Generative AI voice mimicry governance : Doubao voice cloning risk analysis
Li, Xiuzhu (2026)
Kandidaatintyö
2026
School of Engineering Science, Laskennallinen tekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2026061672397
https://urn.fi/URN:NBN:fi-fe2026061672397
Tiivistelmä
This thesis explores the governance issues and risks associated with generative AI voice mimicry, taking the voice-cloning function of Doubao as an example. While voice cloning has its advantages in terms of personalisation and convenience, it also poses potential challenges related to privacy, security, and the right to one's own voice.
The study is of mixed methods including document analysis, policy review and semi-structured interview with 8 student users. It examines the technological, informational, economic, social, ethical and legal risks of voice cloning based on Wirtz et al.'s (2019) six categories of AI risk. It also examines current regulation in China and the EU, and platform-level governance.
The results indicate that even though voice mimicry has its merits, it also has great challenges such as privacy leakage, fraud, impersonation, and violation of the right to one's own voice. The thesis suggests an integrated governance system that includes openness and transparency, user consent and authorisation, technological risk control and safeguards, and multi-stakeholder collaboration, based on these results.
The results add to the current debate on deepfake regulation and voice ownership, and can be used to guide regulators, platform developers and users.
The study is of mixed methods including document analysis, policy review and semi-structured interview with 8 student users. It examines the technological, informational, economic, social, ethical and legal risks of voice cloning based on Wirtz et al.'s (2019) six categories of AI risk. It also examines current regulation in China and the EU, and platform-level governance.
The results indicate that even though voice mimicry has its merits, it also has great challenges such as privacy leakage, fraud, impersonation, and violation of the right to one's own voice. The thesis suggests an integrated governance system that includes openness and transparency, user consent and authorisation, technological risk control and safeguards, and multi-stakeholder collaboration, based on these results.
The results add to the current debate on deepfake regulation and voice ownership, and can be used to guide regulators, platform developers and users.