'The extraordinarily high energy requirements of AI data centres will mean that productive agricultural land will increasingly be used for renewable energy projects with potentially serious consequences for Australia’s food security.'To mitigate this, the Bill establishes a moratorium on the construction or major expansion of AI data centres. This moratorium can only be lifted once Parliament passes legislation addressing several 'essential' matters, including:
The bill is a necessary application of the precautionary principle to a technology with non-linear and potentially catastrophic risks. By mandating a 'kill switch,' the state ensures that human agency remains the ultimate arbiter of machine behavior, preventing 'loss-of-control scenarios' [Explanatory Memo page 3]. Furthermore, the moratorium on data centres recognizes that digital progress cannot come at the expense of physical survival; protecting agricultural land and water security is a utilitarian necessity for national food security. Finally, the requirement for community consent empowers local populations against the top-down imposition of massive infrastructure, strengthening democratic legitimacy and ensuring that the benefits of AI are balanced against the interests of local communities [Judgment].
The bill risks making Australia a technological backwater by creating immense regulatory uncertainty that will likely deter the billions in capital investment required for modern computing infrastructure. From an epistemic standpoint, the concept of a 'kill switch' for advanced, distributed AI may be a technical fallacy, providing a false sense of security while imposing heavy compliance costs.[1] Additionally, the bill infringes on propertarian rights by allowing the state to halt development on private land and grants the Minister broad, potentially arbitrary emergency powers to interfere with private enterprise without sufficient judicial oversight [Judgment].
The technical feasibility of a 'kill switch' for large-scale, distributed, or decentralized AI models remains a subject of significant debate among computer scientists, with many arguing that such a mechanism could be bypassed or rendered ineffective by the system's own architecture.
2026-09-07
House of Representatives
Before House of Representatives
GEE, Andrew, MP
Unspecified
Science / Technology, Agriculture, Climate Change / Environment