

News
The team has undertaken extensive literature reviews concerning AI and voice cloning across the disciplinary areas and conducted focus groups.
​
Follow this page to find out more about our research, events and preliminary findings!
​
​

Blog #1 Research focus group, 2024
Our first workshop intended to present the premise of our project and set the stage for focus group discussions, allowing our team to conduct ethnographic research to gain insights into stakeholder perspectives on contemporary developments in generative AI.
The workshop began with a short talk introducing the project and our core research questions. This talk included key information on the voice from a linguistic perspective, outlining what the voice is, how it works, and in what ways AI may potentially challenge our core ideas surrounding the voice. This was meant to prompt both a scientific but also more thoughtful and holistic perspective of the voice, which we hoped participants would consider throughout the focus group discussions. We also provided some contemporary context around AI voice cloning from the past couple of years. Following the introduction, the workshop broke up into two groups, each led by one of our project leaders, in which we led open discussions guided by prompt questions meant to explore sociological, linguistic and legal perspectives on AI and AI voice cloning. After this, we regrouped and led a plenary discussion that allowed us to collect a broad scope of perspectives. After a short lunch break, we then hosted two further talks: one from one of our research assistants on the findings of their literature review on the public perception and sociological aspects of AI voice cloning, and the next from an Intellectual Property Officer working for the UK government, who provided an introduction to current laws and gave us an insight into the considerations involved in approaching regulation around generative AI.
Some key findings
In our focus groups, participants acknowledged the duality of AI as a beneficial tool but also a source of concern for many reasons. They reached a consensus that AI is neither good nor bad, but that its impact depends on the intention behind and context of its use. So, in order to approach legal regulations and protections around voice cloning, our findings suggest it is crucial to establish why AI is used in a given situation, and what its intentional purposes are. This idea of intention and context was a reverberating theme throughout the discussions at the workshop, guiding both participant perspectives of the technology and their ideas around approaching regulation.
Voice cloning concerns
Despite varying but broadly ambivalent attitudes toward AI as a whole, voice cloning, in particular, was met with apprehension and primarily negative attitudes (even by those who viewed AI as a whole more positively). They viewed the technology to be prone to misuse in malicious contexts such as deepfakes and its unique risks to certain professions, as well as challenging notions of consent. They expressed discomfort about the idea of having their voices cloned and uncertainties around potential implications, also stating that legal options to combat exploitation in this regard are limited. One participant even shared an experience with an AI-generated scam call, explaining that this heightened their awareness of the risks of AI voice cloning. Such concerns arguably underscore the need for research projects like this one, seeking protections for the voice.
Trust
There were various themes that emerged throughout our discussions, one recurring theme being the question of trust in AI. In general, despite enabling efficiency and various creative opportunities, participants stressed how AI raises alarms about job displacement, the spread of misinformation, and several other ethical issues. Thus, a lack of trust, and scepticism, became evident, with one participant stating:
“Based on the information I have gathered, I know not to trust AI.”
There are arguably several reasons for lower levels of trust, including perceived flaws in technology, personal experiences, or job-specific vulnerabilities, to name a few, all of which underscore the complexities of public perception and the need to conduct further research to explore specific social groups. Intriguingly, however, our data revealed a generational divide in trust in and attitudes towards AI voice cloning, where a younger participant displayed a more light-hearted perspective, emphasising its creative potential and remaining both more trustworthy and less critical of its risks:
“I’ve taken everything to be true and not really questioned it before.”
This seemingly contrasted older participants who mainly focused on trust and criticality, thus demonstrating how trust in AI is also context-specific and shaped by various factors; it is not possible to generalise public trust.
Future regulation
Nonetheless, our groups ultimately viewed AI as an inevitability; essentially, we have to learn to live with it, no matter what. Thus, there is an urgent need for adequate regulation, and participants offered some suggestions as to how to approach this; it must be noted, however, that there remained an uncertainty of how such suggestions may work in practice:
“We need regulation but it won’t come into place until we have a strong footing of how to discern what is AI and what isn’t.”
Nonetheless, in line with key factors established earlier, the group debated the importance of the difference between malicious intent and inadvertent misuse, suggesting that legal frameworks pay specific attention to intent and context in regulating AI. Furthermore, participants encouraged policies to emphasise things like public education to regulate misguided levels of trust in AI, a focus on increasing transparency (one suggestion included the development of mandatory markers for AI-generated audio to enhance transparency, for instance) and looking at past legal precedents.
The insights gained in these focus group discussions provide a foundation for understanding the implications and perspectives of AI voice cloning for certain stakeholders, who ultimately reaffirm the urgent need to regulate and find adequate protections for the voice. Our findings encourage further multidisciplinary efforts to address AI’s social and legal implications and navigate the balance between AI innovation and ethical practice. We hope that through critical, evidence-based approaches, we can find ways to introduce protections for the human voice, whilst upholding its integrity and accommodating the rapid developments within the AI industry. If AI is, in fact, an inevitability, safety in its future relies on responsible development, use, and regulation. We look forward to continuing our research with further studies to look toward providing an evidenced legal, linguistic and sociological perspective of voice cloning.
A full write up of our research is forthcoming.
Background
This workshop focused on AI voice cloning in the creative industries and the legal protection of voice rights. This workshop functioned as a knowledge exchange opportunity and research focus group.
​
Why host a workshop?
AI deep fakes create ‘misuse’ of an individual’s personal image, but little is known about whether this ‘personal image’ includes the voice. Despite recent emphasis on the legal and ethical implications of rights pertaining to image and voice, the legal and ethical implications of AI vocal deep fakes are still relatively under-explored, particularly for those whose lives and livelihoods are directly threatened by AI ‘voice cloning’ technologies. With a focus on the creative industries, we seek to understand the characteristics of the voice and how to protect it by understanding the legal, linguistic and ethical dimensions of ‘personality of the voice’. This extends to tensions and transformations taking place across the creative industries and can apply to music, performance and licensing issues. In this workshop we ask; what are the characteristics of personality in voice, and can they be protected? What are the concerns affecting the creative industries about AI use? Where do the opportunities lie?
Our aim was to engage directly with stakeholders in this area, allowing our research to benefit directly from the experiences of those who are most affected by this issue.
This event was organised by:
Dr Jenn Chubb (Sociology)
Dr James Tompkinson (Language and Linguistic Science)
Dr Peter Harrison (Law School)
This event was funded by Yorvoice, Sparks funding which is drawn from the Research Development Fund funded directly by the University of York.
​
​
​

