Dilan Thampapillai, University of Wollongong

The ground is shifting beneath the feet of Australian authors and other creatives. A leaked proposal seen by the ABC suggests the government might grant artificial intelligence (AI) companies access to the copyrighted works of Australians under an opt-out system.

There are no firm plans as yet, but it does appear the government wants to incentivise big AI companies such as OpenAI and Anthropic to bring their business to Australia. Deputy Prime Minister Richard Marles stated:

The opportunity for Australia economically is enormous in collaborating with frontier companies.

It seems the government also wants to see big AI companies investing in data centre development in Australia.

What’s in play is a choice between the security offered by having local data centres versus the long-term interests of Australian writers and other creatives.

While details of the plan are not publicly available, the fact an opt-out system is being looked at suggests a backtracking from the Albanese government’s promise there would be no text and data mining exception to copyright for AI companies.

Under such an exception, Australian copyright holders would be compensated for the use of their works in AI training. However, Minister for Industry and Innovation Tim Ayres has denied a reduction in copyright protection.

The perils of opt-out

An opt-out system doesn’t gel with Australia’s current copyright laws. Under the Copyright Act, the owner of the copyright in a piece of writing or photo has the right to exploit those rights as they choose.

Forcing copyright owners to opt out of a given system – such as their works being included in training materials for AI models – impedes their rights, because it forces them to take action to protect those rights. Our copyright law dictates the burden should run the other way: it should be the AI companies asking for permission.

The problem with an opt-out system is that by the time you get around to actually opting out, your materials have likely been inside the neural network for a while. Once materials are in the model, the AI system can’t “unlearn” them.

This is where the heart of this dispute lies.

The problem is not just the taking of individual copyright-protected works. It’s the ever-increasing ability of AI systems to produce quality materials as a result of training on copyright-protected works. The better the data, the better the outputs.

This obviously has implications for Australian writers and creators. There are already studies that suggest readers prefer AI-produced works over those of human writers. Widening the range of materials that can permissibly be used for training likely further hastens the evolution of AI.

It also has implications for less obviously affected parties, such as academics and other educators. Another recent study suggested that law professors preferred AI answers over those written by their human colleagues.

The real problem is replacement

So should creatives be paid for their works used in training under a royalty scheme? That might not be enough because the prospect of being replaced or displaced by AI is the real problem.

AI has already made its way into creative industries and other areas of intellectual endeavour.

Visual artists, graphic designers and illustrators have been reporting how generative AI has encroached on their work, resulting in lower wages and even complete replacement.

A short film with AI-generated actor Tilly Norwood was released last year. And the popular music streaming service Spotify has recently been making an effort to remove millions of “AI slop” tracks from its platform.

In other industries where AI has taken over some of the work, such as the translator industry, there’s been a decline in job satisfaction for human workers. In effect, they’ve been displaced from primary translation work into secondary work, such as checking the accuracy of AI translations.

In law, another language-based discipline, AI is threatening to impact graduate employability. And graduate lawyers are being shifted onto verification tasks to ensure AI-generated content is accurate. This is a downgrading of their work tasks.

A recent government report has noted those occupations at risk of AI replacement experienced slower employment growth than those that were not at risk. Notably, the report listed advertising and marketing professionals as being among those most exposed to AI replacement.

All of the anxieties that exist in other industries are keenly felt by Australian authors and creatives. None of our worst fears has come to pass just yet, but giving AI industries a free pass on Australian copyright is the first step to a future in which there are some winners and some real losers.

That must not happen without a longer and more open consultation process.

Dilan Thampapillai, Interim Pro Vice-Chancellor (Regional & Metro), University of Wollongong

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Alexandra Hansen, The Conversation and Signe Dean, The Conversation

Community opposition to large-scale data centres is growing. Critics are calling for a pause on data centre development, in line with other jurisdictions such as New York, but Prime Minister Anthony Albanese has ruled this out.

We put a call-out to our audience: what questions do you have about data centres? We were quickly overwhelmed with responses and had to close the survey.

This avid response cemented our decision to take a deeper dive into data centres. From almost 350 questions, here were the most common ones.

Click a question to jump to the answer.

1. What data is stored in them, and who owns it?

2. Are they regulated? How?

3. Do communities get a say in where they go?

4. Why do they need to be located near where people live?

5. What’s in it for us? Why does the government want them?

6. Where do the profits go?

7. Can the water be recycled?

8. What happens to them when the AI bubble pops?

9. Are they/will they become military targets?

---

Data centres have existed for decades. But something has changed, and this essential infrastructure is now at the centre of a major policy debate.

This article is part of The Conversation’s series on data centres – what they are, why we need them, and why they’re suddenly so controversial.

---

1. What data is stored in data centres, and who owns it?

A data centre is a bit like a modern library. It holds data in a range of formats and organises it so the data is easily searchable and accessible. Like libraries, data centres have archives and backups, and can hold items until you need them. Like librarians, computer servers inside data centres process requests and send the right data to your device.

To do all this, data centres work with huge amounts of onshore and offshore data. The largest and most complex workloads are handled by “hyperscale” data centres – these facilities are optimised for AI and equipped with specialised chips that can perform millions of computer calculations in parallel.

It’s useful to distinguish between data centre operators and tenants. The operator runs the facility, while the tenant rents space or server racks inside.

Sometimes the operator and tenant are the same company, such as Google. More often though, an operator has multiple tenants and bills each one for the amount of data centre capacity they use. NextDC, CDC and Airtrunk are among the biggest operators in Australia, but they don’t tend to publicly disclose their tenant agreements.

Generally speaking, a data centre is the custodian of the data it hosts. The data remains the property of the individual, or of the service provider that creates or controls the data.

However, data ownership is often complicated by overlapping laws and jurisdictions, as well as the commercial contracts or terms of service between data centre operators, tenants and end users.

Importantly, having data onshore doesn’t guarantee the data can’t be accessed or controlled offshore. This is especially true in cloud computing and AI, where different parts of the workflow can be spread across different servers and locations. A prompt to ChatGPT submitted from Sydney is likely to touch servers well beyond Sydney before returning a response.

Data centres physically located in one jurisdiction must operate under its rules. This means a data centre in Australia abides by Australian privacy laws, environmental regulations, critical infrastructure rules and so on.

Certain types of sensitive data may be subject to strict local requirements. For example, Australians’ My Health records are not allowed to be held, processed or stored overseas.

In other circumstances, data may be subject to overseas access. The US Cloud Act allows US authorities to request American companies to hand over data within their “possession, custody, or control”, even if that data is stored outside the United States.

As data becomes vital currency in an AI-driven economy, questions of data sovereignty – who has genuine control over what – are likely to become even more contested.

Back to top

---

2. Are data centres regulated? How?

In short, they’re not.

Despite increasing community concern about large date centres both here and overseas, there is no consistent national framework for regulating them or their increasing demand on power and water. The only laws that currently apply are the standard planning processes that apply to most infrastructure projects.

Instead, states and territories have varying and largely unenforceable requirements at local levels. None is legally binding.

Of the seven jurisdictions, four have set some policy. Victoria has a data centre action plan and New South Wales has guidelines incentivising developers who meet them with fast-tracked approval.

Tasmania and South Australia both have data centre strategies and inquiries underway to update them.

These strategies all attempt to regulate data centres and to attract them under the right conditions. The strategies all note the pipeline of billions of dollars of investment data centres represent. Most of this investment is occurring in Victoria and New South Wales.

Queensland, Western Australia, and the Northern Territory have no dedicated data centre policy or regulations. In these jurisdictions, data centre planning applications and operations are managed under existing industrial frameworks. This means in many cases, applications can only be assessed as warehouses or general industrial uses.

To address this, the federal government recently announced it will

work with state and territory governments to develop consistent mandatory standards for data centre energy, water and land use […].

These standards are yet to be drafted. The government also hasn’t yet said who will enforce these standards, or what powers and resourcing they will have to ensure compliance.

