Key takeaways
- The Australian Government has announced a proposed a national framework for large-scale data centres, with mandatory requirements expected to cover power, grid connection, energy efficiency and water use.
- The framework is intended to streamline currently fragmented approval pathways across different jurisdictions, although further consultation will shape the final settings.
- A new Office of AI has been established to coordinate the government’s broader AI policy and regulatory work.
- Copyright reform remains unresolved, with the government signalling that creators should continue to retain control over the use and value of their work in AI training.
On 15 July, Prime Minister Anthony Albanese delivered a speech at the University of Sydney titled AI in Australia's interests, in which he foreshadowed a new national framework for “large-scale” data centres, including mandatory rules for operators to be “net-generators” of power, requiring them to put as much power into the grid as they use and meet power and water efficiency requirements (National Framework).
The Prime Minister also announced the establishment of a new Office of AI within the Department of the Prime Minister and Cabinet, effective immediately. The Office will coordinate the design of the National Framework and bring together work already underway across government on AI-related issues. In his speech, Mr Albanese said “AI touches on the work of every minister and department” and that, although the government’s response has so far been “issue-by-issue, sector by sector”, Australia now needs a more coordinated approach.
A sector-by-sector approach supported by a central body is a sensible approach to AI regulation, as was outlined in our original submission to the government in response to the 2023 Safe and Responsible AI Discussion Paper. This is similar to the approach taken in the UK and combines the benefits of sector-specific domain expertise, while enabling central regulatory and government coordination and capability to tackle cross-sectoral issues, ensure consistency of objectives and approach and help ensure that any resulting regulatory reform is harmonised and optimised where possible.
It is currently unclear how the new Office of AI will interact with the existing AI Safety Institute, which has, as part of its mandate, a similar goal to support regulators and agencies in responding to emerging AI risks and harms.
World-first new rules for large-scale data centres
In March of this year, the government released non-binding “expectations” for data centres and AI infrastructure developers. The National Framework is proposed to be based on those expectations and bring them into a single regulatory framework that is “clear, consistent and mandatory”. Under the National Framework, new large-scale data centre operators may be required to underwrite or supply their own power, pay their full share of new grid connectivity so these costs are not passed to consumers or businesses, maximise energy efficiency, minimise water usage, and pay for any additional water infrastructure that is required. However, further industry consultation may shape and refine these concepts, including which data centres are subject to the new rules.
The National Framework is also expected to seek to optimise the location of data centres through consultation with states, territories and local communities.
The Prime Minister said he will seek agreement from state and territory governments on the National Framework at National Cabinet next month, with the aim of introducing legislation early next year.
The Prime Minister described Australia as the first country in the world to bring these issues into a single national framework. Speaking to ABC Radio National earlier on Wednesday, Assistant Minister for Science, Technology and the Digital Economy, Dr Andrew Charlton, described the proposed regime as a “world-first” for a regulated national framework for large AI data centres.
The National Framework is intended to replace the existing piecemeal approach involving different levels of government and provide a streamlined process for verifying compliance. Currently, State and Territory approval pathways for data centres remain fragmented and can vary significantly across jurisdictions. For example, NSW classifies large-scale data centres as a state significant development, while Victoria commonly channels major proposals through ministerial assessment. South Australia has released a dedicated data centre strategy with approval via the essential infrastructure pathway. Referring to the new National Framework, Mr Albanese stated that “Getting this right will enhance our appeal to international investors. By delivering greater clarity and speed for approvals, and a streamlined process for verifying compliance”.
The announcement comes as state and federal Labor governments are also reportedly considering requiring data centre operators to provide a financial benefit to the local communities where they establish operations, largely in response to growing community opposition to rapid data centre development. Dr Charlton told ABC Radio National that the government wants to avoid overseas mistakes where data centres have “damaged local communities”, “sucked up energy” and affected local drinking water sources. The Australian Financial Review has also reported that internal Treasury briefing notes released under FOI reveal that the government is considering incorporating its data centre expectations into the foreign investment review process.
