by Jose Carlos Erdozain, Of Counsel PONS IP

Phenomena such as Mary of Gold (Finding Me Again has surpassed 4 million streams on Spotify alone) or Breaking Rust (an AI-generated country music project whose track Walk My Walk has accumulated nearly 29 million streams, also on Spotify) have shaken consciences and opened the eyes of those who still doubted the impact of artificial intelligence on intellectual property. I confess that, the first time I listened to songs by these performers, unaware that they were AI-generated projects, I thought: wow, what extraordinary voices! Genuinely moved by the nuances of their vocal performances and the sighs that could almost be perceived through a good pair of headphones, I eagerly sought out more from these “artists”. To my disappointment, however, I quickly realised that there was no past to discover… as though they had never existed. And, indeed, until very recently, they did not exist, at least not in the physical world. Both cases, like an increasing number of others, are nothing more than sophisticated products of artificial intelligence.

Faced with examples such as these, which are surely only the tip of the iceberg of the known uses of AI in artistic creation (according to some commentators, more than 30% of the music we hear daily has been created with, or predominantly by, AI), we must reflect on the role we wish this technology to play within the creative industries.

To begin with, and apologies for revisiting what has by now become an old debate, much has been said about whether AI should be regarded as an author. Considerable attention continues to be devoted to this issue, often by individuals who appear to possess little understanding of the legal foundations of copyright. Yet one need only listen to some of the songs produced by the aforementioned “authors” to appreciate how anomalous and misguided it would be to recognise AI as an author. Beyond the doctrinal problems inherent in treating an entity that is arguably more artificial than intelligent as an author, a role historically reserved for natural persons, practical difficulties would also arise concerning the management of royalties generated by the exploitation of its creations. Should such an entity be registered with a collective management organisation? If so, which one? Into which bank account would the remuneration to which it was entitled be transferred?

Let us acknowledge the reality: AI, or perhaps more accurately the major technology companies and research centres leading this race, has laid siege to the fortress of intellectual property and seeks to conquer it. In truth, this has been underway for several years. Few are willing to predict the consequences. Nevertheless, within academic circles, specialised institutions and even among AI developers themselves, concern is increasingly evident. One need only consider statements such as those made by the CEO of Anthropic, warning of the need to pause AI development. Yet these concerns have done little to slow the race towards ultimate technological supremacy.

The real question, therefore, is not whether AI itself should be considered an author, but rather whether we ought to recognise authorship in favour of natural persons who use AI and claim authorship over the resulting outputs. The answer undoubtedly depends on the degree of reliance on AI. Just as no one would question the authorship of a photographer who uses Photoshop as a creative tool, few would deny that a composer or novelist may still qualify as the author of a work even where AI has assisted in the creative process. Intensity of use is, in all likelihood, the determining factor. It is the criterion that will distinguish what should benefit from exclusive rights from what is merely commonplace and devoid of creativity. Without an author there is no work. Without a work there is no author. The pressure exerted by certain technological sectors appears motivated by a desire to gain a foothold within intellectual property law, perhaps through the recognition of a new form of exclusive right or remuneration attached to intelligent creations. This phenomenon is not unprecedented. The examples of broadcasters, software developers and database producers demonstrate that such efforts can ultimately prove successful.

For its part, case law offers little assistance in restoring clear boundaries around what ought to be protected as a copyright work. From the judgment in Infopaq (Case C-5/08) to the recent decision in Gândul Media (Case C-598/24), and through the landmark Cofemel ruling (Case C-683/17), judicial resistance to lowering the threshold of what may qualify as a work has progressively diminished. The subjective approach has prevailed, placing emphasis on the “expression of the author’s intellectual creation”, an empty formula that increasingly blurs the distinction between the ordinary and the genuinely original. The Court of Justice of the European Union appears to have sacrificed objective creativity on the altar of market pragmatism, favouring protection whenever an author can invoke the manifestation of a creative spirit. Everything becomes protectable, provided it results from human expression, apparently equated with creativity and, consequently, with intellectual property. Have we fully considered the implications? Are we prepared to grant an exclusive right to authorise and prohibit the use of content consisting of merely two lines, one line, or a social media post expressing little more than an idea or opinion?

The debate therefore ceases to concern what constitutes a work and shifts instead towards whether any intellectual activity undertaken by a human being can or should be regarded as a work. The continuing dilution of the originality requirement, encouraged and endorsed by the courts, combined with the increasing use of AI, may well lead us to a scenario in which the mere entry of a basic instruction into a prompt results in the recognition of copyright and the appropriation of the output by the human user. This is not merely a question of categorising works as “hybrid”, fashionable though that term may be. Rather, it represents a genuine erosion of one of the fundamental pillars of intellectual property law: the original work itself, the master key that unlocks all the protections granted by the system.

Examples such as those mentioned above clearly demonstrate that theoretical debate has been overtaken by reality and technological necessity. Whatever position one adopts regarding AI-generated authorship, the potential disruption AI may cause to the foundations of intellectual property law, or the optimal means of balancing the interests of AI developers, creators, rights holders and consumers, the reality is being written by subscriber numbers, views and likes. More than half a million monthly listeners and a steady upward trajectory speak more loudly than academic discussions ever could. In any event, when it comes to country music, I still prefer Ella Langley and her delightful Dandelion. I have yet to encounter any AI capable of replicating that.