It has released general guidelines, called “Expectations of data centres and AI infrastructure developers”, but these are not laws. They are guiding principles that reflect emerging areas of focus in the regulation of data centres globally. This includes energy and water use, as well as impacts on natural environments and cultural heritage sites.

Increasingly, the government is also emphasising that the high energy use of data centres should not increase energy costs for other users including households.

The proposed national standards also include a requirement for data centre operators to contribute to the national interest through innovation and local jobs pipelines, themes echoed in the existing state frameworks and policies.

The federal government’s proposals also clearly signal it sees the vast and growing investment in data centres as a lever in the transition to renewable energy. The so-called “causer pays principle” will require data centre operators to (at a minimum) fully offset their energy use with new renewable energy sources.

But both Queensland and the Northern Territory governments reject this idea. The impact will also be limited by the fact it would be a new requirement. This means existing data centres and the more than 200 centres currently going through the planning process will not be subject to these requirements.

So in sum, there are currently very few laws governing data centres, the policies that do exist are non-binding, and future attempts to regulate will require cooperation between the federal, state and territory governments.

Back to top

---

3. Do communities get a say in where data centres go?

The short answer is – no, not really.

Data centre locations are determined predominately by developers and state planning departments, with relatively little influence from local councils and communities.

Developers select and purchase land based on four key requirements:

1. Land large enough to accommodate the facility and future expansion as demand for digital and AI services grows.

2. Energy and water infrastructure capable of powering and cooling the facility, both now and in the future. As AI services become more intensive, developers are seeking access to increasingly large volumes of energy and water.

3. Proximity and connectivity to urban users and/or submarine cable networks, to reduce latency and ensure ease of connection to global fibreoptic networks.

4. A stable regulatory environment in which governments are supportive of data centre investment and expansion.

Sydney and Melbourne are now among the world’s most attractive locations for data centre development because they offer this combination of land, infrastructure, connectivity and regulatory support.

Once a developer identifies a suitable site, it can purchase the land and proceed through the planning process.

State governments play two important roles.

Firstly, they establish the regulatory settings intended to attract investment while managing impacts on communities, infrastructure and the environment.

Secondly, state planning departments assess many of these large-scale developments. In New South Wales, data centres are classified as State Significant Developments. This means applications are assessed by the state government rather than local councils.

Laws in Victoria can similarly allow data centre projects to bypass councils.

This leaves councils with little influence over development decisions. Local planning controls, such as the distinction between light and heavy industrial land, may be overridden by state-level planning processes.

So where does this leave local communities?

Communities generally cannot determine where a data centre is located, but they can have some influence over how it’s designed. Through developer engagement processes and written submissions on exhibited applications, residents may push for changes such as reduced building heights, tree retention or measures to minimise overshadowing.

Community opposition can also affect how a proposed centre is assessed. For instance, in NSW developments receiving more than 50 submissions are referred to the Independent Planning Commission rather than being determined by the Department of Planning. We recently saw an example of this, with the Goodman Group’s Project Mars proposal in Lane Cove West receiving more than 374 submissions, with nine of these supporting the proposal.

What could we do better?

Firstly, data centres could be more appropriately classified and differentiated in the state planning systems, rather than automatically being accommodated within industrial areas that may sit close to homes, schools and transport hubs.

Secondly, local councils and communities could be involved much earlier in decisions about where data centres should be located. Councils have a rich knowledge of their communities, infrastructure, and industrial areas that is often not picked up until late in the application process.

Involving councils and communities early on, rather than only consulting them once a site has been selected and the infrastructure designed, could produce outcomes that benefit both developers and the communities hosting data centres in the long term.

Back to top

---

4. Why do data centres need to be located near where people live?

The short answer is: because proximity is a friend of speed and convenience.

The more complicated answer is: they don’t, at least not all data centres. The world’s biggest data centres, like OpenAI’s Stargate, are not where people are, but where the energy is abundant.

Things are changing though. The rapid take-up of AI services in large cities means more data centres – bigger, more energy intensive – are now being built close to major populations.

The owners of most of the world’s AI compute (processing power) – companies like Microsoft, Amazon, Google and, increasingly, Anthropic – are investing billions to scale up capacity, leading to dramatic new developments in populated cities around the world.

The digital “cloud” has very much descended to earth.

As I show in this data centres map, much of Australia’s data centre pipeline is landing in and around major population centres, and is being fast-tracked by state governments in Victoria and New South Wales. Victoria is ranked lowest for transparency on how data is governed.

The pipeline of approved data centres will require a reported 6 gigawatts of power to fuel them – a massive scale up from the 100 megawatts consumed by the existing generation of data centres. A gigawatt is enough to power about 500,000 homes at once.

So why is this new build-out happening so close to where people live? It’s because the majority of existing data centres in places like Australia were built for another era: for things like streaming and cloud hosting. These applications kept things feeling instant by holding copies of popular content close to users, while the heavy lifting happened far away.

Now, with the expansion of AI services, this method for delivering applications doesn’t quite cut it. While the AI models running ChatGPT and Claude can still be trained in data centres far from where people live, growing use of these models means much more compute capacity is needed closer to where people are.

Every time you ask ChatGPT or Claude a question, the model has to generate a fresh response; it can’t be cached. This is called “inference”, and unlike a cached TV episode, it can’t be stored in advance close to you.

As AI spreads across business and consumer applications, more inference capacity is needed closer to where people are.

Yet another factor here is power. Connecting a new site to the grid can take years, so operators favour locations where high-voltage transmission and substation capacity already exist. These are concentrated in and around the cities Australians already live in.

When developers choose these sites, they can meet that need for speed again – which some state governments are facilitating by fast-tracking development applications.

Back to top

---

5. What’s in it for us? Why does the government want data centres?

Nations have long contested control of land, sea and air spaces. Data is the newest frontier, and just like the others, it’s physically located somewhere, under someone’s law. Data centres are where the digital economy resides, and the country that controls the underlying land shapes how the data can be used. This is known as data sovereignty.

Sovereignty matters because our most valuable data is increasingly processed by AI. Consider our medical records, banking details, farm and business records, minerals exploration surveys, and research data.

These are assets of national and personal importance, yet when this data is sent to offshore commercial AI services, it becomes subject to foreign contracts, and may become subject to foreign law, putting our control and intellectual property at risk. Data centres built on Australian soil keep our data subject to Australian laws.

Another reason is economic. Data centre investment in Australia is booming. In NSW alone, investment in the sector has grown by around 75% a year over the past three years, with more than 60 facilities operating or under construction and a further A$50 billion of projects in the pipeline.

Similarly, Victoria’s action plan could deliver a pipeline worth more than $25 billion. Our governments are shaping this growth: the federal government released the National AI Plan in late 2025, and NSW released its Data Centre Policy Framework in August 2026.

Here’s the catch: location alone doesn’t guarantee complete sovereignty. The AI models running inside the data centre matter too. Models come in two main types: closed and open-weight. Closed models may be subject to foreign law and export controls, and today’s access and prices are not guaranteed tomorrow.

Open-weight models running on Australian soil carry less risk: there are no per-token fees, and overseas companies can’t switch off our access to them. The real question then becomes not whether Australia should build data centres, but what kind of data centres Australia should be building and operating.

Then there is energy. Both federal and state governments now expect data centres to incorporate new clean energy generation or storage to offset their energy demands. Regulated well, this could help bankroll the renewable energy transition. Done poorly, it can strain the grid.

So what’s in it for us? Potentially a great deal: sovereign capability, a funded energy transition and regional growth. But only if we are selective. The question is not whether Australia should host data centres. It is whether we choose the ones that serve us.

Back to top

---

6. Where do the profits go?

The short answer is: mainly overseas.

When it comes to following the money in data centres, think of them not as tech companies, but rather commercial landlords that rent out space to tech companies.

In a normal warehouse, a tenant pays based upon size and location. For a data centre, the key factor is power, rather than space.

