Australian researchers are preparing to launch a project designed to test how legal controls could be incorporated into autonomous AI systems. The “Autonomy under Law” project brings together legal, scientific and engineering expertise to explore what it calls a legal harness architecture for AI.
Rather than considering AI governance purely through policies and regulation, the project will experiment with technical mechanisms intended to reduce AI harms and produce evidence that could inform Australian state and national policy.
Key facts
- The project is led by researchers and organisations including the University of Western Australia and Curtin University.
- It will examine legal safeguards for autonomous AI.
- Researchers intend to test risks through controlled experiments.
- The work combines legal, scientific and engineering expertise.
- Findings are intended to help inform Australian AI policy.
- The formal project launch is scheduled for 3 September.
Our take
This is worth watching from New Zealand because it tackles one of the harder AI governance problems: translating principles and legal requirements into controls that actually constrain autonomous systems.
As agentic AI becomes capable of taking actions rather than simply generating content, governance will increasingly need technical enforcement alongside policies and human oversight. Australian experiments in this area could provide useful evidence for New Zealand regulators and organisations developing their own agentic AI controls.
Sources
KiwiGen.AI 