Voices in self-defense: the legal battle against AI cloning
With the emergence of the major protagonist of this decade, Artificial Intelligence, numerous concerns have arisen in the audiovisual sector. Particularly, the appearance of tools capable of replicating human voices has led to a series of risks for creative professionals and interpreters, such as the cloning of their voices, loss of control over vocal identity, or possible identity impersonations, among others. Ultimately, by feeding AI systems with real recordings, it is possible to generate dialogues and narratives with the voices of people who did not originally interpret them, which raises serious ethical and legal implications.
In light of this situation, PASAVE - the Platform of Associations and Unions of Voice Artists in Spain - mobilized last year to promote the inclusion of contractual limitations on the use of AI with the voices of actors. Its primary objective was to prevent large distributors from being able to use the voice of a dubbing actor or actress in one project, and then replicate it in other projects without their direct participation. In other words, to prevent audiovisual productions that include their cloned voice from being exhibited and exploited without consent or compensation.
In response, a coordinated action was carried out by all the unions and dubbing organizations in Spain in the negotiations of their contractual conditions. Traditionally, as with other creative professionals, the general rule has been the absolute and unlimited transfer of rights in favor of the producer, so that the producer or the third-party financier of the project could monetize the economic investment made by all means possible. As a counterpoint, this left interpreters without control over the subsequent uses of the result of their work.
This process culminated in the drafting of a rights exclusion clause, accepted by multiple key actors in the sector. This clause establishes the following:
“The use of the voice, modulation, tone, gestures, and analogous elements of the speaker or the dubbing actor/actress is neither permitted nor transferred, to be used in training, simulation, or any similar actions in artificial intelligence (AI) programs or projects, robotics, video games, or any methodology that transforms the original interpretation, for purposes other than those detailed in this contract, which is to give voice and interpret one or more characters in a specific production.”
In summary, this clause ensures that the recorded voice of the actor or actress for a project cannot be replicated via AI for purposes other than interpreting a specific character in this production. On the other hand, it seems to leave open the possibility of using AI tools for this purpose, as long as it is within the same project and for the same character.
As a result of this new scenario, several distributors have accepted to incorporate either the “PASAVE” clause or similar contractual limitations that restrict the use of voices through artificial intelligence for purposes other than those agreed upon, and which require obtaining the necessary consent to carry out certain uses. Among the companies that have adopted these clauses are Movistar, Sony Pictures Entertainment, Netflix, AMC Networks, and RTVE.
Along with these contractual limitations, it is necessary to consider current regulation.
The Regulation (EU) 2024/1689, known as the Artificial Intelligence Regulation or AI Law, classifies systems into unacceptable risk levels (prohibited), high (subject to strict requirements), limited (subject to transparency obligations), and minimal or none (unregulated), establishing stricter controls on systems that present greater risks to fundamental rights, health, and safety of citizens.
In the case of deepfakes or ultrasubstitutions—such as happens when the voice convincingly replicates to interpret texts not pronounced by a person through their voice in other interpretations—they would fall under the category of limited risk and, therefore, subject to transparency obligations. Thus, the regulation requires that a clear notice be included indicating that it is generated or manipulated content by AI. However, the regulation itself exempts this obligation when the content is part of a manifestly creative, satirical, artistic, fictional, or similar work, which is the case for films and audiovisual series. In these cases, it is sufficient to adequately inform about the existence of artificially generated content, without interfering with the exhibition or enjoyment of the work.
It is clear that this exception does not imply that a deepfake of any person can be made in audiovisual projects without restrictions, even if there is proper notice about the use of AI. This new regulation must be analyzed together with the Organic Law 1/1982, of May 5, on civil protection of the right to honor, personal and family privacy, and one's own image. Although this regulation was enacted decades ago, it remains the fundamental legal framework to regulate the use of people's images and voices, and it requires the express consent of the affected person for any use of their image and voice, even when used in artistic or technological environments.
In short, the new PASAVE clause and its different variants set a precedent in the sector and could open the door to future similar regulations in other areas. The legal battle against AI cloning has begun and is already yielding its first fruits. And while it is undeniable that exhaustive rights transfers are necessary for producers, platforms, and distributors to monetize their investments and ensure the viability of audiovisual projects, this contractual margin should not become a blank check that allows for the replacement of human labor through cloned voices. The challenge will be to find a balancing point where technological innovation complements professional talent without distorting or eliminating it.
Article written by Clara Collado, senior associate in the Media area of ECIJA Madrid.