Access to reliable power drives the amount of computers that can be stored inside, and determines how much their tenants – tech companies like Amazon, Google, Microsoft and Anthropic – will pay.

For the billions of upcoming investments into Australian data centres, the majority (up to 65%) goes towards computer equipment. Almost all of this money heads overseas to computing giants like Nvidia, which enjoys 70% gross margins on its crucial AI chips.

The remaining money is spent on purchasing land and specialist cooling equipment, while 5% is spent on improving power grid capacity. It’s this money that is more likely to stay in Australia, paying for land, construction, engineering, and other local services.

Once operational, data centres generate remarkably high profit margins of around 50%. These margins are driven by the current 97% global occupancy rate of data centres, and long-term leases where tenants pay for their own power consumption.

In the eyes of investors, these facilities are essentially highly appealing premium commercial real estate with consistent, long-term yields. Australian data centre owners include:

• global investment funds such as Blackstone
• foreign pension funds from Canada and elsewhere
• Australia’s largest superannuation funds including Commonwealth Super, AustralianSuper and UniSuper
• Australia’s sovereign wealth fund, the Future Fund
• stock market investors, for listed companies such as NextDC on the ASX or Digital Realty on the US NASDAQ exchange.

Ultimately, the tech tenants pay rent to their data centre landlords. In turn, the tech companies charge Australian businesses for their services.

While these facilities serve Australian businesses, their primary economic contribution is the short-term sugar hit during the construction phase.

Beyond that, the majority of the money invested, and the profits made, will ultimately end up in the hands of overseas chip companies and investors.

Back to top

---

7. Can the water be recycled?

Potentially.

Data centres require constant cooling to work properly and prevent damage from overheating. There are several ways to keep them cool, but a common approach involves circulating a coolant liquid through banks of servers that store and process online data.

This coolant, often a mix of glycol and water, removes heat generated by the servers and is then sent to external cooling towers. There, a combination of mains water and large volumes of air pulled in by massive fans causes the hot coolant to rapidly drop in temperature. This process mirrors how a person sweats to cool down.

Currently, the water used to cool data centres comes from urban water systems. This puts pressure on drinking water supplies. In the American city of The Dalles, Oregon, Google’s data centres reportedly consume nearly 40% of the city’s total annual water demand.

Sydney Water, Australia’s largest supplier of urban drinking water, is grappling with the growing demands of this thirsty industry. Within a decade, data centres are expected to consume 25% of Sydney’s daily water demand, up from just 1% currently. This means data centres could soon suck up roughly 139 billion of the 556 billion litres Sydney residents consume each day.

So where else could this water come from?

A new data centre slated for Melbourne might have the answer. Amazon Web Services is set to build a new data centre in the city’s west that will be cooled with recycled wastewater. This wastewater will be treated and supplied by the Melton Recycled Water Plant, making it the first Victorian plant to deliver recycled water to a data centre.

This is part of the Victorian government’s broader ambition to use recycled water across various industries including data centres, manufacturing and even health care.

Sea water is another option. The Google data centre in the Finnish city of Hamina, for instance, is cooled using seawater from the Gulf of Finland. However, this approach is expensive and difficult to pull off because evaporating sea water produces a salty brine that can corrode infrastructure, such as data centres.

Generally, data centre cooling water is only used once before it flows back into the local sewerage system. But the American city of Quincy, where a special facility has treated cooling water from Microsoft’s data centre since 2021, suggests data centres could reuse it. But treatment is crucial because evaporating cooling water makes it extra salty and more likely to damage data centre servers.

Back to top

---

8. What happens to data centres when the AI bubble pops?

If the AI bubble bursts, data centres won’t suddenly become empty buildings, because the underlying technology won’t disappear. Some investors may lose money, and some proposed projects may never be built. But the world will continue producing, storing and using more data.

The first projects to suffer would be those without confirmed customers, reliable electricity, or strong internet connections. Some would be cancelled before construction starts. Others might be delayed, reduced in size, or sold to another company. Operators that built too much capacity may have to lower their prices.

The computer chips inside data centres would be most at risk. They are expensive and can become outdated quickly. In January 2026, Microsoft said about two-thirds of its quarterly infrastructure spending went towards shorter-life equipment, mainly computer chips. The remaining spending went towards assets expected to earn revenue for at least 15 years.

If demand for AI slows, newer chips could still be used for cloud computing, everyday AI services, medical research, engineering simulations, weather forecasting and digital media. Older chips that consume too much electricity may no longer be economical. They would need to be reused, sold or recycled responsibly.

The buildings themselves are a different story. Their electricity connections, fibre networks, cooling systems, security systems and serviced land can remain valuable for many years.

Banks, hospitals, government agencies, universities, retailers and streaming services all depend on data centres. These needs will continue even if excitement around generative AI declines.

We’ve seen this pattern before. During the internet boom of the early 2000s, companies in the US built too many fibre-optic networks. The dot-com crash caused serious financial losses, but the internet kept growing and the infrastructure supported its next stage.

The International Energy Agency has also tested what could happen if AI grows more slowly than expected. Data centres used about 415 terawatt-hours of electricity worldwide in 2024, nearly one-and-a-half times the 283.9 terawatt-hours of electricity generated across Australia that year.

Even under a slower-growth scenario, global data centre electricity use reaches about 700 terawatt-hours in 2035, approximately 70% higher than in 2024.

Australia should therefore continue developing data centres, but do it carefully. New data centres should be built in stages as real customers are secured. Developers should pay for the electricity infrastructure their projects require. Buildings should also be designed so their equipment and cooling systems can be changed as technology evolves.

If the AI bubble bursts, weak projects will fail. Well-designed data centres will adapt. Success will depend on building the right facilities in the right places, not simply building as many as possible.

Back to top

---

9. Are they/will they become military targets?

They already are. When the US and Israel launched their war against Iran earlier this year, one of the first sites Tehran struck in retaliation were data centres in the United Arab Emirates and Bahrain.

Experts called this moment a “sea change in warfare”.

Our most valuable data has been subject to cyber attacks for decades. But now, governments and tech companies must also think about how to physically protect their data when a conflict breaks out.

Some data centres are obvious targets due to their sheer size alone. But size is not the most important factor – the significance of the data is what matters most.

An adversary would likely want to cause the most damage to the wellbeing of a nation for an extended period of time. This means targeting the data that forms part of a nation’s critical infrastructure.

How to recognise and prepare for these new threats is still poorly understood, which is why I’ve started a research program at the University of Canberra to study this. These are three issues I believe require urgent examination.

First, there is the danger of attacks from missiles and drones, even as far away as Australia. Iran used drones to attacks data centres across the Middle East. And Ukraine has sent waves of drones thousands of kilometres into Russia to attack warehouses and logistics centres belonging to online retailers.

This has prompted much debate about how little is being invested in defending these sites in Australia. One expert has posited that fortifying data centres could cost hundreds of millions of dollars.

Second, Ukraine has proven so resilient to repeated Russian attacks on its infrastructure because it backed up data from 50 government institutions to cloud servers outside the country in just three months.

This is something Australia could replicate. But for Australian data to be protected by our own laws, the data would need to be physically stored at Australian-based data centres. To have our backup data protected on foreign servers, we would need very strong partnerships with the countries involved.

Last, but not least, there’s the danger of what’s known as cascading failure. Due to the interconnectedness of our critical infrastructure, an attack on, say, the power grid or undersea cables could take out our data centres, too.

The best example of cascading failure was the CrowdStrike outage in 2024 that caused computer systems around the world to crash. It wasn’t caused by an attack. But it could have been.

The threats to data centres will only continue to evolve, and not just in the cyber domain. Protecting them needs to be a much bigger priority. Putting a fence around them and locking the doors is just not enough.

Back to top

---

Disclosures

Amir Karton receives funding from the Australian Research Council. He is a visiting scientist at Microsoft AI4Science, working on fundamental quantum chemistry in a scientific collaboration unrelated to data centre infrastructure, commercial AI services, or government policy.

