Five decisions that will define Australia’s AI future

Prime Minister Albanese’s speech on 15 July 2026 marked a decisive shift in Australia’s approach to artificial intelligence (AI), from voluntary guidance to mandatory national regulation. The speech announced the establishment of: 

• Australian Standards for AI – consolidating economic, social, security and environmental considerations into a single national framework that reaches across energy, copyright, workforce, education, and national security. The newly established Office of AI, housed within the Department of the Prime Minister and Cabinet (PM&C), will coordinate the design of the standards in consultation with industry and trading partners. 
   
• Copyright licensing requirements – confirming that AI developers will not be permitted to use Australian creative works for training without the rightsholder’s control, and that no text and data mining exception will be introduced into the Copyright Act 1968 (Cth). 
   
• A national planning and approvals framework for data centres – delivering streamlined approvals in exchange for mandatory sustainability conditions, including net power generation, full grid connection costs, water minimisation, and energy efficiency obligations.  
   
• Energy and project structuring obligations – requiring large-scale data centres to underwrite new power supply and put at least as much energy into the grid as they take out, with implications for Power Purchase Agreement (PPA) structures, co-located generation, and capital cost modelling. 
   
• Workplace protections – tasking the Minister for Employment with engaging employers, workers and unions on AI's role in the workplace, with a focus on supporting and creating good jobs rather than replacing them.

This new approach departs significantly from Australia’s current, non-prescriptive, principles-based approach to regulating AI that relies on key tech-agnostic legislation, sector regulators and voluntary AI guardrails. As things stand, the operative AI policy instruments remain the Voluntary AI Safety Standard and the National AI Centre guidance, neither of which is binding. They also don’t address the infrastructure, environmental or industrial questions the PM proposes to address in the new framework.

Underpinning the framework is the concept of sovereign AI. In the Australian context, sovereign AI refers to Australia's capability to develop, control and operate AI using its own infrastructure, data and workforce, subject only to domestic law. The PM warned that inaction would mean “subcontracting our sovereignty and security to the control of foreign monopolies”. What this means in practice is stark: if Australia fails to invest in and develop the data centres, compute, data governance, foundational models and workforce required to design, develop and power the AI that our defence, healthcare, financial industries, and general public have come to rely on, we will always be dependent on another country's models, energy, compute, and experts. We are also subject to the controls imposed by foreign laws that govern the models on which Australian industries are reliant. This includes laws such as the Export Administration Regulations, which dictate the terms on which critical AI capabilities, including models such as Mythos 5, can be accessed and deployed outside of the United States. Where a foreign AI model is subject to regimes such as these, a foreign government can simply revoke or restrict access to a model on which Australian systems depend, with little or no recourse available. This creates a critical point of failure that grows in severity as our reliance on AI deepens. Done well, sovereignty provides not just security of supply for critical systems, but control over the cultural lens through which AI operates.

The Prime Minister’s vision is ambitious, however, there is a significant gap between the sovereign AI future outlined in the speech and the current legislative landscape. Delivering on that ambition will require the government to resolve five critical decisions in the months ahead, each of which will determine what sovereign AI means in practice for Australian industry.

1. What will the mandatory AI standards require?

The proposed AI Standards will form a single national framework spanning energy, copyright, workforce, education and national security, which the PM described as “[c]lear, consistent and mandatory”. Importantly, the standards are not a reboot of the mandatory guardrails considered by the government in 2024, which narrowly targeted AI safety in high-risk settings. The proposed standards are broader in scope and more ambitious in intent, and sit alongside the existing safe and ethical AI guardrails. Consequently, organisations need to continue to invest in AI governance frameworks. They need to ensure their AI use is safe, ethical and fair, and meets existing technology-agnostic legislative obligations.

The scope of the proposed standards raises significant questions about their legislative form. Will they be prescriptive and risk-tiered, like the European Union’s AI Act, or principles-based and outcomes-focused, as the United Kingdom has pursued? The Prime Minister's statement that “it is not our goal to try and legislate for every possible eventuality or risk” points toward the latter. However, the breadth of the framework suggests that sector-specific rules will be required. 

The Office of AI's placement within the Department of PM&C suggests centralised coordination capable of that responsiveness, but the interaction with existing state and territory AI offices, strategies, and procurement guidelines remains unresolved. Whether national standards will override or sit alongside state and territory frameworks is a critical question for National Cabinet. We have already seen how contested this question is, with the Northern Territory Government referring to federal plans to prevent the development of gas-fuelled power stations as an "overreach". The pushback from state and territory governments in the weeks after the Prime Minister’s speech has demonstrated that intergovernmental alignment on the shape of any national framework cannot be assumed. This will directly affect whether Australia can present a coherent proposition to the investors it needs to build sovereign capability at scale. 

