Minister rules out copyright exemptions for AI

On 6 August 2025, during an interview on ABC News Breakfast, the Minister for Industry, Mr Tim Ayres, addressed the Federal Government’s position on the regulation of artificial intelligence (AI) in Australia.  At approximately four (4) minutes and fifty-three (53) seconds into the segment, the Minister confirmed that there are “no plans to make changes” to the existing legal framework governing AI platforms and their use of copyrighted works. 

Minister’s statement about copyright and AI

At approximately four (4) minutes and forty-one (41) seconds into the interview, the Minister was asked should “big tech companies get access to Australian copyrighted material to train their AI systems?”.

The Minister’s response was as follows:

[4 minutes and 56 seconds]

“Australian copyright law applies now that’s one of the things that different from Australia to other jurisdictions,”

[5 minutes and 00 seconds]

“There is no ambiguity about this question Australian law [copyright] whether criminal or civil applies now and [uh] we have no plans to make changes in relation to this question,”

[5 minutes and 27 seconds]

“We are as a government focused absolutely on securing the advantage here for Australia, that means , investment in digital infrastructure and data centres, [uh] diffusion of this technology through Australia workplace, so that we maximise the productivity advantage….”

What this means in the context of AI

Training AI models typically involves “scraping” or copying vast quantities of online content, text, images, video, and audio, to feed machine learning algorithms.  Much of this material is protected by copyright.  This has raised concerns from copyright owners who argue that their work is being used without permission in a way that breaches the Copyright Act 1968 (Cth) (Copyright Act).

AI software developers and advocates argue that certain uses, such as text and data mining for training purposes, should be exempt from copyright restrictions under a fair use exception regime.

An example of these conflicting positions was shown in the US class action lawsuit Bartz & ors v Anthropic PBC [2025] No. C 24-05417 WHA ND Cal (Anthropic).  In the case of Anthropic, the plaintiffs were three (3) authors who alleged that Anthropic had used their books without authorisation to train their AI system, ClaudeAI, by initially downloading pirated copies and later purchasing and scanning the books. 

Judge Alsup ruled on 23 June 2025 that the latter use of the lawfully purchased books qualified as fair use, comparing ClaudeAI to a human who learns from copyright material.  The matter will proceed to trial in relation to the earlier use of the pirated copies of the books.

The legal issues raised in Anthropic are not unique, as shown by an article published by The Atlantic on 20 March 2025 which explained that Meta utilised millions of pirated copyrighted works obtained from the website Library Genesis (LibGen), to train it’s AI models.  The article alleges that internal discussions at Meta considered the possibility of licensing books and articles for AI model training, however, decided against this due to the cost and time disadvantages of acquiring the legal rights to every copyrighted work used.  

While no legal action has yet been taken in relation to this use of copyright material, the situation highlights a global challenge for copyright law as AI software developers access and use vast amounts of online content without clear or consistent legal frameworks.

Current Australian copyright law

While Australian Courts have yet to decide any cases involving legal issues related to AI platforms, it’s likely that the outcome would differ to international copyright law as Australia has different statutory exceptions for the use of copyright material.

Although some jurisdictions such as the United Kingdom have adopted specific text and data mining exceptions for certain uses,^[1] Australia has not.

Under the Copyright Act, copyright owners have the exclusive right to reproduce, communicate, and adapt their works.  Any use of copyright works, including for the purpose of training generative AI models, requires permission, unless an existing exception applies.

The Copyright Act does not define a ‘fair dealing’.  Rather, specific fair dealing exceptions exist for the purposes of:

• research or study;^[2]
• criticism or review;^[3]
• parody or satire;^[4]
• reporting news;^[5] and
• a legal practitioner, registered patent attorney or registered trademarks attorney giving professional advice.^[6]

These exceptions are unlikely to apply in a commercial AI training context, since any use must be considered ‘fair‘ to qualify.

Implications of not making amendments to the Copyright Act

The current refusal to introduce copyright exemptions for AI data mining means that AI software developers, researchers, and companies building AI systems in Australia must ensure compliance with existing copyright laws.  This may involve:

• securing appropriate licences for the use of third-party content;
• limiting datasets to non-copyright or openly licensed material; and
• implementing internal compliance processes to mitigate infringement risks.

Although an interim report published August 2025 by the Productivity Commissioner proposed a text and data mining exception to be added to the Copyright Act,^[7] the Minister’s statements suggest that the current Government has no intention of doing so.

Links and further references

Australian Legislation

Copyright Act 1968 (Cth)

Copyright cases by Dundas Lawyers

Native Extracts Pty Ltd v Plant Extracts Pty Ltd (No 2) [2024] FCA 106 – damages for flagrant breach of copyright

Gardner Industries Pty Ltd as trustee for the S M Gardner Family Trust v Telstra Corporation Limited [2021] FCA 294 – site blocking orders under section 115(A) of the Copyright Act 1968 (Cth)

Motorcycle Aftermarket Spares Pty Ltd v Tamworth Cycle Tune Pty Ltd [2014] FCA 1433  – consideration of an assessment of damages arising out of infringements of copyright in photographs and particular trade text

Further information

If you need advice on copyright compliance regarding AI technology, contact us for a confidential and obligation-free discussion:

Disclaimer

This article contains general commentary only.  You should not rely on the commentary as legal advice.  Specific legal advice should be obtained to ascertain how the law applies to your particular circumstances.

---

[1] Copyright, Designs and Patents Act 1988 (UK) section 29A.

[2] Copyright Act 1968 (Cth) sections 40(1), 103C(1).

[3] Copyright Act 1968 (Cth) sections 41, 103A.

[4] Copyright Act 1968 (Cth) sections 41A, 103AA.

[5] Copyright Act 1968 (Cth) sections 42, 103B.

[6] Copyright Act 1968 (Cth) section 43(2).

[7] Productivity Commission, Harnessing Data and Digital Technology, Interim Report (August 2025) 9–24.

Related insights about artificial intelligence law