Australia | The Copyright Stalemate Deepens As Albanese Centralises AI Policy

The Australian Government’s refusal to introduce a text and data mining (TDM) exemption for AI training (see our previous article here) has become the centrepiece of an escalating standoff between global AI companies seeking to invest billions in Australian infrastructure and creative industries determined to preserve the integrity of the Copyright Act 1968 (Cth) (Copyright Act).

This week, the landscape shifted dramatically, as Prime Minister Anthony Albanese announced that his Parliamentary Office will take direct control of Australia’s AI agenda. While it seems clear that Australia is unlikely to adopt an AI focussed exception to copyright protection any time soon, a number of other options to facilitate AI training in Australia remain on the table.

Anthropic’s $15 billion gambit

In recent days, new reports have publicised a Treasury briefing note indicating that AI giant Anthropic told the federal government that its proposed US$15 billion investment in Australian AI infrastructure is “contingent on clarity of copyright settings”. The investment would make Australia Anthropic’s second-largest base for AI model training behind the United States, with plans to secure 1.4 gigawatts of local data centre capacity. OpenAI has likewise been pursuing investment in Australian infrastructure, but has indicated it faces the same copyright constraints that prevent it from using local facilities for model training purposes.

The Treasury briefing note elaborated on Anthropic’s motivation, noting: “This is driven by a desire to have certainty over their liability to rights holders. While Anthropic can negotiate with larger rights holders, they purport there is a ‘long tail’ of smaller rights holders which impedes efforts to identify and purchase licensing rights.”

Reports have referenced a proposed “creatives fund”, whereby AI companies would pay into a government-managed fund in exchange for training data access.

Creative organisations swiftly condemned the proposal. Copyright Agency chief executive Josephine Johnston described it as fundamentally incompatible with the copyright framework: “dismantling copyright and replacing it with a fund removes the economic framework for how creators are rewarded for their intellectual endeavours and incentivised to create new work. It would put their rights at the whim of future Governments’ priorities and the tech sector largesse”. Others have argued that licensing deals with AI developers are already proving workable across news, music and publishing.

Reform options under consideration

While the TDM exemption appears to remain off the table, the Attorney-General’s Department is consulting on alternative models for compensating rights holders. The current consultation has its origins in the Productivity Commission’s August 2025 interim report, which floated a fair dealing exception for text and data mining as a means of reducing friction for AI developers. Attorney-General Michelle Rowland ruled that out in October 2025, confirming the government had no intention of weakening copyright protection for AI. In its place, the Copyright and Artificial Reference Group was convened to consult on major priorities, including a paid collective licensing framework. Against that backdrop, three broad options are now reportedly being canvassed:

1. Statutory licensing scheme — A compulsory licence under the Copyright Act providing AI developers with a clear legal avenue to access copyright material in return for paying a fair licence fee.
2. Collective licensing model — Administered through bodies such as the Copyright Agency, allowing one or more organisations to negotiate and distribute payments for groups of rights holders.
3. Voluntary, negotiated licensing framework — The existing approach, whereby AI companies negotiate directly with rights holders.

In addition, the government is reportedly considering a new enforcement regime to make it easier for smaller rights holders to pursue technology companies for uncompensated use of their work.

None of the reform options on the table would replicate the exceptions available to AI developers in the United States; and key Asian markets, including Japan, Singapore and China, operate some of the most permissive TDM regimes in the world. Rather the Australian options all require some form of licence and payment, with varying degrees of compulsion. This makes Australia a materially more complex and costly jurisdiction for AI model training by international standards, and that is unlikely to change in the near term.

This raises a deeper question for policymakers: whether such a regime will deter investment in domestic AI infrastructure and push model training through more permissive offshore jurisdictions, eroding Australia’s ambitions for national sovereignty in AI. There is also a potential risk that the largest developers simply treat any infringement exposure as a cost of doing business, particularly while enforcement options remain uncertain. Whether the copyright framework acts as a genuine deterrent or merely adds friction for compliant operators will depend heavily on the design of whatever licensing and enforcement regime ultimately emerges.

Albanese’s AI policy u-turn: The Office of AI

In a significant policy pivot Prime Minister Albanese announced on 15 July 2026 the establishment of an Office of AI within the Department of the Prime Minister and Cabinet. This is not the first time Australia has pivoted on AI regulation: an earlier policy framework included mandatory guardrails, before the December 2025 National AI Plan reversed course, adopting a “technology-neutral” approach with no standalone AI Act and no mandatory guardrails. The latest announcement appears to reverse that position once again, returning to a more interventionist regulatory posture.

The new Office will seek to work with other Ministries to coordinate Australian standards for AI, and bring together work already underway on data centres, defence and copyright protections, a shift Albanese framed by comparing AI to civil aviation in the 1920s and genetics in the 1990s. Albanese framed Australia as the “first country in the world to bring these issues into a single, national framework”.

In contrast to earlier reports that Albanese’s speech would say little on copyright, the Prime Minister made a notably strong public statement on behalf of creators. Musicians, writers and publishers have urged the government to resist pressure from AI companies and protect their work and Albanese appeared to answer that call. He declared that Australian creative content was not “up for grabs”, and that the new laws will provide “the strongest possible protection for Australian artists and Australian media”. He went further, stating: “no company should use Australian books, music, art or news to build or train AI without the artist’s control. That includes the artist’s control of the price and value of their work. Anything less is theft.”

Implications for practitioners

For copyright, media and technology practitioners, for now:

• Continued permission-based regime: Australia will not adopt a US-style fair use or EU-style opt-out TDM exception. Any AI training using Australian content currently requires licensing and payment. The level of compulsion to license content is still under consideration.
• Timing uncertainty: Assuming some legislative change is likely, the Attorney-General’s Department has yet to commit to a legislative timetable. The Consultative Committee is only now reconvening to discuss submissions from its December consultation.
• Enforcement evolution: A new enforcement mechanism for smaller rights holders is under consideration, potentially lowering the barriers to pursuing AI companies for unlicensed use.

However the underlying policy conflicts between investment attraction and creator protection remain unresolved, and the government’s new regulatory framework and approach to copyright will be closely watched by rights holders and AI developers alike.