Andrew Cullen is a Board Member of Music Technology Australia, a member of the World Intellectual Property Organisations AI Technical Exchange Committee, and the managing consultant at EthiCan AI consulting.

Ian A. Wright has received research funding from local government bodies, the New South Wales government and the federal government. He has previously worked for Sydney Water as a scientist and a catchment officer.

Frank den Hartog is the Cisco Research Chair in Critical Infrastructure at the University of Canberra.

Bronwyn Cumbo receives funding from the Australian Public Policy Institute for the project Grounding the Cloud: Co-creating a sustainable, community-inclusive data centre industry.

Ehsan Noroozinejad, Jacqueline Boaks, Olivia Shen, and Sarah Barns do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.

Alexandra Hansen, Deputy Editor and Chief of Staff, The Conversation and Signe Dean, Science + Technology Editor, The Conversation

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Michael Noetel, The University of Queensland

Earlier this week, researcher Jacob Coxon quit Anthropic, saying the firm and its competitors are “gambling with our lives”. “We really do earnestly believe AI could kill all humans,” added current Anthropic researcher Evan Hubinger in a post on X.

Coxon isn’t the first to down tools over fears of AI doom. The idea that AI could wipe out humanity, advanced in Nick Bostrom’s 2014 book Superintelligence and the influential LessWrong forum, has long circulated among researchers. There are many scenarios for how this could happen, but the core idea is that AI smarter than humans could escape our control and destroy us.

In 2024, Jan Leike and Daniel Kokotajlo quit OpenAI over safety concerns. This year, Anthropic safety chief Mrinank Sharma departed, warning “the world is in peril”. Alex Turner left Google DeepMind after it signed a deal with the Pentagon permitting “killer drones”.

But Coxon’s resignation has made waves, with more researchers admitting they think AI might kill everyone. So if the people building AI believe it could cause extinction, why keep building it? There are three main reasons.

Some think the risk is worth it

AI leaders acknowledge the risk of losing control and killing everyone. In 2023, the chief executives of OpenAI, Anthropic and Google DeepMind agreed that AI extinction risk should rank alongside pandemics and nuclear war. Anthropic’s Dario Amodei puts the chance of things going “really, really badly” at 10–25%.

Yet Amodei also promises a world without poverty or disease, while Elon Musk speaks of AI-enabled “universal high income”.

This is the first reason for pursuing AI: the belief that the benefits outweigh the risks. Perhaps so, but that decision arguably deserves a more democratic process.

Some say you can’t study the danger from a distance

The second reason: you can’t learn to make dangerous AI safe without building it first – like a spacecraft, you can study safety from afar, but can’t really test it without going to space.

OpenAI’s plan is “iterative deployment”: release each model, learn from its problems, and fix them in the next one. The idea is like getting as close to the cliff edge as possible to see what the jump looks like.

Some feel it’s winner-takes-all

The third and perhaps most important reason is the race. OpenAI’s Sam Altman recently said “we are close to creating a genie that can grant any wish”.

The trouble is everyone wants to hold the lamp – it would be hugely profitable, and each company doubts the judgement of rivals to use their wishes wisely.

So they race, reasoning that if they slow down, someone else will get there anyway, so it’s better to arrive first as the “responsible one”. Some fear even a mutual agreement would be broken in secret. So they press on.

AI making better AI

You might doubt runaway AI is plausible. But when the companies themselves raise the alarm, we should listen.

AI firms already report signs of “recursive self-improvement”, where each model helps to build a better successor. According to OpenAI’s chief scientist, models are improving faster than humans’ ability to control them.

In July, hundreds of AI employees signed an open letter calling for a slowdown. But the dynamics of the race make that hard for any single company – or country – to do alone.

A classic arms race

AI research has the hallmarks of an arms race. OpenAI doesn’t want to lose to Anthropic, and the United States doesn’t want to lose to China.

History offers a template for how to manage a situation like this, with rules binding all players, and enforcement everyone can verify.

Nuclear weapons are the classic case. Treaties and verification systems haven’t eliminated the risk of nuclear war, but they have slowed proliferation, and no nuclear weapon has been used in conflict for 80 years.

Rules for AI

In the US, where most cutting-edge AI research happens, the Trump administration shows little sign of slowing AI development.

In its first week it scrapped the old AI safety rules. Now it is trying to override state-level rules, arguing caution risks losing the race to China.

Some politicians are pushing back. California recently passed laws supporting independent assessment of AI systems. US senator Bernie Sanders introduced a bill to ban superintelligence, and British MP Alex Sobel introduced a similar bill.

Companies have moved too. OpenAI paused its most advanced training after a swarm of its agents hacked another startup in August. The company’s head of policy now says that when safety and speed conflict, safety should win.

Still, without binding rules, we’re relying heavily on the goodwill of a handful of companies.

What happens now?

In mid-2025, researchers published what might be our best guide to the coming years: a detailed scenario called AI 2027. Since then, AI capabilities have advanced faster than predicted.

Unless something changes, staffers who quit over safety will simply be replaced, AI models will help build better AI models, and each generation will grow harder to monitor and control.

Is the situation hopeless? I hold out three hopes.

First, that more people recognise AI escaping human control is a bigger risk than AI’s water use.

Second, that governments listen to their people. In the US, two thirds say AI is moving too fast.

Third, that we have a good plan ready before a crisis hits. The best plan, in my view, looks something like this: delays, transparency and verification to slow the race and keep humans in control.

Insiders at the world’s top AI companies say our current safety plan isn’t good enough. If they’re leaving their jobs over safety fears, we should listen to what they have to say.

Michael Noetel, Associate Professor of Psychology, The University of Queensland

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Andrew Cullen, The University of Melbourne

Over the weekend, Anthropic chief executive Dario Amodei called for artificial intelligence (AI) companies, including his own, to slow down their work. Sam Altman and Elon Musk, heads of rivals OpenAI and xAI respectively, agreed.

The move reflects concern across the AI industry and more broadly about the dangers of new, rapidly improving systems. Recent high-profile incidents such as OpenAI AI agents hacking another company and hijacking a public website have shown current systems can break out of safety confines – and even more capable systems are in development.

Further complicating AI safety is the tension between safety and performance. Companies will be reluctant to limit the performance of their models in the name of safety for fear of losing ground to competitors. US President Donald Trump has also rejected calls for a slowdown, for fear of losing ground to China.

This means any successful effort at “pacing the rate of capabilities advancement so that risk prevention has time to keep up”, as Amodei puts it, will require significant cooperation between rival companies – and nations.

Risk minimisation

New technologies often bring new risks, and new concerns. Often governments, researchers and companies do find ways to manage those risks.

In 1975, the Asilomar conference on then-new DNA technologies did much to ensure research didn’t get ahead of our understanding of safety and risk. Similarly, in the 1990s, the US government attempted to build a consensus on limiting cryptography and computer security.

The AI situation has an extra twist: the industry is in the middle of a gold rush. Scientific rivals may be able to restrain themselves, but commercial rivals rarely hold back.

There are clear precedents for self-regulation failing in the face of competitive pressure. In the Boeing 737 Max disaster in 2018, for example, pressure to catch up to rival Airbus led Boeing to hide the limitations of the Max, which ultimately cost lives.

In the AI race, the scale of the competitive tension is even greater. Anthropic and OpenAI are both competing to establish market dominance before pursuing share market listings that could raise tens or hundreds of billions of dollars.

And at the nation-state level, the stakes are higher again, with the US and China each hoping to use the new technology for geopolitical advantage.

The politics of a pause

Amodei’s slowdown proposal centres around a three-step plan: embedding independent third-party safety reviewers, establishing coordinated industry safety standards within democratic nations, and eventually securing global agreements. This would include strict limits on AI chip exports to companies and countries that do not agree to prioritise AI safety.

Anthropic and OpenAI have already agreed to the first phase of this plan, despite their history of suing, publicly insulting, and undercutting each other in pursuit of market dominance.

