Matters of Public Importance: Artificial Intelligence
Parliament of Australia
READ
Details
- Date Published
- 16 Sept 2026
- Priority Score
- 4
- Australian
- Yes
- Created
- 17 Sept 2026, 04:01 pm
Description
Search engine for Australian parliamentary documents and media.
Summary
This parliamentary transcript addresses the necessity of quantified risk measurement to enable proportional AI regulation, drawing parallels to the risk-tiering approach of the EU AI Act. The debate highlights the challenge of 'emergent behavior' and 'black-box reasoning' in frontier models, which complicate traditional regulatory foresight. The contribution emphasizes that while AI acts as an amplifier of human creativity, its existential peril lies in the hands of 'naive' or 'villainous' actors, necessitating a shift from reactive policy to an informed, ready-to-act governance posture. This discourse is highly relevant to Australia's ongoing development of domestic AI safety frameworks and its alignment with international governance standards.
Body
Mr FARLEY
(Farrer) (15:59): AI represents creative destruction at an unprecedented scale and speed. To quote the late Richard Pratt:
You don't know what you don't know 'til you know it.
Quantifying AI turns a debate that is currently running on speculation, antidote and vibes into one the government can act on. A few reasons this matters specifically draws on what's already surfaced in the AI governance conversation. You can't regulate proportionally without measuring risk and autonomy. The emerging governance consensus is that oversight should scale with an AI systems autonomy and the stakes of its actions and not be applied uniformly. That principle is unworkable without a quantified way to classify systems by risk level, which is exactly what the EU AI Act attempts—unacceptable-, high-, limited-risk tiers, each triggering different obligations. No tiering system, no proportional regulation. You're left with either overregulating everything or underregulating the system that matters the most.
The caveat worth naming honestly—quantification has limits. AI's own opacity makes it worse. Emergent behaviour and black-box reasonings mean some of these risks generally resist being reduced to a number in the advance. The goal isn't to pretend AI can be fully measured before it's regulated, but to close the gap between 'we're speculating' and 'we're measuring what we can and building institutions capable of updating that measurement as the technology changes'. That is really the difference between the reactive regulation and the informed, ready-to-act posture the 'urgent and ambitious' argument depends on.
AI is not an artificial creator. It's an amplifier of human creativity. AI is creative destruction on steroids. It creates new industries, new opportunities and new productivity gains whilst simultaneously disrupting existing job skills and business models. The promise of AI is in the hands and the security of science and industry. The peril of the nation is in the hands of the naive, the arrogant, the villains and the government.