In August 2026 I lodged a submission with the Senate Environment and Communications References Committee’s inquiry into artificial intelligence and data centres. This is a short summary of what I argued, in my own words. The views are mine; they are not the committee’s.
Status: the committee secretariat has confirmed my submission was received and is being processed. The committee has not yet decided whether to publish it. If it is published, I will link to it here. Until then, this page summarises it; it does not reproduce it.
I am not against AI, and I am not against data centres. My work is in search and AI, and Australia should want this industry. My argument is about terms. A very large AI data centre is a power station’s worth of demand, a water user and a piece of national digital infrastructure all at once. Right now it is mostly approved like an ordinary development, one site at a time, under rules that differ from state to state. That leaves room for the costs to land on households, farmers and communities while the benefits flow elsewhere.
The principle I put to the committee is simple: whoever creates the new demand should carry the new cost.
The Australian Energy Market Operator’s 2026 Electricity Statement of Opportunities forecasts data-centre electricity use in the National Electricity Market rising from about 5 TWh in 2025–26 to around 34 TWh by 2035–36, roughly 3% to 13% of grid supply. AEMO also flags real uncertainty about how fast that growth arrives. That combination, big and uncertain, is exactly where clear national rules earn their keep.
Energy that is genuinely new. New AI data centres should bring their own additional renewable generation, firming or storage, verified properly, rather than buying certificates from projects that would have been built anyway. Ireland already requires something similar of new data centres (Commission for Regulation of Utilities, 2025).
No cross-subsidy from households. Proponents should pay the full cost of the grid connections and network upgrades their projects cause, with the cost-allocation method published, and large sites should be able to cut demand when the grid is under stress.
Water with a hierarchy. Recycled and non-drinking water first, the lowest-water cooling that is practical, a site-specific water budget, and assessment at the level of the whole catchment, using climate projections rather than past rainfall alone.
Cumulative impact, not just single projects. Three facilities in one region can each look acceptable on paper and still overload the local grid, water, roads and housing together. Planning should assess them together, and include heat, noise, biodiversity, decommissioning and end-of-life e-waste.
Communities and First Nations, early. Plain-English information before a site is locked in, independent technical advice paid for by the proponent, and a national standard for engaging Traditional Owners as rights-holders before decisions are made, not after.
A public national register. Ordinary Australians should be able to see each large facility’s electricity and water use, emissions, cooling method, on-site fuel and battery storage, and grid requirements. Voluntary reporting is not enough when public infrastructure is at stake.
Safety and supply chains. National fire and life-safety standards for on-site lithium-ion batteries and diesel, worked out with local fire services; workplace safety rules suited to data-centre hazards; and human-rights due diligence for the minerals and chips the hardware depends on.
Distinctly Australian protections. A spectrum-interference check for any facility proposed near Western Australia’s radio-quiet zone, which protects the Murchison Radio-astronomy Observatory and the SKA-Low telescope. I also asked that regulators look at how exposed default superannuation funds are to AI infrastructure assets, so that risk is visible to members.
A server can sit in Sydney and still be controlled by a company that answers to another country’s laws. So the real questions are who controls the data, who holds the encryption keys, and who can be compelled to hand it over. I recommended that Australians’ privacy protections travel with their information wherever it is processed, with transfer-risk assessments for sensitive data, Australian-controlled encryption keys for the most sensitive categories, and acceptance of Australian regulatory jurisdiction as a condition of operating strategically significant facilities here. The European Union, the United Kingdom and Singapore all offer workable models.
There is a newer layer to sovereignty too, close to my own research. As people move from search engines to AI assistants, the rules about which Australian websites those systems can read and cite become a kind of infrastructure in their own right. Australia should keep independent capability to measure that, so local businesses and public information do not quietly disappear from the systems people now ask.
The federal Government has already set out expectations for data centres: support the energy transition, use water responsibly, pay a fair share of grid costs, invest in Australian skills. My core ask is that those expectations become enforceable, measurable and audited, through a national standard applied in proportion to a facility’s size, with approvals conditional on compliance rather than promises.
Australia has real advantages for AI infrastructure: renewable resources, skilled people, stable institutions. We should use them, on terms that keep Australian interests, environmental limits and community rights at the centre.
Doug Lord (Douglas Lord) is an Australian search and AI-visibility researcher, founder of Digital Dominator, and creator of the Periodic Table of Digital Authority. He is also co-founder of the Brisbane rock band Moonjuice. This article summarises his own submission and does not represent the views of the committee or the Parliament.