While recent high-profile hacks may have forced their hands, the leading AI companies may benefit from a development pause or slowdown. For one thing, it could put off strict legislation such as US senator Bernie Sanders’ proposed Ban Artificial Superintelligence Act. For another, by creating expensive safety standards and limiting chip exports, it could block smaller competitors – especially Chinese companies such as DeepSeek and Alibaba – from catching up.

A pause would also give the overstretched frontier labs a chance to recoup, recover, and focus on profits over progress.

These AI labs have spent enormous amounts to produce their existing models, which must be repaid. At the same time, progress is hitting speedbumps as new high-quality training data gets harder to find, and building the massive data centre infrastructure needed to sustain development is a challenge in itself.

A coordinated safety pause could be a convenient public reason for a plateau in AI model performance.

What can be done

Making AI safe won’t be easy. The governance of software is notoriously difficult, as past attempts to legislate encryption software have shown.

However, unlike other software, AI is dependent on relatively scarce physical hardware. It needs advanced silicon chips and the massive data centres required to power them.

This is where governments have real leverage. Here they have ways to monitor and control AI development, if they can find the legislative will.

In the meantime, there are steps businesses and governments can take to minimise how exposed we all are to AI-driven harms. This should include ensuring that critical safety infrastructure – like power grids, water supplies, and military systems – are not only isolated from AI, but potentially even isolated entirely from the internet.

And the question of liability is also important. It can’t just be the users of AI systems who face legal jeopardy for acts assisted by AI, but also the people responsible for making AI systems.

Fundamentally, harm produced by AI – even by “autonomous” AI systems – isn’t an abstract technological byproduct. It is the direct result of decisions made by both those making AI, and those using it.

Andrew Cullen, Senior Research Fellow, School of Computing and Information Systems, The University of Melbourne

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Jak Kakhkharov, Australian Catholic University

Australia is in the middle of a data centre construction boom. Driven by artificial intelligence (AI) and growing demand for digital services, global technology companies are expected to invest up to A$150 billion in Australian data centres by 2030.

The scale of the investment is hard to ignore. However, a more important question is whether Australia will benefit from higher productivity and economic growth, or simply from a brief construction boom.

Data centres can create jobs, strengthen digital infrastructure and support AI development. But they can also put pressure on housing, electricity networks and skilled labour.

---

Data centres have existed for decades. But something has changed, and this essential infrastructure is now at the centre of a major policy debate.

This article is part of The Conversation’s series on data centres – what they are, why we need them, and why they’re suddenly so controversial.

---

Why is Australia a favoured location?

Data centres are packed with computers that store, process and move digital information, the backbone of cloud computing. In theory, they can be built anywhere, so why does big tech have Australia in its sights?

Australia offers several advantages for global technology firms. It is politically stable, has strong institutions, and sits close to the fast-growing Asia-Pacific region.

The direct economic benefits are significant. Construction projects create work for engineers, electricians and builders. They also increase demand for materials, electricity infrastructure and specialised equipment.

Yet the boost to economic growth may be smaller than the headline investment figures suggest.

A large share of the spending goes towards imported goods, rather than production in Australia: servers, processors, chips and networking equipment. That means much of the money leaves Australia and does not reach the local economy.

The hidden costs of the boom

Every major investment boom comes with trade-offs. One of the biggest risks is competition for construction labour. Data centres require many of the same workers needed to build houses, roads and renewable energy infrastructure.

Australia already faces shortages of up to 72,000 electricians, engineers and skilled trades. There are media reports that electricians with just two years’ experience are being offered salaries of $200,000 a year to work on data centres.

If data centres hire more of these workers, labour costs could rise across the economy. Housing projects may become more expensive, infrastructure projects could face delays and renewable energy developments may struggle to secure workers.

This matters because housing affordability remains one of Australia’s biggest economic challenges.

Power, water and land are also under pressure

As more data centres are built, demand for electricity and water will increase. If supply doesn’t keep pace, prices could rise for households and businesses.

Land use is another challenge. Some sites allocated for data centres might otherwise have been used for housing, logistics facilities or other productive purposes.

These costs do not necessarily outweigh the benefits. But they form part of the economic trade-offs policymakers need to consider.

The inflation risk

The Reserve Bank of Australia has raised concerns about capacity pressures in the Australian economy and noted data centres could contribute to inflation in several ways.

First, stronger competition for skilled workers may push wages higher in construction and related industries.

Second, increased demand for construction materials could drive up project costs across and increase prices in the economy.

Third, a large inflow of investment can add to demand at a time when policymakers are trying to keep inflation under control. These factors could make it harder for the Reserve Bank to bring down inflation, which is well above the 2-3% target band.

A surge in copper

Data centres also use large quantities of copper for cables, power systems and cooling equipment. As data centres are built around the world, demand for copper is expected to grow. S&P Global forecasts that total global copper demand will surge by around 50%, climbing from 28 million metric tonnes in 2025 to 42 million metric tonnes by 2040.

Higher copper prices will benefit Australia’s miners. Indeed, BHP, the world’s largest copper producer, said last month copper contributed more than half of its earnings for the first time, overtaking iron ore, as the price of copper surged almost 50% in the year.

But there is also a downside. More expensive copper raises costs for domestic construction projects, electricity networks and renewable energy developments.

This highlights a broader theme of the data centre boom: the same economic trend can create gains for some industries while imposing costs on others.

Who captures the value?

Perhaps the most important question is who captures the value data centres create.

Owning the buildings does not necessarily mean owning the technology. Many data centres are operated by multinational companies that also control the software, intellectual property and AI systems running inside them.

If Australia hosts the infrastructure but develops few local AI-related capabilities, much of the long-term value may flow overseas. As assistant minister for the digital economy, Andrew Charlton said recently:

Simply building data centres in Australia does not necessarily mean that Australia captures the economics of AI. We can supply the land. We can supply the electricity. We can host the machines. And still find that much of the value […] flows offshore.

The biggest gains are likely to come from building domestic expertise in AI, software development and digital services.

How do we maximise the benefits?

The key policy question is how to maximise the benefits of data centres while minimising the costs.

This means expanding the supply of skilled workers, electricity infrastructure and industrial land so that construction does not crowd out housing, renewable energy and other important investments.

Training more electricians, engineers and technical specialists will be part of the solution. But training takes time. In the short term, carefully targeted migration policies may help ease labour shortages in critical occupations.

Australia’s data centre boom brings investment, jobs and the promise of a sophisticated digital economy, but it also creates new pressures on housing, energy and skilled workers.

The real challenge for policymakers is ensuring Australia captures enough of the value these investments create.

If governments strike the right balance with their policies, the current boom could boost productivity and strengthen Australia’s position in the global digital economy. However, if they do not, Australia may find that the benefits of the construction boom are fewer than the investment figures suggest.

Jak Kakhkharov, Senior Lecturer in Finance, Australian Catholic University

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Michael Vardon, Australian National University

Australia’s pipeline of huge data centre projects has raised questions about how much energy and water these giant server farms will use.

These questions aren’t simple to answer because we don’t have the data.

We can’t check water or energy use for individual facilities or inspect national data centre statistics and no operator is obliged to disclose these figures.

That may change. Last month, the National Cabinet agreed to develop mandatory national standards for large data centres, covering energy, water and land use. The legislation is due early next year.

No detail is available, and little has been said about their lifetime environmental footprints, from land clearing to noise issues to electronic waste to greenhouse gas emissions. These must be counted too.

The fastest way to settle arguments about the environmental impact of the data centre boom is to collect and publish the data.

---

Data centres have existed for decades. But something has changed, and this essential infrastructure is now at the centre of a major policy debate.

This article is part of The Conversation’s series on data centres – what they are, why we need them, and why they’re suddenly so controversial.

---

Hungry for power

In 2020, data centres were estimated to use about 1.5% of the world’s electricity. But since 2022, the artificial intelligence boom has led to rapid growth.