Critically, the mandatory standards will need to regulate risk, incentivise domestic capability and delineate what Australia must own domestically and what can be safely sourced from trusted partners. Getting this balance right will be essential: standards that overreach risk deterring the investment that the framework seeks to attract, while standards that defer too heavily to market participants may fail to deliver the sovereign capability that is at the centre of the national agenda. Decisions made before the legislation is introduced will determine not only compliance obligations but the broader commercial and operational environment for every organisation with exposure to AI in Australia.

2. How will copyright licensing work in practice?

The Prime Minister made it clear that “[n]o company should use Australian books, music, art or news to build or train AI without the artist’s control”, going so far as to characterise “[a]nything less” as “theft”. This position is consistent with earlier statements by the Australian Attorney-General confirming that, as part of forthcoming copyright reforms, the government will not introduce a text and data mining (TDM) exception into the Copyright Act 1968 (Cth). Such an exception would permit AI developers to use the works of Australian creators free of charge and without permission for the purposes of training AI systems.

This places Australia firmly at the rights-protective end of the international spectrum in respect of AI training. Unlike the European Union and the United Kingdom, each of which have adopted or proposed some form of TDM exception, Australia will generally require AI developers to obtain permission from rightsholders before ingesting copyright material for training purposes.

There is, however, a tension at the heart of the government’s position. The Prime Minister warned that Australia must not become “the last link of the digital supply chain” and emphasised the ambition to build sovereign AI capability, rather than being “a data warehouse for AI products made overseas”. Achieving that ambition will require domestic and international developers to access high-quality Australian training data at scale. This in turn demands a copyright licensing regime that is both efficient and enforceable. 

The government’s position that artists retain “control of the price and value of their work” suggests that a voluntary, negotiated licensing framework may be preferred (i.e. where the rightsholder elects to license their work through direct agreements). Alternatively, a collective licensing model administered through bodies such as the Copyright Agency would allow organisations to negotiate and distribute payments on behalf of groups of rightsholders. Another alternative is a statutory licensing scheme requiring the compulsory payment of remuneration in return for the right to use and reproduce copyright material, analogous to existing statutory copyright licensing schemes. A key question is whether any of these approaches can operate at the speed and volume that AI development requires.

3. How will the national planning and approvals framework work for data centres?

The Prime Minister has committed to delivering “greater clarity and speed for approvals, and a streamlined process for verifying compliance” for large-scale data centres. This is a direct response to the fragmented approval pathways that currently exist across Commonwealth, state and local government regimes. The Prime Minister’s foreshadowed new “Australian Standards for AI” acknowledged that overlapping and inconsistent powers across all levels of government and under state-based regulatory regimes are “precisely why we need national standards”. He sought agreement from state and territory governments on the new national framework at the Energy and Climate Change Ministerial Council on 28 July 2026. Ministers agreed to progress the framework, however Queensland and the Northern Territory opposed key elements of the proposed reforms, including the requirement for data centres to underwrite new renewable power supply and the proposed nationally consistent standards.

The framework will be considered by National Cabinet later this month, with the aim of introducing legislation early next year. It would consolidate economic, social, security and environmental considerations into one coordinated structure. Under the new framework, mandatory conditions would attach to approvals, including requirements for data centres to be net power generators, pay full grid connection costs, minimise water use, maximise energy efficiency and fund any additional water infrastructure required. These obligations go well beyond the non-binding expectations for large AI data centres published by the government in March 2026. They represent a significant escalation in the regulatory framework for data centre development.

The commitment to a single, national framework is welcome from an investment certainty perspective. However, it raises significant questions in the environment and planning regulatory landscape. Questions that remain unclear include:

• Will the Commonwealth retain a call-in power over nationally significant proposals?
   
• What transitional arrangements will apply for projects already in the planning system?
   
• How will the framework interact with the concurrent Environment Protection and Biodiversity Conservation (EPBC) Act reforms, including the National Environmental Standards and expanded compliance powers of the National Environmental Protection Agency, which commenced on 1 July 2026?

For proponents, the trade-off is clear: faster approvals in exchange for demonstrable compliance with sustainability and safety standards. Operators would need to model energy self-sufficiency and net grid contribution, which may require co-located generation or surplus renewable capacity. Planning and environmental approvals strategy would now be conceived at a national level, with energy modelling, community engagement and environmental compliance embedded from the outset. The decisions made before legislation is introduced will shape not only individual project timelines, but the broader operating environment for data centre investment in Australia.

4. What does the ‘net power generator’ obligation mean for data centre project structuring and how could it work?

The net energy contribution requirement flagged in the Prime Minister's speech builds on policy signals previously published in the government's expectations of data centres and AI infrastructure developers. 

In particular, the speech anticipates regulatory obligations requiring:

1. “Mandatory net energy contribution: Data centres must contribute at least as much energy into the grid as they take out.
    
2. Full grid connection cost allocation: Data centres are responsible for covering all grid connection costs, ensuring that no expenses are shifted to homes or businesses.”

The objective is to ensure that the rapid growth in AI and data centre infrastructure does not increase electricity prices. Instead, those costs would be borne by data centre developers, investors or customers.

