ABSTRACT
This paper argues that Australia’s moral rights regime can be applied and extended to the unauthorised use of copyright-protected works to train artificial intelligence models. Building on prior work which set out a normative basis to reject the authorisation of such training without consent, the paper examines how the right of integrity of authorship can be repurposed towards these actions. The paper highlights how the existing scheme is limited by restrictions like the requirement for infringing acts to be done in Australia and the exception for reasonable infringing use. In response, the paper recommends the adoption of a dedicated moral right against unauthorised machine learning, with a focus on transparency as to training (or a presumption to that effect), allowing creators to withdraw consent for training, standards for verifying model ‘unlearning’, protection against contractual abuses that could bypass moral rights protections, flexible territoriality, and effective remedies grounded on deterrence.
Yuvaraj, Joshua, I Do Not Consent: Moral Rights and Machine Learning in Australian Copyright Law (October 7, 2025), (2025) Copyright Reporter (forthcoming); University of Auckland Faculty of Law Research Paper Series 2025.
Leave a Reply