Australia’s 165 operating data centres use an estimated 3% of the power from the National Electricity Market, Australia’s largest grid.

By the mid 2030s, the Australian Energy Market Operator predicts this will reach roughly 13%. This would put them on par with aluminium smelters.

Where will the water come from?

Water is much harder to quantify. My rough estimate suggests current national water use is a fraction of 1%, tiny compared to agriculture.

It’s not the national figure that matters, but where, when and how much water is taken.

A new data centre drawing 100 megalitres of water annually from a rural catchment may not matter over a wet winter. In a drought it might.

Some centres are proposed for hot, dry areas where water is already a concern, such as Bundey and Tailem Bend in South Australia. Their water use will depend on the cooling technology used.

At most data centres, only the operator has the data to calculate water use. They are not currently required to disclose use. The water utility sees a connection, not what happens inside the fence. Regulators could look, but it seems nobody has asked them to.

Sydney’s data centres currently draw about 3.5 billion litres a year, less than 1% of the city’s water supply. But this could rise to 25% within a decade, according to Sydney Water estimates.

By comparison, industry modelling suggests a figure of 2% by 2030.

This tenfold difference in forecasts for the same city is due to the fact much has to be assumed given the lack of data.

Trade-offs and complexities abound

If a data centre runs on renewables and uses the same water in a closed loop or new direct-to-chip cooling, the impact will be much smaller than using evaporative cooling, where much water is lost.

Similarly, power drawn from fossil-fuel dominated grids means water is used indirectly, as coal and gas plants use lots of water – especially compared to renewables.

The trade-off is direct: less water for cooling usually means more electricity.

The electricity used ends up as heat. Field measurements show downwind air temperatures average 0.7 to 0.9°C warmer.

While energy efficiency is improving and advanced liquid cooling can cut water use by up to 50%, this isn’t a full solution. Australia could build giant water-efficient data centres and still end up with climbing water use.

Lessons from abroad

In Ireland, data centres use 23% of electricity, close to all households combined (28%). But data centre demand is rising fast, jumping 10% last year – while household demand rose 2%. We know this because Ireland’s statistical office measures it. Ireland measures. Australia estimates.

Ireland’s transparency lets policymakers make hard decisions such as limiting new grid connections and requiring the use of renewables to meet 80% of demand.

The main lesson from the United States is that the most intense impact is often local. Communities chosen to host new data centres often push back hard based on issues around noise, energy and water use, and visual impact. Backlash against new data centres is becoming intense in many areas – and bridges political divides.

In Virginia, the constant hum of cooling fans drove local authorities to rewrite zoning rules and reduce noise limits.

Planning for growth

At a national scale, today’s data centres have modest environmental costs and potentially large economic benefits.

But we don’t know what this looks like when translated to local impacts.

The real issue is what’s in the pipeline. Right now, 225 data centres are proposed or under development. If all are approved, this would represent 67 gigawatts of new power demand – a massive jump on the 2.2 gigawatts currently operating.

Some of these projects won’t get up. Over a third of the planned projects listed in 2025 were cancelled. But some will be built. Because company disclosures are voluntary, selective and unaudited, they cannot be compared with each other or checked.

As a result, anxiety is filling the space where data should be.

What should Australia do?

Ideally, new data centres would draw power from new renewables with minimal indirect water use.

At a federal level, authorities want large data centres to source their own clean power. But Queensland and the Northern Territory want to use gas. This issue is yet to be resolved.

For water, maximising efficiency is the mantra. While efficiency is good, it is not enough. Water is extracted, not efficiency. What matters is total use against what is available and how much competition there is.

Energy use is expected to include reporting, a regulator and testing. Water use will need the same, and the standards should include noise, electronic waste and emissions. If, as the industry and government claim, there is no need to worry, neither should object to such standards.

Done poorly, Australia’s data centre boom could suck up scarce water supplies and increase the use of fossil fuels. But if done well, we could grow the industry using clean power and ensuring water use is monitored and minimal.

The difference comes down to having data. Much of the extra data we need to track data centre impact can be gathered by the Australian Bureau of Statistics – just as Ireland has done.

Transparency can help the data centre industry improve its social licence to operate. With data we can see if the largely national and international economic benefits of data centres justify the mostly local environmental costs.

Michael Vardon, Associate Professor of Environmental Accounting, Australian National University

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Johanna Lim, University of Sydney

Until recently, data centres attracted relatively little public attention. They were largely treated as invisible pieces of digital infrastructure: essential but rarely discussed outside technical and industry circles.

But over the past 12 months across Australia, data centres have become the subject of intense political debate, community opposition, planning disputes and parliamentary inquiries.

Questions are being raised about how much electricity and water they use, where they should be built, who should pay for the infrastructure needed to support them, and whether Australia benefits from their continued expansion.

These questions are feeding into government policy. Following Prime Minister Anthony Albanese’s speech at the University of Sydney in July, a National Cabinet meeting in August reaffirmed plans to legislate nationally consistent mandatory standards for large data centres by early next year. These will include requirements around their energy, water and land use.

So how did a piece of digital infrastructure that once attracted relatively little public attention become such a prominent policy issue? Data centres themselves are not new. What has changed is their scale, purpose and the resources required to support their growth.

---

Data centres have existed for decades. But something has changed, and this essential infrastructure is now at the centre of a major policy debate.

This article is part of The Conversation’s series on data centres – what they are, why we need them, and why they’re suddenly so controversial.

---

The evolution of data centres

Data centres are specialised physical facilities that house servers, networking equipment and data storage systems. They are foundational infrastructure for the modern digital economy, supporting a wide range of services such as streaming, social media, banking, emergency response systems, and artificial intelligence (AI).

The origins of data centres can be traced to the 1940s, when early computers were so large they needed dedicated spaces to house them. As computers became smaller, more powerful and more accessible, governments and businesses increasingly adopted their own IT infrastructure. They often operated on-premises server rooms to manage email, file storage and other internal systems.

The internet then drove a shift towards larger co-location data centres in the 1990s, where multiple customers could rent space for their servers in a single shared facility.

Then came cloud computing. This provided customers with on-demand access to computing resources over the internet, and further accelerated the growth of data centres. In 2006, Google opened its first hyperscale data centre. Amazon Web Services also launched its first cloud computing services in the same year. Cloud service providers subsequently built increasingly large facilities to meet growing demand for these services.

What was once a room or floor serving a single organisation evolved into massive dedicated facilities – often called hyperscale data centres – capable of supporting millions of users. Today, the scale of a data centre is often measured by its power capacity, in megawatts or gigawatts. This reflects how much electricity the facility can draw at any one time.

Australia’s data centre boom

Australia’s own data centre market reflects this shift in scale.

Australia currently has over 160 operational data centres, with most located in New South Wales and Victoria. There are at least another 90 facilities in the development pipeline.

The data centres now being proposed are also considerably larger than many existing data centres.

---

---

---

---

For example, the proposed 1.2 gigawatt Mamre Road Data Centre in Sydney would cover an area equivalent to the size of 52 rugby fields.

If built to its maximum capacity, it would become Australia’s largest single electricity user.

Why AI is driving bigger data centres

Hyperscale data centres typically contain at least 5,000 servers, occupy at least 10,000 square feet of physical space, and can draw over 100 megawatts of power. That’s enough to meet the annual electricity needs of more than 50,000 households.

In 2025, hyperscale operators accounted for 48% of global data centre capacity.

Cloud computing and the growth of everyday digital activity initially drove the expansion of these large, centralised facilities. But since OpenAI launched ChatGPT in late 2022, AI has rapidly accelerated this growth.

AI workloads are far more computationally intensive than traditional digital services. They are expected to account for approximately 70% of data centre demand by 2030.

Training advanced AI models requires dense arrangements of specialised chips working simultaneously to process large volumes of data. This can run continuously for weeks or months.

