The voice, one of the most intimate components of identity, constitutes both an immediately recognizable distinctive feature and a genuine intangible asset of significant economic value. It also represents a particularly vulnerable resource, insofar as it is constantly recorded, digitized, and disseminated through online content and streaming platforms, thereby becoming increasingly susceptible to reproduction thanks to recent advances in artificial intelligence systems.
The rapid evolution of artificial intelligence and voice cloning technologies now makes it possible to generate highly realistic synthetic voices with remarkable ease from just a few seconds of recording, reproducing timbre, rhythm, and intonation. This phenomenon has contributed to the proliferation of AI covers on platforms such as YouTube and TikTok, where synthetic voices of well-known artists, voice actors, speakers, podcasters, and digital creators reinterpret songs, speeches, public interventions, or non-existent vocal content, and are even used in unauthorized promotional materials.
This phenomenon, which has affected numerous artists, including deceased individuals, has generated significant legal uncertainty and produced substantial impacts across the music, social media, dubbing, and misinformation sectors, raising both legal and ethical issues linked to the use of a technology that is creative in nature but frequently non-consensual.
In this context, a form of protection of personal identity based on sound marks is gradually emerging. Traditionally associated with advertising jingles or short musical sequences and primarily used for promotional purposes as a tool of commercial identification, the sound mark enables an immediate auditory association between a company and a specific product or service.
The protection of personal identity is based on a composite normative framework, structured around several legal foundations. It notably includes the right to one’s image, extended to its sound dimension, the neighboring rights of performing artists, the protection of personality, as well as the general clause of civil liability.
Moreover, recent developments in several European national legal systems, as well as in the European Union’s regulatory framework on artificial intelligence, have significantly contributed to reshaping this legal framework, intervening in copyright law, text and data mining, and criminal law provisions. In particular, in various national legal orders, mechanisms now allow right holders to expressly opt out their works from any use for the training of artificial intelligence systems, thereby strengthening their control over algorithmic exploitations of content.
From a jurisprudential perspective, certain national courts, particularly in Europe, have held that ownership of phonographic masters does not, in itself, confer authorship of the underlying work, thereby correspondingly excluding the possibility that mere physical possession of recordings may constitute a sufficient legal basis for the exploitation of a voice for the purposes of training artificial intelligence systems. The Court of Justice of the European Union (judgment of 4 December 2025, joined cases Mio and Konektra), for its part, adopted an expansive approach to the concept of originality, holding that copyright protection may extend to creations exhibiting a limited degree of artistic input, provided that they reflect an intellectual expression attributable to their author.
Although sound marks are fully recognized under EU law, their function remains strictly limited to their distinctive purpose in the marketplace and cannot extend to the broader protection of personal identity. Indeed, they protect the registered sound sign only insofar as it serves to distinguish the goods or services of one undertaking from those of its competitors, thereby fulfilling an exclusively identifying and commercial function, to the exclusion of any protection of the entirety of the individual’s vocal identity.
Nevertheless, considering the proliferation of artificial intelligence technologies, both legal responses and strategic protection mechanisms are progressively being developed to mitigate the risks posed to artists and other stakeholders.
In the United States, various initiatives aimed at the registration of personal marks — undertaken notably by public figures such as Matthew McConaughey and Taylor Swift — form part of a defensive strategy designed to prevent the unauthorized use of identity through artificial intelligence systems. These signs encompass vocal, visual, and behavioral elements intrinsically linked to the individual concerned, with the aim of creating a genuine legal barrier against improper exploitations, particularly those carried out by generative AI systems.
In Europe, similar dynamics have recently emerged following the filing of sound mark applications with the EUIPO by artists such as Giusy Ferreri and Luca Ward, consisting of short and easily recognizable vocal sequences. However, these filings do not seek to protect specific artistic performances or interpretations—already covered by copyright and neighbouring rights—but rather aim to establish the voice per se as a distinctive sign, within a preventive framework designed to guard against potential uses by artificial intelligence systems.
In this context, sound marks tend to establish themselves as a complementary instrument to traditional legal mechanisms. As registered distinctive signs, they constitute intangible assets capable of economic exploitation through licences, assignments, and contractual arrangements, while their registration entails a deterrent effect, insofar as it enables immediate identification of both the right holder and the precise scope of the claimed right. From a contentious perspective, they also open access to trademark-specific remedies, thereby providing a more direct and structured form of protection than that derived solely from personality rights. Furthermore, they contribute to strengthening the control over the use of voice in artificial intelligence training processes, including with regard to Directive (EU) 2019/790 on copyright and related rights in the Digital Single Market and its provisions on text and data mining, which permit such uses in principle, subject to the express opt-out right of the rights holder.