The Prime Minister’s announcement has been followed in quick succession by: 

1. The Australian Energy Market Commission (AEMC) publishing the package of reforms it has recommended to the Energy and Climate Change Ministerial Council on how data centres are to bring their own clean, firmed energy, operate flexibly and connect efficiently so that other electricity consumers are not worse off as data centres connect to the grid;
    
2. Chris Bowen, the Commonwealth Energy Minister, lodging two rule change requests with the AEMC to amend the National Electricity Rules (NER) to ensure data centres cover the grid connection costs they require or cause; and 
    
3. the NSW Government releasing the Electricity Infrastructure Investment Amendment Bill 2026 to give the state minister the power to regulate the connection of data centres to the grid.

The proposed reforms have the potential to reshape the commercial model for large-scale data centre developments in Australia. They are likely to influence project economics, site selection, energy procurement and transaction structuring from the earliest stages of project development. The most immediate implication is the potential for a significant increase in capital expenditure.    

As noted above, data centres are also intended to bear a broader share of the connection costs. This will include those directly associated with funding the connection of the projects to the electricity network. It may also involve those associated with network augmentation, shared infrastructure or other costs that have traditionally been recovered across the network. 

The data centre sector is estimated to have invested approximately $3.1 billion in grid infrastructure between 2020 and 2025, with that figure projected to reach $7.2 billion by 2030. As developments increasingly move to greenfield locations requiring significant transmission and distribution infrastructure, those costs are likely to continue to rise.

Site selection will be critical. Regional locations will offer greater opportunities to integrate generation infrastructure. This may reduce the wider grid augmentation and stabilisation costs that data centres may be exposed to (subject to whether non-renewable behind the meter firming generation is permitted). It will also give rise to commercial and operational challenges that will need to be managed, including land access, fibre connectivity, workforce availability, construction logistics and latency considerations. 

Data centres’ requirement that all of their electricity is sourced from new renewable energy generation on a firmed basis will have significant cost implications, whether the generation is to be co-located with the data centre or otherwise. The proposed framework will change the way energy is procured for large data centres. 

5. What will workplace protections look like?

Ministers have been appointed and forums convened, but a change in approach to workplace protections does not currently feature on the AI legislative agenda. With no evidence of broad labour market upheaval in the most recent AI monitoring report published 9 July 2026 by the Department of Employment and Workplace Relations (DEWR), there is limited impetus to overhaul the existing legislative frameworks. Even if such impetus existed, it is unlikely any major reforms will be pushed through with the same speed (and lack of Senate committee oversight) as the Workplace Relations (Building Cooperative Workplaces) Bill assented to last month.

Currently, government policy is focused on standards and coordination, safety priorities (including AI safety in the workplace) and a clear employment narrative around supporting and creating good jobs, not replacing them. The DEWR’s monitoring framework, together with the tripartite AI Workplace and Employment Forum’s role in progressing ‘AI safety in the workplace’, frames the case for ongoing, data‑driven assessment rather than immediate industrial law change.

Any shift from this incremental trajectory will likely be informed by the DEWR’s foreshadowed ‘gap analysis’, assessing the adequacy of current frameworks. Until then, employers should continue to assess exposure and manage regulatory risk under the current framework of workplace protections, which is already being tested in two key areas: consultation obligations and psychosocial hazards. By way of illustration:

1. Unions are leveraging consultation obligations in awards and enterprise agreements by characterising AI implementation as a major workplace change and seeking to embed AI-related terms through bargaining;
    
2. New South Wales has legislated a duty of care requiring businesses to ensure “digital work systems” (including AI) do not risk worker health and safety. The NSW Government has granted WHS entry permit holders powers to access and inspect relevant digital systems with 48 hours’ notice, signalling closer regulator access and scrutiny of AI‑enabled systems in practice; and
    
3. Safe Work Australia has now published guidance on managing the health and safety risks associated with AI and “digital technologies”.

While legislative reform is not on the current agenda, intensified regulatory focus and scrutiny of AI systems calls for day-to-day compliance with existing industrial and safety laws. Employers should map AI use‑cases, embed explainability and human review for high‑stakes decisions and integrate privacy, discrimination, and health and safety controls into an integrated governance framework. 

Australia’s AI future 

What unites the five decisions examined above is the underlying tension between ambition and implementation. The government’s vision of sovereign AI that is domestically controlled, environmentally sustainable and economically productive is compelling. However, sovereign AI is not a single policy lever; it is the cumulative product of decisions about regulatory design, copyright licensing, planning approvals, energy structuring and workforce protections. 

Translating that vision into a coherent, enforceable legal framework that operates across sectors, jurisdictions, and levels of government will require careful design, sustained consultation, and difficult trade-offs. The practical implications for organisations are immediate. The consultation process preceding legislation will be the critical window in which to engage, influence, and prepare. Organisations should audit their current AI governance frameworks against existing obligations, monitor the development of the national standards, and begin scenario-planning for the compliance requirements that will follow. The decisions made in the coming months by both government and industry will define Australia’s AI future.