Once trained, AI models also require computing power to respond to users. This process is known as inference. While a single interaction requires considerably less computing power than training a model, that demand adds up across millions of users.

Inference represents an increasing share of AI’s energy demands. This will continue to increase alongside AI adoption.

A 2025 survey found 88% of organisations reported regularly using AI in at least one business function – an increase from 78% a year earlier.

The shift toward large-scale data centres is also about efficiency.

Larger facilities tend to be more efficient. They benefit from economies of scale and advances in facility design, such as optimising power distribution and using higher-performance chips and servers.

In Australia, on-premises servers are estimated to consume over seven times more electricity to perform the same computation as hyperscale and co-location data centres.

But while hyperscale data centres can use energy more efficiently for the computing they perform, the sheer scale and growth of demand mean their overall electricity consumption is still significant.

Data centres accounted for around 3% of electricity supplied through Australia’s main grid in 2025–26. This share is projected to reach 13% by 2035–36.

What comes next for Australia?

The growth of Australia’s data centre market is unlikely to slow anytime soon.

Australia remains a competitive destination for data centre investment, with strong government support, continued interest from major technology companies and an estimated A$150 billion in data centre buildouts by 2030.

But community opposition is emerging as a significant risk for new developments. This opposition is, in part, because of the land, power and water data centres consume.

But it’s also because there is uncertainty over whether the significant investment in AI infrastructure will truly generate sufficient economic returns – and who will benefit from those returns.

The future growth of Australia’s data centre sector may therefore depend less on whether there is demand or capital to build them, and more on where they can be built, who bears the costs, and whether communities benefit from hosting them.

Johanna Lim, Research Associate, Strategic Technologies, University of Sydney

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Marion Burgess, UNSW and John Davy, RMIT University

In the United States, areas such as Michigan have been hotspots for residents pushing back against data centre construction in their neighbourhoods. Now, similar concerns about noise – and potential health impacts – are gaining traction in Australia.

But we currently have very little data to link environmental noise from data centres to health effects. And we also can’t assume the US and Australia impose the same limits on industrial noise.

So, what do we know about the noise these large data storage facilities produce? And is this noise actually bad for our health?

---

Data centres have existed for decades. But something has changed, and this essential infrastructure is now at the centre of a major policy debate.

This article is part of The Conversation’s series on data centres – what they are, why we need them, and why they’re suddenly so controversial.

---

How much noise is safe?

Decibels (dB) are units that measure sound intensity. But when talking about how the human ear actually perceives that sound, we use a weighted scale: dBA.

For context, the sound of leaves rustling is about 30dBA, background music about 60dBA and average office noise around 70dBA.

Noises like these, below 70dBA, are generally considered safe for our health even if we are exposed to them in the long term.

But hearing damage can occur at higher levels. For example, in Australia workers must not be exposed to noise at 85dBA (the sound of a lawnmower or leaf blower) for any longer than eight hours, to prevent hearing damage.

According to the World Health Organization (WHO), excessive noise is also linked to conditions such as high blood pressure, heart disease and sleep disruption.

What noise do data centres produce?

The industrial noise data centres produce is typically described as a low-frequency hum.

Data centres need to continuously manage the heat they generate. This cooling requires large air-handling units (which regulate and circulate air), air-cooled chillers and cooling towers, as well as pumps to circulate water – all of which produce noise.

Data centres also have backup generators. These are used when there is a disruption to the power supply, or for a few minutes of testing during a maintenance check.

Both the air-handling units and the backup generators can produce noise levels above 85dBA within the facility.

However, this noise is reduced by barriers, distance and other mitigation measures to meet the compliance noise limits in surrounding areas.

Noise mitigation measures include choosing quieter cooling equipment, careful building design and use of soundproofing equipment such as acoustic louvres, physical barriers and silencers.

How is noise regulated?

Like any new industrial building, to gain environmental approval a company proposing a data centre will need to prove the sound won’t exceed noise limits.

These are set by the state and territories and are based on the surrounding land uses.

The company must submit an independent noise assessment that outlines all estimated noise the data centre will produce, and the measures it will use to mitigate excessive noise.

A critical requirement is compliance with the nighttime limit. In residential areas in Australia, this is typically 35–40dBA.

These noise limits for residential areas at night are similar to the levels the WHO recommends in its 2018 guidelines for environmental noise in Europe.

The environmental approval process in Australia also provides controls to prevent “creeping background noise”. The aim is to limit the total cumulative noise in the area, based on the way surrounding land is used for residential, commercial and industrial purposes.

This means when a new data centre applies for development, its individual noise limit may be adjusted downward if existing noise is already close to the acceptable limit.

This is an important safeguard as it becomes common for multiple data centres to be installed in one area, as in Lane Cove in Sydney.

Noise continues to be contentious

Last month, Victoria’s Environmental Protection Authority (EPA) objected to the construction of a data centre for the first time. The EPA cited concerns about the impact noise could have on nearby residents in Campbellfield in Melbourne’s north.

In 2025, an environmental noise impact assessment was submitted for a proposed data centre in Hazelmere in Western Australia that would have 64 rooftop chillers and 88 diesel generators.

While noise-mitigation strategies – such as soundproofing, acoustic louvres and silencers – would be required to ensure the noise complied with the 35dBA residential limit, the assessment concluded these would not be difficult to achieve.

However, in May the local council rejected the Hazelmere proposal due to concerns about the potential noise impact on nearby residents and a school, as well its proximity to the river and Aboriginal cultural sites.

What do we actually know about the health effects?

Evidence about the health effects from any form of environmental noise requires long-term studies, with large populations. When it comes to data centres, we simply don’t yet have the research data that identifies any health effects.

But similar concerns have arisen previously about the low-frequency hum from wind farms being built near residential areas. So a comparison could be relevant in the meantime, as their noise level limits are similar.

In 2015, the Australian National Health and Medical Research Council released an independent, systematic review of the available evidence about wind farms and human health. It concluded that, in relation to noise, “there is no direct evidence that exposure to wind farm noise affects physical or mental health”.

While it takes some time for health effects from excess noise to be identified from studies, any new noise source in an area does lead to annoyance in the short term.

Annoyance can lead to stress-related health effects. However, studies have shown that as long as the noise is limited to levels similar to those recommended in the WHO guidance, the annoyance reduces with time.

The bottom line

The noise limits that apply in Australia to data centres are in broad agreement with the WHO guidelines to avoid health effects from environmental noise. To date, we don’t have any contrary evidence that shows the noise from data centres has an impact on health.

Marion Burgess, Honorary Associate Professor, School of Built Environment, UNSW and John Davy, Adjunct Professor, School of Science, RMIT University

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Matthew Benetti, Climateworks Centre and Eleanor Guerra, Climateworks Centre

Australia is in the middle of a rapid data centre rollout. Warehouses full of computer servers that run the artificial intelligence (AI) tools society increasingly uses are being built at pace, and they draw heavily on essential resources.

Data centres need large amounts of electricity, water and land. Without the right planning and oversight, developers will secure what they need at the expense of Australia’s economy, nature and communities.

Without careful design, data centres could increase power bills, put pressure on water supplies, and slow the transition to net zero. With it, they can do the opposite, stabilising the electricity grid, attracting investment to regions, and creating economic growth.

It’s tempting to think futuristic technologies could make these problems disappear – things like putting data centres underwater or in space. These may eventually matter, but for now they are small-scale experiments, a long way from being able to absorb the gigawatts of demand Australia needs to plan for over the coming decade.

The Australian government is currently designing the rules that will govern AI and data centres, with national legislation expected in early 2027. So what does a good data centre future look like? And what needs to go into those rules to make this a reality?

---

Data centres have existed for decades. But something has changed, and this essential infrastructure is now at the centre of a major policy debate.

This article is part of The Conversation’s series on data centres – what they are, why we need them, and why they’re suddenly so controversial.

---

The right kind of energy

Energy is a significant pressure point. The federal government wants new data centres to match their electricity demand with new renewable generation, backed by “sensible firming arrangements”, which means reliable backup power for when renewable supply is low.

