “fed”.