Submission 427
Mr Isaac Dugdale
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.
The author, giving a lay judgement about AI extinction risk, proposes rapid data-centre construction without harm or environmental damage to gain leverage over frontier development. This rests on an understanding that chip supply constrains global capacity, so Australian construction substitutes for overseas capacity. The submission proposes compute conditions, a collective training licence, binding infrastructure obligations, diplomacy and AI Safety Institute reform.
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.
Argues that loss of control of superintelligent AI could cause human extinction within a decade and that Australia can reduce the chance through global cooperation to pace frontier development.
The loss of control of superintelligent AI could cause human extinction within the next decade. Australia can reduce the chance of this happening by being involved in global cooperation efforts to pace the frontier of development.
Argues for rapid data-centre construction to reduce existential risk and gain economic benefits, conditional on preventing harm or environmental damage.
To reduce existential risk and capture economic benefits, Australia should prioritise building data centres as quickly as possible, while maintaining social license by ensuring their building does not cause harm or environmental damage.
Supports the government's rejection of a text and data mining exception and proposes a collective licence for AI training available only for training done in Australia.
Legislate a collective licence for AI training, and make it available only for training done in Australia. The government ruled out a text and data mining exception in October 2025, which I support.
Argues that data centre ownership and end use should be public before approval, with guaranteed domestic access to a share of compute.
Ownership and end use of every large data centre proposal should be public before approval. The six confidential Tasmanian proposals, and the Firmus irrigation water episode, show what the current settings allow. Australia should also reserve guaranteed domestic access to a share of the compute built here, for public services, research and the Safety Institute.
Says the AI Safety Institute has $29.9 million over four years against roughly $120 million a year for the UK equivalent, and advises rather than regulates.
This only works if the AI Safety Institute can do the work. It has $29.9 million over four years, including $3.5 million this financial year, against roughly $120 million a year for the UK equivalent. It also advises rather than regulates, and sits inside a department.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Prioritise building data centres as quickly as possible while maintaining social licence by ensuring they do not cause harm or environmental damage.
To reduce existential risk and capture economic benefits, Australia should prioritise building data centres as quickly as possible, while maintaining social license by ensuring their building does not cause harm or environmental damage.
Condition Australian frontier-model training/serving approval on incident/dangerous-capability reporting to the AI Safety Institute, Institute pre-deployment evaluation access, and commitment to pace frontier development in line with the July 2026 industry statement.
Approval to train or serve frontier models on Australian land and power should carry three obligations: Report safety incidents and dangerous capability evaluations to the Australian AI Safety Institute. Give the Institute pre-deployment access to evaluate models trained or served here. Commit to pacing frontier development, in line with the July 2026 industry statement.
Legislate a collective licence for AI training, available only for training done in Australia.
Legislate a collective licence for AI training, and make it available only for training done in Australia.
Legislate declared collecting-society licensing of AI training using existing Australian licensing machinery, alongside the AI standards bill in early 2027.
I'd propose the following, legislated alongside the AI standards bill in early 2027: A declared collecting society licenses AI training use, using the machinery that already administers statutory and voluntary licences in Australia.
Let rightsholders opt out of the collective licence and negotiate directly.
Rightsholders can opt out and negotiate directly, which preserves the artist control the Prime Minister promised.
Have the Copyright Tribunal set the licence rate where the parties cannot agree.
The Copyright Tribunal sets the rate where the parties can't agree, so a deadlock can't stall the scheme for years.
Require licensees to declare their training corpora so payment reaches the right people and artists can check whether their work was used.
Licensees must declare their training corpora, so money reaches the right people and artists can check whether their work was used.
Place First Nations cultural material outside the general licence, requiring community permission.
First Nations cultural material sits outside the general licence and requires community permission, which several Aboriginal artists have asked for and which a blanket licence would not deliver.
Legislate the announced data centre obligations as binding: underwriting new power supply, paying full grid connection costs, being net generators rather than net users, and minimising water use.
The data centre obligations announced on 15 July should be legislated as binding rather than left as expectations: underwriting new power supply, paying full grid connection costs, being net generators rather than net users, and minimising water use.
Make ownership and end use of every large data centre proposal public before approval.
Ownership and end use of every large data centre proposal should be public before approval.
Reserve guaranteed domestic access to a share of the compute built in Australia, for public services, research and the AI Safety Institute.
Australia should also reserve guaranteed domestic access to a share of the compute built here, for public services, research and the Safety Institute.
Back international frontier-development pacing through Five Eyes, relationships with the United States and China, and regional partners.
Australia should back international work on pacing frontier development, through the Five Eyes relationships, our standing with both the United States and China, and our regional partners.
Look at both the funding and the structure of the AI Safety Institute, which has $29.9 million over four years, advises rather than regulates, and sits inside a department.
This only works if the AI Safety Institute can do the work. It has $29.9 million over four years, including $3.5 million this financial year, against roughly $120 million a year for the UK equivalent. It also advises rather than regulates, and sits inside a department. I'd ask the committee to look at both the funding and the structure.