Rights-holders to retain control, but reform model still to be developed
The announcement of the National Framework comes at a time when Anthropic is seeking to make Australia its second-largest training ground behind the US. It has been reported that Anthropic's planned investment of up to USD15 billion ($21 billion) to buy 1.4 gigawatts of capacity from Australian data centres is linked to clarifying Australia’s copyright settings and any licensing pathway for using Australian content to train Claude in Australia.
The Australian reported that internal Treasury briefing notes recorded that Anthropic would tell the Treasurer that “investment in AI model development capability and associated infrastructure, like data centres, is contingent on clarity of copyright settings”. Treasury said this was “driven by a desire to have certainty over their liability to rights holders” and noted Anthropic’s position that a ‘long tail’ of smaller rights holders impeded efforts to identify and purchase licensing rights. Instead, AI firms have proposed to pay into a fund for musicians and artists in exchange for a copyright exemption. However, the proposal has been firmly rejected by rights-holder groups: the Australasian Performing Right Association and Australasian Mechanical Copyright Owners Society (APRA AMCOS) described the “long tail” argument as a “convenient smokescreen”; and the Australian Writers’ Guild described the proposal as “corporate welfare”, adding that “Copyright is clear”. The sound recording industry, through ARIA, and the writers, publishers and artists represented by the Copyright Agency, have also been vocal in opposing moves to circumvent copyright protections for the creative sector.
Mr Albanese’s speech sought to address rights-holders’ concerns about any broad copyright carve-out or a pooled-fund model. Mr Albanese said: “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.” He added that Australian creators "must retain ownership and control of their work. Our laws will spell that out, plain as day”.
The government previously ruled out introducing an EU-style text and data mining (TDM) exception to the Copyright Act 1968 (Cth), and has instead convened the Copyright and AI Reference Group to work with rights holders and AI developers on a copyright model that ensures rights holders are paid if their works are used for AI training, while also making it easier to facilitate licensing deals. The Attorney-General is leading consultations on reforms, with three key models currently under consideration:
- statutory licences (providing rights in return for a set payment)
- a licensing regime through collecting agencies such as the Copyright Agency
- the existing voluntary system.
While Mr Albanese’s comments pointed to the government working on a model that would allow AI training while preserving rights-holder control, they did not identify the model the government will adopt. He stated that “No country has got this right yet. Nowhere do artists or rights holders have sufficient control of their work, when it comes to AI training”.
What to watch next – Labor’s draft national platform
The announcement of new mandatory rules for large-scale data centre operators reflects a more interventionist approach to AI policy since the release of the government’s National AI Plan in December 2025, which saw the government pull back from introducing mandatory guardrails for high-risk AI use cases and focus on existing laws as the foundation to address and mitigate AI-related risks.
It also comes ahead of the 50th ALP National Conference in Adelaide from 23–25 July 2026, where delegates will vote on a substantially amended draft national platform which contains significant new commitments on AI oversight and worker protections. Among other things, the draft national platform is reported to propose a model of "co-design" between businesses and workers on AI implementation, under which employers would be expected to be transparent about AI use and meaningfully consult workers on decisions regarding AI when there is workplace change that affects jobs and work. This is likely to be a contentious proposal.
While the Prime Minister’s speech did not outline any employment-related AI reforms, it comes against a backdrop of emerging debate about the role of workers in AI implementation. The government cited recent Department of Employment and Workplace Relations research indicating that AI has not yet had a significant impact on employment and highlighted Australia's strong labour market as a foundation for AI investment. Business groups have relied on the same findings to argue against introducing mandatory workplace consultation requirements, warning that the proposed "co-design" model could increase implementation complexity. The outcome of that debate will be an area to watch as the government's broader AI agenda develops.
Our AI experts at G+T will continue to monitor the government’s policy developments in the AI space. For organisations active in data centre development, financing or operations, our data centre guide provides a practical overview of the legal, commercial and regulatory issues relevant to building and scaling data centres in Australia.