Submission 441
Miss Michele Bray
Submitter group: Individuals.
Position labels are not shown for individual submitters.
An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.
Summary
Read the original submission for the full text.
Argues that existing laws are inadequate for frontier AI and that companies should not set their own pace or rules. Cites an expert survey and statements by an AI company chief executive, and recommends legislated independent verification, mandatory guardrails, a better-resourced AI Safety Institute and consideration of slower rollout.
Key points
Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.
States that a stress test found existing Australian measures inadequate across all five categories assessed, with over 93% of experts rating current measures for loss-of-control risk insufficient.
Good Ancestors' AI Legislation Stress Test surveyed 64 experts spanning AI, public policy, cybersecurity, national security and law, assessing Australia's laws against five categories of AI threat. It found existing measures inadequate across all five. For loss-of-control risk specifically, over 93% of experts rated current measures insufficient, the least adequately covered of the five categories assessed.
Argues that the stress test it cites is an expert-led answer to the Commission's call for a gap analysis and points toward stronger legislated measures, not further delay.
In substance, that gap analysis already exists: the stress test above is a direct, Australian, expert-led answer to the question the Commission asked, and its finding points toward stronger legislated measures, not toward further delay.
Argues that frontier developers are not in a position to set the terms of their own safety and that self-regulation is a conflict of interest.
An industry racing itself is not in a position to set the terms of its own safety, and self-regulation in this context is a conflict of interest, not a substitute for governance.
Argues that the pace and standards of frontier AI development should be set and verified by government and independent institutions.
It follows that the pace of frontier AI development, and the standards it is held to, should be set and verified by government and independent institutions, not left to voluntary commitments from the labs themselves.
Argues that the Australian AI Safety Institute is not comparably resourced to the UK body, which it says has £66 million a year and over 100 technical staff.
The UK's equivalent body operates on £66 million per financial year with over 100 technical staff. Australia's AI Safety Institute was only recently announced, and its current mission clarity and resourcing are not comparable.
Argues that Australia can have a strong, independent say in governing frontier AI without building frontier models itself.
Australia does not need to build frontier AI models itself to have a strong, independent say in how they are governed here, and I would encourage the Committee to treat that precedent as evidence of capability, not as an exception.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Weigh an AI company leader’s case for independent, government-backed oversight as evidence from inside the industry.
In other words, the head of one of the world's leading AI companies is arguing, in his own words, that companies cannot be left to mark their own homework, and that independent, external oversight, backed by government, is required. I ask the Committee to weigh this as evidence from inside the industry, not merely from its critics.
Treat Australia’s under-16 social-media ban as evidence of its capability to govern frontier AI independently without building frontier models.
Australia has shown it can lead on exactly this kind of question before, most recently with the world-first ban on social media access for under-16s. That policy was widely doubted, on the grounds that it could not be enforced or that Australia was too small a market to influence global platforms, and a number of other countries are now examining it as a model. We did not wait for the industry's permission on that one. Australia does not need to build frontier AI models itself to have a strong, independent say in how they are governed here, and I would encourage the Committee to treat that precedent as evidence of capability, not as an exception.
Legislate requirements for independent external verification of frontier AI systems before and after deployment.
That the Committee recommend legislated requirements for independent, external verification of frontier AI systems, before and after deployment, rather than relying on voluntary industry commitments.
Treat the Good Ancestors stress test as satisfying the gap analysis and proceed with mandatory guardrails for high-risk AI without further delay.
That the Committee treat the Good Ancestors AI Legislation Stress Test as substantively satisfying the Productivity Commission's call for a gap analysis, and recommend the government proceed with mandatory guardrails for high-risk AI applications without further delay.
Give the Australian AI Safety Institute funding and statutory authority broadly comparable to the UK body (£66 million annually, 100+ technical staff), sufficient for independent evaluation.
That the Committee recommend funding and statutory authority for the Australian AI Safety Institute at a level broadly comparable to the UK AI Security Institute (£66 million annually, 100+ technical staff), sufficient to perform an independent evaluation role.
Treat the pace of AI capability development as a regulatory question in its own right, including when government should be able to require a slower rollout pending adequate oversight.
That the Committee treat the pace of AI capability development as a live regulatory question in its own right, including the circumstances in which government should be able to require a slower rollout pending adequate oversight.