However, a sound mark cannot be regarded as an autonomous and self-sufficient instrument of legal protection; it must necessarily be embedded within a multi-layered protection strategy, cumulatively combining the reservation of online content, the inclusion of specific contractual provisions, as well as recourse to personality rights and the rules of personal data protection law.
Indeed, personality rights have traditionally ensured the protection of voice and image according to an essentially ex post logic, contingent upon the cumulative demonstration of infringement, the identifiability of the individual, and the existence of compensable harm. However, the development of generative artificial intelligence — and in particular voice cloning technologies — is undermining this classical framework, insofar as such systems do not merely reproduce pre-existing content but are capable of generating so-called “similar” voices, thereby rendering the boundary between lawful imitation and genuine appropriation of vocal identity particularly uncertain. In addition, there is a structural difficulty in attributing liability, stemming from the plurality of actors involved in the value chain of generative systems, including developers, intermediary platforms, and end users. Copyright law likewise reveals its intrinsic limitations, as it is designed to protect specific works and recordings, whereas voice cloning systems are capable of producing original outputs without any direct reproduction, while nevertheless remaining strongly linked to the vocal identity of an identified artist.
Similarly, sound marks face inherent structural limitations, in that they protect only the recorded sound sequence as such, to the exclusion of vocal identity considered in its entirety, remain subject to the principle of specialty, and are primarily embedded within a framework of commercial use. Furthermore, artificial intelligence systems may reproduce the distinctive vocal characteristics of an individual without reproducing the registered sign itself, thereby potentially escaping qualification as trademark infringement under trademark law.
Accordingly, none of the traditional legal instruments, when considered in isolation, provide fully adequate protection. The protection of the voice in the age of artificial intelligence therefore requires the implementation of an integrated, multi-layered approach, based on a systemic articulation between trademark law, personality rights, personal data protection law, and the regulatory instruments applicable to artificial intelligence.
The emergence of voice cloning based on artificial intelligence has, moreover, in several European countries, opened a significant gap within contractual frameworks, particularly in the field of recording agreements, making it necessary to undertake a thorough revision of publishing contracts as well as synchronization licences. Indeed, a substantial proportion of contractual instruments concluded prior to 2023 in the recording industry do not expressly address the use of artificial intelligence technologies, thereby creating a genuine normative gap likely to generate significant interpretative and enforcement difficulties. In this context, certain particularly broad contractual clauses—especially so-called “all media now known or hereafter devised” provisions—could, under an expansive interpretation, be regarded as capable of encompassing voice cloning technologies as well, notwithstanding the absence of any explicit reference to such technologies.
By contrast, the most recent contracts now include specific provisions expressly dedicated to artificial intelligence-related uses, such as the prohibition on using recordings for the training of AI systems without the prior, express, and separate consent of the rights holder, as well as without independent financial remuneration. They also more explicitly prohibit any digital reproduction or reconstruction of the artist’s voice and timbre. In parallel, there is a growing development of contractual clauses providing for periodic review and updating of contractual terms, as well as strengthened mechanisms for monitoring, traceability, and oversight of the exploitation of masters.
The central element remains consent, which must be simultaneously express, informed, and objectively traceable, as the voice is increasingly regarded as an intrinsic and inseparable component of the artist’s identity. From this perspective, new licensing frameworks tend to regulate exploitation modalities, conditions, and limitations with greater rigor, while also providing specific safeguards against any abusive, deceptive, derogatory, or defamatory use.
In a context where artificial intelligence is now capable of dissociating personal identity from the corporeal support that carries it, the voice can no longer be reduced to a mere vehicle of expression: it is situated within a hybrid legal qualification space, at the intersection of intellectual property law, personality rights, and emerging technologies.
The law is thus confronted with a genuine turning point, marked by the gradual emergence of the sound mark as an additional instrument for the protection of digital identity.
Situated at the intersection of trademark law, copyright, and personality rights, the voice now occupies an intermediate and still evolving legal space, within which the future development of identity law will be shaped—one that will be called upon to determine whether it remains possible to draw a clear distinction between what falls within the sphere of the subject holding vocal identity and what, instead, pertains to the technological capacities for its reproduction and simulation.