The explosion of generative artificial intelligence presents a thicket of copyright challenges, swallowing up creative industries, technology developers, and legal practitioners into a realm of uncertainty. While Australia, like many jurisdictions, relies on existing frameworks not designed for this technological moment, international courts and policymakers are locked in fierce debates over issues ranging from training data use, outputs, and the copyrightability of content generated by artificial intelligence. Three persistent questions underpin legal disputes: does training a generative artificial intelligence model infringe copyright, do its outputs violate exclusive rights, and can those outputs themselves receive copyright protection?
In Australia, despite robust governmental reviews and recommendations—including a call from a Senate committee for transparency and licensing requirements for developers using copyrighted works in artificial intelligence training—no reforms or litigation have yet meaningfully altered the long-standing legal landscape. Australian copyright law’s territorial limitations and narrow exceptions keep rights holders cautious, especially without a broad ´fair use´ exemption. Meanwhile, copyright owners must watch foreign legal developments closely, as offshore decisions may shape Australia’s future approach.
Internationally, the UK’s much-anticipated Getty Images v. Stability AI case saw Getty Images suddenly withdraw its primary copyright infringement claims, leaving only trade mark and secondary claims in play, sending shockwaves among rights holders. The UK government’s consultation has generated policy options ranging from total inaction to mandatory licensing and broadened text and data mining exceptions, with further legislative action likely as artistic communities grow restive.
The United States—undisputed epicenter for copyright and artificial intelligence litigation—witnesses a cascade of lawsuits from journalists, authors, publishers, and artists against major artificial intelligence providers. Judicial opinions to date reveal the complexity of applying ´fair use´ to artificial intelligence training, with some cases emphasizing market harm, transformative use, and the necessity of evidence for infringement or dilution. Notably, rulings have sometimes gone in favor of companies when plaintiffs failed to demonstrate actual damages or focused on technical rather than creative claims. The recent RossAI v. Westlaw decision, finding against an artificial intelligence developer for overstepping fair use by competing directly and threatening an established market, may also influence global legal perspectives.
In the EU, the emerging regulatory approach is defined by the AI Act (Regulation (EU) 2024/1689), blending copyright obligations with a flexible text and data mining exception under the 2019 Copyright in the Digital Single Market Directive. The Act creates new duties for general purpose artificial intelligence providers regarding rightsholder ´opt-outs´ and mandates transparency about training data. Judicial review looms as the Court of Justice of the European Union considers whether artificial intelligence-generated outputs or training practices constitute acts of reproduction or communication to the public and how exceptions apply across jurisdictions.
For businesses, especially in Australia, the imperative is clear: do not assume copyright is irrelevant when developing or using generative artificial intelligence. Organizations should tighten contractual arrangements around intellectual property, maintain vigilant legal governance, monitor rapid policy shifts, and anticipate heightened demands for enforcement and licensing. Above all, responsible artificial intelligence deployment depends on bridging the gap between emerging technological capabilities and the evolving legal framework, ensuring human creativity, oversight, and accountability remain central to innovation.
