Cambridge centre urges Australia to make AI data and data centre reporting mandatory
The Minderoo Centre for Technology & Democracy at Cambridge (MCTD) has told the Australian Government that transparency should be a legal requirement for large data centres and frontier AI developers. The centre submitted written evidence to the consultation 'Getting it right: Building AI infrastructure that works for Australia'.
- The MCTD at Cambridge submitted written evidence to the Australian Government consultation 'Getting it right: Building AI infrastructure that works for Australia'.
- It says large data centres should report energy use, water use, efficiency and current and future grid demand.
- It cites EU reporting requirements as a model for consistent, comparable data centre reporting.
- It argues frontier AI developers should disclose where training data comes from and how it is used and updated, to the public and regulators.
- The centre says transparency should be a legal requirement, not a voluntary commitment.
Two areas of focus in the submission
According to the MCTD, its evidence concentrates on two matters: reporting requirements for large data centres, and Training Data Transparency rules for AI developers. The centre says transparency is "the precondition for accountability" and for the public trust on which Australia's AI ambitions depend.
The centre also makes a point about international coordination. It argues that the choices Australia makes now will shape, and be shaped by, decisions elsewhere. In its view, interoperable rules in one jurisdiction lower compliance costs and raise standards in others.
What data centres should report
The MCTD argues that large data centres should be required to share basic information on their resource use. Specifically, it lists:
- how much energy and water they use;
- how efficient they are;
- how much demand they place on the electricity grid, now and in the future.
The centre points to reporting requirements introduced in the EU as a model. It says these provide reporting that is consistent and easy to compare across different countries.
Openness about AI training data
On AI models, the MCTD says frontier AI developers should be required to be more open about the data used to build and train their systems. It draws comparisons to the EU and California in making the case.
The centre says companies should provide clear information on where their training data comes from and how it is used and updated. That information should be available both to the public and to regulators. It adds that regulators and independent researchers should be able to access more detailed information when needed.
Legal duty, not voluntary pledge
A central claim of the submission is that transparency should be a legal requirement rather than a voluntary commitment by AI companies. The document does not detail specific legislative text or enforcement mechanisms in the summary published by the centre.
For Australian policymakers, the submission offers a framework that links infrastructure oversight, such as energy, water and grid demand, with oversight of how AI models are trained. For companies, it signals that reporting rules aligned with those in the EU and California could reduce the burden of complying with different regimes in different countries, as the centre sees it. The full evidence and an accessible version are available for download from the MCTD website.
Minderoo Centre for Technology & Democracy: Transparency as the foundation: Written evidence to Australian Government consultation on AI training and large data centres
