The Ethical Implications of Generative Audio Models: A Systematic Literature Review

Ethical consideration is severely lagging in generative audio research: While 65% of papers discuss positive impacts, fewer than 10% address potential negative consequences, revealing a significant gap in responsible development practices within the field.

High-quality audio generation creates serious misuse risks: The human-like quality of recent generative audio models has enabled realistic fraud, deepfakes, and voice cloning—concerns that are documented in the limited research examining harms, but largely ignored by the broader research community.

Copyright and intellectual property remain unresolved: Generative audio models trained on existing music and speech raise substantial copyright infringement concerns that researchers are not adequately addressing in their work.

The field is at a critical juncture for establishing ethical norms: This systematic review identifies this moment as crucial for guiding more conscientious research practices before generative audio technology becomes more widespread, suggesting that establishing ethical guidelines now could prevent future harms.

Researchers need to systematically evaluate dual-use potential: The paper's findings suggest that generative audio researchers should routinely consider and document both beneficial applications and potential misuse scenarios as a standard part of their work, rather than treating ethics as an afterthought.