Australia is currently debating whether Artificial Intelligence (AI) systems should be allowed to learn from copyrighted work without creators’ permission.
A recent Productivity Commission proposal to introduce text and data minging (TDM) exception has sparked backlash from authors, artists, and unions, who argue that AI companies are benefiting from creative labour without paying for it.
Meanwhile, AI tools are increasingly generating creative works – music, writing and visual art – that closely resemble human-made work, raising urgent questions about ownership and authorship.
According to the Australian Copyright Council, only human-made works are protected by copyright, leaving AI-generated creations in a legal grey area – even as they are built from datasets containing copyrighted material.
Australian Copyright law protects works that show “independent intellectual effort” from a human author – AI cannot hold copyright or moral rights.
Making digital copies of copyrighted material to train AI models, even for “data mining”, generally constitutes infringement unless permission is obtained from the copyright owner. While facts and ideas themselves are not protected by copyright, their original expression is.
As the Australian Copyright Council states (2025), “The permission of a copyright owner is generally required to use copyright material to train AI systems, including by way of text and data mining.”
In practice, the consent is not always obtained, and artists’ works are frequently scraped and reused by AI models without consent.
The Media, Entertainment & Arts Alliance’ (MEAA) Stop AI Theft campaign calls for stronger protections to prevent the unauthorised use of creative works in AI training.
The MEAA survey found that 71 percent of members are extremely concerned about the loss of human-led creativity, while 69 percent strongly agree that stronger government intervention is needed to regulate AI.
The campaign calls for an Australian AI Act and a levy on big tech developers to ensure they pay for the creative work used to train AI.
The World Intellectual Property Organisation (WIPO) notes that AI intersects with intellectual property (IP) law at multiple levels, including authorship, ownership, infringement, exceptions and data use.
WIPO distinguishes between “AI-assisted” and “AI-generated” works, noting that varying degrees of human skill, labour, and effort influence originality and ownership across jurisdictions.
It also stresses the growing need for new frameworks to address AI and copyright – such as licensing systems, sui generis rights (special legal rights designed for works that do not fit existing copyright laws), or clearer statutory rules.
Australia’s fair dealing exceptions (for purposes such as research, criticism, parody/satire and news reporting) are narrower than the US fair use system, offering less flexibility for AI training.
Using copyrighted material to train AI could cause market substitution – where AI outputs compete directly with human work – which is now a major factor in assessing copyright risk.
The Senate Select Committee on Adopting Artificial Intelligence report (November 2024), recommended that the Australian Government “require the developers of AI products to be transparent about the use of copyrighted works in their training datasets, and that the use of such works is appropriately licensed and paid for.”
The Productivity Commission’s report proposed TDM exception to the Copyright Act, allowing AI systems to use copyrighted works without explicit creator permission under certain conditions. However, this has sparked debate between promoting AI innovation – which could boost the economy by an estimated A$116 billion – and protecting authors’ rights.
As Lucy Howard, CEO of the Australian Society of Authors argues, “Why should we create a situation where billion-dollar tech companies can profit off authors’ work, but not the creators who made the work?”
Currently, Australian copyright law has not yet caught up with the pace of AI development, leaving both creators and policymakers navigating uncertainty.
As the Senate Select Committee on Adopting Artificial Intelligence notes, reforms are needed to ensure transparency around training data and fair licensing arrangements. MEAA argues that without clearer safeguards, AI risks undermining not only income but the cultural value of human-made work. The Australian Society of Authors similarly warns that creators must retain the right to decide how their work is used.
For journalists and media practitioners, the responsibility is to report on AI developments accurately and model ethical use of these tools by crediting sources, avoiding unlicensed content and maintaining transparency about when AI is used in production.
Protecting creative labour is not only a legal issue — it is a matter of cultural integrity and public trust.
