Submission 206
Associate Professor Michael Noetel
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.
Writing personally, the author argues for national-security regulation of frontier AI, a superintelligence-development ban and international agreement, while retaining existing law for deployment. Proposes expanded compute infrastructure and statutory safety powers, leaving the institutional vehicle to government. Says they know of no law covering frontier development. Draws on expert judgments with uncertain percentages and a public-attitudes survey.
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.
Says Delphi risk estimates reflect expert judgment rather than frequency data, with uncertain percentages but risks consistently above tolerable thresholds.
Delphi point estimates aggregate expert judgement rather than frequency data, so the exact percentages are soft; the finding that experts place these risks far above tolerable thresholds is robust across rounds.
Says 94% of Australians expect AI to meet safety standards at least as strict as commercial aviation and 72% want government to prioritise managing risks over driving innovation.
Australians seem to want that gap closed. 94% expect AI to be held to safety standards at least as strict as commercial aviation. 72% want the government to prioritise managing risks over driving innovation.
Argues that Parliament should legislate a prohibition on developing superintelligence in Australia, mirroring bills before the US and UK legislatures.
Parliament should legislate a prohibition on developing superintelligence in Australia, mirroring the bills from the US and UK. It costs nothing, signals intent, and means Australian law is ready when an agreement arrives.
Argues for removing barriers, especially copyright barriers, to data centres, describing compute as Australia’s primary leverage given the cost of reaching the frontier.
Remove barriers to companies building data centres here (esp. copyright barriers). Given the cost of reaching the frontier, compute is the primary leverage Australia will have.
Argues that existing law is largely adequate for how AI is used, with narrow amendments where AI opens a new route to an old harm.
For how AI is used, largely yes. Fraud, scams, deepfakes, discrimination, defamation, privacy breaches and harm to children are already illegal. Where AI opens a new route to an old harm, the fix is usually a narrow amendment.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
State in every relevant forum that Australians should not be exposed against their will to catastrophic risks from decisions made in San Francisco and Beijing.
Government should now state plainly, in every relevant forum, that Australians are on the receiving end of decisions made in San Francisco and Beijing, and that we expect the people making them not to expose us to catastrophic risk against our will.
Adopt a stated position that frontier loss-of-control risk is a national security concern for Australia, and pursue a coalition of middle powers to press for binding international rules.
Government adopt a stated position that loss-of-control risk from frontier AI is a national security concern for Australia, and pursue a coalition of middle powers to press for binding international rules.
Legislate a prohibition on developing superintelligence in Australia, mirroring the US and UK bills.
Parliament should legislate a prohibition on developing superintelligence in Australia, mirroring the bills from the US and UK.
Support an international agreement prohibiting superintelligence development, with compute-based verification.
Recommendation: Parliament legislate a prohibition on the development of superintelligence in Australia, and government support an international agreement to the same effect with compute-based verification.
Consider copyright amendments, dedicated zones, fast permitting and energy policy aimed at tens of gigawatts as likely requirements for enabling data centres.
To enable those data centres, we likely need amendments to copyright law, dedicated zones, fast permitting, and energy policy aimed at tens of gigawatts.
Set a national target for AI compute capacity, establish dedicated data-centre zones with streamlined approvals, and negotiate access guarantees with frontier providers as a condition of hosting.
Government set a national target for AI compute capacity, establish dedicated data-centre zones with streamlined approvals, and negotiate access guarantees with frontier providers as a condition of hosting.
Consider obtaining Australian data-centre water-use figures, recognising that local water conditions differ from US conditions.
I have not found equivalent Australian figures and the Committee may wish to obtain them; the proportions should be similar but the local water context differs.
Promote current AI deployment under existing law, patching gaps narrowly.
Recommendation: Promote deployment of current AI under existing law, patching gaps narrowly.
Give AISI compute-tracking authority and intelligence access; require frontier developers operating in Australia to disclose specifications, internal capabilities and safety cases; protect safety whistleblowers; and establish standing advice to government and diplomatic missions.
Term of reference (m) asks whether the Australian AI Safety Institute can identify and respond to emerging risks. Today it cannot, because it lacks the authority or the funding. It should have: - authority and intelligence access to track AI-relevant compute in Australia and the region; - power to require frontier developers operating here to disclose model specifications, internal capabilities and safety cases; - statutory protection for AI-company employees who raise safety concerns; and - a standing role advising the government and Australia's diplomatic missions.22
A program corrected obvious text-recognition errors in this quote for display.
Leave the choice between AISI and a new body to government, while ensuring a body holds the proposed authority.
Whether the AISI or a new body is the right vehicle is a design question for government; the substance is that some body has this authority.
Read the original
In some quotes a program corrected obvious text-recognition errors for display.