Most states and territories agree. However, Queensland and the Northern Territory want data centres to have the flexibility to use fossil fuels both as their primary supply and for firming.

From a cost and climate perspective, neither the federal nor state and territory governments have got this quite right. The use of gas and diesel, whether as primary or backup supply, is inconsistent with Australia’s climate objectives and keeping costs down. It also raises concerns about local air quality and community health.

Australia has committed to net zero emissions by 2050. We have an interim target of 62–70% reductions by 2035.

Without the right oversight, data centres could compromise this commitment. Data centres are being built by sophisticated, well-resourced operators. They are capable of firming their supply with batteries and other clean technologies rather than defaulting to fossil fuels.

Operated right, data centres could help the electricity grid work better, too. By shifting some computing tasks to off-peak times they could smooth the electricity demand peaks that increase costs and make the system less reliable.

Using water wisely

Water is the next pressure point. For instance, Sydney Water anticipates that data centre water needs, primarily for server cooling, could reach 250 megalitres a day by 2035.

The good news is that technologies already exist that can significantly reduce how much water they use. Closed-loop systems recycle the same water instead of letting it evaporate. Some new designs barely use primary water sources at all.

The National Australian Built Environment Rating System is developing a water standard for data centres to sit alongside its energy rating. To ensure efficient use of both electricity and water, data centres would be required to hit the top six star rating on both.

Location, location, location

Where data centres get built determines how much pressure they place on local resources, and how welcome they are in the communities hosting them. Around 160 already operate in Australia.

Most are clustered in Melbourne and Sydney for access to skilled workers, supply chains and low latency. That concentration is now pushing local power grids to their limits, putting pressure on network planning and potentially increasing costs for consumers.

Steering projects towards regions with spare capacity would ease pressure on the grid and use renewable power that might otherwise go to waste. These projects could also act as reliable buyers for new renewable developments, giving investors the revenue certainty they need to bring new capacity online.

At the same time, it will be important that this new demand does not compete with other kinds of industry for the same renewable energy supply, which could push up costs and slow the broader industrial transition away from fossil fuels.

Community benefits

Finally, the communities that host these facilities also deserve to benefit from their presence. That means genuine consultation and meaningful benefits as a condition of approval.

This could include community batteries or e-waste recycling programs to prevent critical minerals from entering landfill. It could mean engagement on where facilities are sited and the type of land they take up or upgrades to local water infrastructure.

In cities and regions, training pipelines could be established with local TAFEs and universities to help workers participate in the industry.

The good data centre future

Taken together, these measures describe a good data centre future. It has three key features.

First, data centre operators need to invest in new supplies of renewable energy rather than drawing on existing generation. Second, efficiency standards for energy and water must be mandatory rather than voluntary. And third, decisions about where data centres are located need to weigh resource use and community impact from the outset.

The legislative detail is still being drafted. Building in these guardrails now will ensure data centres minimise their strain on public resources and make a positive contribution to Australia’s economic outlook and energy transition.

Matthew Benetti, Policy Manager (Energy), Climateworks Centre and Eleanor Guerra, Policy & Stakeholder Coordinator, Climateworks Centre

This article is republished from The Conversation under a Creative Commons license. Read the original article.

T.J. Thomson, RMIT University; Daniel Angus, Queensland University of Technology; Jake Goldenfein, The University of Melbourne, and Kylie Pappalardo, Queensland University of Technology

Australians are among the most anxious in the world about artificial intelligence (AI).

This anxiety is driven by fears AI is used to spread misinformation and scam people, anxiety over job losses, and the fact AI companies are training their models on others’ expertise and creative works without compensation.

AI companies have used pirated books and articles, and routinely send bots across the web to systematically scrape content for their models to learn from. That content may come from social media platforms such as Reddit, university repositories of academic work, and authoritative publications like news outlets.

In the past, online scraping was subject to a kind of detente. Although scraping may sometimes have been technically illegal, it was needed to make the internet work. For instance, without scraping there would be no Google. Website owners were OK with scraping because it made their content more available, according with the vision of the “open web”.

Under these conditions, scraping was managed through principles such as respect, recognition, and reciprocity. In the context of AI, those are now faltering.

A new online landscape

Many news outlets are now blocking web scrapers. Creators are choosing not to use certain platforms or are posting less.

Barriers are being put in place across the open web. When only some can afford to pay to access news and information, then democracy, scientific innovation and creative communities are all harmed.

Exceptions to copyright infringement, such as fair dealing for research or study, were legislated long before generative AI became publicly available. These exceptions are no longer fit for purpose in an AI age.

The Australian government has ruled out a new copyright exception for text and data mining. This signals a commitment to supporting Australia’s creative industries, but leaves great uncertainty about how creative content can be managed legally and at scale now that AI companies are crawling the web.

In response, the international nonprofit Creative Commons has proposed a new voluntary framework: CC Signals.

Creative Commons licences allow creators to share content and specify how it can be used. All licences require credit to acknowledge the source, but various additional restrictions can be applied. Creators can ask others not to modify their work, or not to use it for commercial purposes. For example, The Conversation’s articles are available for reuse under a CC BY-ND licence, which means they must be credited to the source and must not be remixed, transformed, or built upon.

---

How would CC Signals work?

The proposed CC Signals framework lets creators decide if or how they want their material to be used by machines. It aims to strike a balance between responsible AI use and not stifling innovation, and is based on the principles of consent, compensation, and credit.

Simplistically, CC Signals work by allowing a “declaring party” – such as a news website – to attach machine-readable instructions to a body of content. These instructions specify what combinations of machine uses are permitted, and under what conditions.

CC Signals are standardised, and both humans and machines can understand them.

This proposal arrives at a moment that closely mirrors the early days of the web, when norms around automated access (crawling and scraping) were still being worked out in practice rather than law.

A useful historical parallel is robots.txt, a simple file web hosts use to signal which parts of a site can be accessed by the bots that crawl the web and look for content. It was never enforceable, but it became widely adopted because it provided a clear, standardised way to communicate expectations between content hosts and developers.

CC Signals could operate in much the same spirit. But, as with any system, it has potential benefits as well as drawbacks.

The pros

The framework provides more nuance and flexibility than the current scrape/don’t scrape environment we’re in. It offers creators more control over the use of their content.

It also has the potential to affect how much high-quality content is available for scraping. Without access to high-quality data, AI’s biases are exacerbated and make the technology less useful.

The framework might also benefit smaller players who don’t have the bargaining power to negotiate with big tech companies but who, nonetheless, desire remuneration, credit, or visibility for their work.

The cons

The greatest challenge with CC Signals is likely to be a practical one – how to calculate, and then enforce, the monetary or in-kind support required by some of the signals.

This is also a major sticking point with content industry proposals for collective licensing schemes for AI. Calculating and distributing licence fees for the thousands, if not millions, of internet works that are accessed by generative AI systems around the world is a logistical nightmare.

Creative Commons has said it plans to produce best-practice guides for how to make contributions and give credit under the CC Signals. But this work is still in progress.

Where to from here?

Creative Commons asserts that the CC Signals framework is not so much a legal tool as an attempt to define “manners for machines”. Manners is a good way to look at this.

The legal and practical hurdles to implementing effective copyright management for AI systems are huge. But we should be open to new ideas and frameworks that foreground respect and recognition for creators without shutting down important technological developments.

CC Signals is an imperfect framework, but it is a start. Hopefully there are more to come.

T.J. Thomson, Associate Professor of Visual Communication & Digital Media, RMIT University; Daniel Angus, Professor of Digital Communication, Director of QUT Digital Media Research Centre, Queensland University of Technology; Jake Goldenfein, Associate Professor, Melbourne Law School, The University of Melbourne, and Kylie Pappalardo, Associate Professor, School of Law, Queensland University of Technology

This article is republished from The Conversation under a Creative Commons license. Read the original article.