[co-author: Joshua Cooley]
In response to a request to identify ‘priority reforms’ under the Australian Government’s productivity growth agenda, on 5 August 2025 the Productivity Commission released an interim report entitled Harnessing data and digital technology. The report calls for consultation on reforms to ensure Australia’s copyright regime remains up to date with advances and widespread adoption of AI technology. Among other policy options, the Productivity Commission proposes the adoption of a text & data mining exception to copyright infringement.
Background
In December 2024, the Treasurer, Jim Chalmers, requested the Productivity Commission “undertake five inquiries to identify priority reforms under each of the five pillars of the Government’s productivity growth agenda and formulate actionable recommendations to assist governments to make meaningful and measurable productivity-enhancing reforms.”1
On 5 August 2025 the Productivity Commission released an interim report: ‘Harnessing data and digital technology’ (the Report). In the Report, the Productivity Commission indicated AI could facilitate an approximately $116 billion increase in GDP over the next decade through improving multifactor productivity. However, the Report also states that “experience shows that we need a flexible approach to facilitating data access across the economy”. In this context, the Productivity Commission singled out Australia’s copyright regime as an area where the Australian Government “can act to bring regulatory clarity”, and called for feedback about whether reforms are needed “to better facilitate the use of copyright materials, in the context of training AI models”.
One of the policy options the Productivity Commission put forward for consultation is a proposal to amend the Copyright Act 1968 (Cth) to include either a ‘fair dealing’ or ‘fair use’ exception for text and data mining (the TDM exception).
Since the publication of the Report, the Treasurer has stated the Government doesn't “have any plans to change or weaken [Australia’s] copyright arrangements.”2
What is a TDM exception?
As we have observed previously, central to almost any AI system is a large mass of data on which the system is trained. These training materials may include original works in which copyright subsists.
The TDM exception proposed by the Productivity Commission would enable the use of copyright protected content to train AI models in Australia, and more generally for uses of copyright protected content in “all forms of analytical techniques that use machine-read material to identify patterns, trends and other useful information”, provided the use of the content is ‘fair’ in the circumstances. The Productivity Commission further identified that implementing the exception may create “a need for legislative criteria or regulatory guidance about what types of uses are likely to be considered fair”.
Various different forms of TDM exceptions currently exist in other jurisdictions, including:
- In the European Union, TDM exceptions exist for scientific research and broader general use, with UK courts finding the exception does cover AI language training.
- In the United Kingdom, the TDM exception applies only to non-commercial research. However, as we have discussed previously, there has been a recent push to expand the exception to apply more broadly.
- In the United States, courts have found that while AI training still needs to meet the ‘fair use’ criteria for the doctrine to apply, the exception is available.
- In Singapore, there is both a specific TDM exception within the Singaporean Copyright Act and a broader fair use exception.
The arguments for and against a TDM exception
The Productivity Commission’s request for submissions about the likely effects of introducing a TDM exception to the AI market in Australia highlights the following considerations:
- Large AI models (including generative AI and large language models) are generally available to be used in Australia and this is unlikely to be affected by the implementation (or not) of a TDM exception in Australia.
- At present, large AI models are trained overseas, not in Australia. It is unclear whether the introduction of a TDM exception would change this trend.
- Large AI models are already being trained on unlicensed copyrighted materials.
- A TDM exception could make a difference to whether smaller, low compute models (such as task-specific models) can be built and trained in Australia, such as by Australian research institutions, medical technology firms, and research service providers.
In his National Press Club address in late July, Atlassian co-founder Scott Farquhar also called for the introduction of a TDM exception, as one of the actions which could be taken by the Government to remove barriers to AI companies training or hosting their models in Australia.3
The Report did note several consultation participants’ concerns about the unauthorised use of copyrighted materials to train AI models. The Productivity Commission emphasised that a TDM exception would not be a ‘blank cheque’ for copyrighted materials to be used as inputs into all AI models because the requirement that use be ‘fair’ would limit the scope of the exception.
Creators and copyright owners have also raised concerns about TDM exceptions, with the Australasian Performing Right Association (APRA), the Australasian Mechanical Copyright Owners Society (AMCOS) and the National Aboriginal and Torres Strait Islander Music Office (NATSIMO) rejecting the Report’s proposals, claiming a TDM exception would “legitimise digital piracy under guise of productivity”.4
NATSIMO highlighted results of a 2024 survey in response to the Productivity Commission’s interim report, which suggested that Aboriginal and Torres Strait Islander communities fear that the use of AI increases risks of cultural appropriation and could stymie efforts to codify safeguards for Indigenous Cultural Intellectual Property (ICIP). The survey found5:
- 89% of Aboriginal and Torres Strait Islanders surveyed think AI has the potential to cause cultural appropriation;
- 67% agree that using AI in music creation makes it harder to protect and enforce cultural rights; and
- 89% think Traditional Guardians or Owners of ICIP should be able to handle copyright violations resulting from AI activity.
Critics have also pointed to the Productivity Commission’s statement that “large AI models are already being trained on unlicensed copyright materials” and expressed concern that, rather than supporting creators who have been affected by this practice, the Commission is instead seeking to legitimise it.6
Other rights holders have questioned the premise of the Productivity Commission’s report; for example, Newscorp Australasia chief executive Michael Miller has stated there is “no genuine evidence that Australia’s copyright laws are stifling innovation or investment”.7
Other options
In addition to a TDM exception, the Productivity Commission identified two other pathways to address concerns raised about the unauthorised use of copyright works:
- utilising copyright licensing regimes to facilitate the authorised use of copyright materials prior to the use of such material to train AI models, for example, through collecting societies who can negotiate licences on behalf of multiple copyright holders at once.
- Copyright Agency Limited submitted that licensing mechanisms are already available, for example its licenses which permit the use of news media content in prompts for AI tools. Copyright Agency Limited is also conducting discussions with its members and licensees about further solutions for collective licensing, including the use of datasets for AI-related activities.
- Addressing any unauthorised use of copyright materials after the event, including through take-down notices, alternative dispute resolution and court action.
- This has been previously explored in the 2022-23 Copyright Enforcement Review by the Attorney-General’s Department, which found that additional regulatory measures are needed to achieve an effective copyright enforcement regime.
What is next?
The Commission is seeking feedback on:
- how a TDM exception would affect the development and use of AI in Australia;
- the need for legislative criteria or regulatory guidance about what uses are considered ‘fair’; and
- whether the TDM exception should be implemented through a broad text and data mining exception or one that covers non-commercial uses only.
Written submissions can be submitted here by 15 September 2025.
[View source.]
