VOICE – AI-generated voices: legal and societal perspectives

About the project

The development of AI-generated – or “synthetic” – voices is advancing rapidly. The technology makes it possible to create voices that sound almost identical to real people, including dialects, breathing sounds and emotional expressions. These voices are already being used in games, films, music, audiobooks, smart assistants and chatbots. This raises important questions:
Who owns a voice? Who profits from it? And what should the legislation look like?

There is currently considerable uncertainty surrounding the legal protection of AI-generated voices, which has significant democratic, cultural and symbolic implications. VOICE aims to create an interdisciplinary research environment that deepens society's, culture's and the legal system's understanding of these issues.

Purpose and objectives

The VOICE research environment (2025–2030) has four key objectives:

1. Create an interdisciplinary space where more perspectives on synthetic voices can come together.
2. Develop a legal basis for issues of equality rights, transparency and anonymisation – taking into account democracy, fundamental freedoms and rights, creative innovation and artistic freedom.
3. Contribute to research in law, literary studies/sociology of literature, human-computer interaction, and ethnomusicology.
4. Develop practical tools that strengthen the rights of citizens and artists in relation to their voices.

Research questions and themes

The project's interdisciplinary work focuses on five overarching themes:

1. Authenticity vs. syntheticity
2. Deceased voices vs. living voices
3. Physicality vs. disembodiment
4. Consumer choice and "voice bubbles"
5. Power and bottom-up legal issues

Participants and networks

VOICE consists of a core team of four senior researchers and one doctoral student, as well as a network of 25–35 researchers and other critical friends who participate in workshops and knowledge exchange.

Project members (selection):

• Katja de Vries, project manager (law)
• Karl Berglund, literary studies (audiobook research)
• André Holzapfel, HCI
• Kıvanç Tatar, ethnomusicology/data science
  

Background: Why VOICE is needed

When AI can reproduce the voices of famous (and deceased) people, a number of complex questions arise:

• Should it be permissible to give your audiobook a voice that sounds like Astrid Lindgren, Snoop Dogg, or a deceased relative?
• Who decides on a voice, and can it be ‘owned’?
• How do synthetic voices affect cultural production, equal treatment and power structures?

VOICE addresses these issues through research that combines technology, humanities, social sciences and law.