How Australia charts its own course on artificial intelligence and copyright will directly affect the country’s future demand for data centre computing capacity.

The connection between the two is deliberate: data centres provide the physical infrastructure for training and deploying AI models, while copyright governs the content that flows through that infrastructure.

The regulatory rules in these two areas will directly determine the commercial viability of AI operations in Australia. For data centre operators, investors and colocation service providers, the copyright regime adopted by the government will be as important a consideration as energy prices, land availability and network connectivity.

On July 15, the Australian federal government released a document introducing what it described as the world’s first unified national AI regulatory framework: a new National AI Office within the Department of the Prime Minister and Cabinet, together with mandatory Australian AI standards. The announcement stated that the government would provide “the highest level of protection for Australian artists and media” and ruled out any text and data mining (TDM) exception under the federal Copyright Act 1968.

Australian Prime Minister Anthony Albanese stated: “The development of Australia’s AI industry must respect the rights of creators.” He also reaffirmed the position consistently maintained by the Australian government over the past year: no company should be permitted to use Australian books, music, artworks or news reports to train artificial intelligence without the explicit consent of right holders and without allowing them to control the price and value of their works.

Why Copyright Matters to Artificial Intelligence

Large language models and generative AI systems rely on vast quantities of human-created content as their core inputs. The quality and cultural characteristics of training data directly determine the quality of AI-generated outputs.

The location where AI training actually takes place determines which copyright regime applies. This affects demand for Australian data centre facilities relative to those in competing jurisdictions.

Collecting copyrighted works to train AI models involves acts of reproduction and adaptation. In the absence of a statutory exception or authorization under a licence, such acts may constitute copyright infringement.

Unlike the United States, Australia does not have a broad fair use regime. Australia’s fair dealing defenses are limited to specified purposes, such as research or study, criticism or review, parody or satire, and reporting news. They are generally considered too narrow to accommodate large-scale commercial AI training.

Australia has also neither followed the European Union nor adopted the text and data mining exception previously proposed by the United Kingdom. The absence of both a general fair use defense and a specific text and data mining exception means that AI developers training large models in Australia face significant legal uncertainty over whether they may lawfully access copyrighted content without the explicit consent of right holders.

There is an inherent tension between attracting economically beneficial investment in AI infrastructure and establishing adequate mechanisms to ensure that Australian copyright holders are remunerated for the use of their works. The government nevertheless considers its copyright regime and its objective of attracting AI investment to be compatible.

Three Potential Licensing Models

With a text and data mining exception having been ruled out, the government is drawing on the work of the Copyright and Artificial Intelligence Reference Group (CAIRG) to examine paid collective licensing and voluntary licensing frameworks, together with opt-in and opt-out mechanisms for right holders. Three types of licensing models are currently under consideration:

Statutory Licensing

Under this model, a compulsory licence would be established pursuant to the Copyright Act. Right holders would not be able to opt out, and the amount of remuneration would be negotiated or determined by the Copyright Tribunal. This model would provide AI users with cost certainty but would conflict with the government’s emphasis on allowing right holders to retain control over their rights.

Collective Management Organization Licensing

Copyright collective management organizations, including the Australasian Performing Right Association (APRA) and the Australasian Mechanical Copyright Owners Society (AMCOS), could grant licences on behalf of their members, accompanied by an opt-out mechanism. Legislative amendments would be required if an extended collective licensing model were adopted. This model would introduce a degree of cost variability for data centre operators.

Voluntary Direct Licensing

Under this model, copyright owners and AI companies would engage in direct bilateral negotiations. To some extent, this approach has already been implemented through content licensing agreements, such as those between Nine and Microsoft and between OpenAI and News Corp Australia. In our view, however, voluntary direct licensing would create the greatest cost uncertainty for AI training in Australia because of the potential challenges involved in identifying right holders and the risk of protracted negotiations.

The Global Position

Australia’s position has both policy and commercial significance. The licensing regime ultimately implemented by the government will be an important factor in determining whether Australia can become a preferred destination for large-scale investment in AI computing capacity.

The government is still considering feedback from participants in CAIRG. Meanwhile, further details of the AI standards framework will be submitted to the National Cabinet for consideration in August, and the relevant legislation is expected to be introduced into Parliament in early 2027.

Australia’s policy will also be influenced by the positions adopted in other countries. The UK Intellectual Property Office has launched a consultation on artificial intelligence and copyright to examine options relating to text and data mining. Earlier this year, the UK government decided not to extend the text and data mining exception under copyright law to the field of AI training.

The UK’s approach differs from that of the European Union in certain respects. EU policymakers are conducting a formal statutory review of Directive (EU) 2019/790 on Copyright in the Digital Single Market, including a licensing-fee proposal advanced by the European Parliament and a parallel regulatory framework being developed by the European Commission for data centre infrastructure.

The next phase of Australia’s AI and copyright policy will focus on translating the government’s commitment to “consent and payment” into a workable licensing framework. The way in which that framework is designed will directly affect future demand for data centre capacity in Australia.