AI Inquiry Atlas

An unofficial, independent guide to the public records of the Joint Select Committee on Artificial Intelligence. Not connected with the Parliament of Australia or the committee. Its makers also made a submission (no. 481; see About). Summaries and scores are written by AI and can be wrong: check the original before relying on them.

Submission 483

Australian Academy of Science

Submitter group: Research institutes.

An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.

Open in the interactive guide

Summary

Read the original submission for the full text.

Argues that Australia should make strategic investments in sovereign AI capability, particularly high-performance computing, research-sector AI uptake, AI skills, research data stewardship and attribution. Makes eleven recommendations, including a 10-year national computing strategy proposed at $200 million per year, and enforceable data centre expectations and AI standards.

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.

  1. Argues that Australia faces escalating risk by relying on offshore AI capability, and that sovereign capability is critical to manage AI development, regulation and adoption in the national interest.

    Australia faces an escalating risk by relying on offshore AI capability. Sovereign AI capability is critical to ensure that Australia has the domestic ability to manage the development, regulation, and adoption of AI in our national interest.

  2. States that in June Australian researchers and companies lost access to leading AI models under US export controls, briefly, and that the duration of any next occasion is difficult to predict.

    In June, Australian researchers and companies lost access to the world's most advanced AI models when the United States imposed export controls restricting their use to US nationals. This left Australia without access to frontier AI capability. That occasion was brief, but it is difficult to predict the duration of the next one.

  3. Says the rapid uptake of AI by Australia's scientific workforce is transforming the practice of science and could accelerate discovery and innovation.

    The rapid uptake of AI by Australia’s scientific workforce is transforming the practice of science and has the potential to accelerate discovery and innovation.

  4. Says both of Australia's Tier-1 computing facilities are oversubscribed and proposes $200 million per year over 10 years for sector planning, upgrades and Tier-0 investment.

    Both of Australia’s Tier-1 facilities are oversubscribed and demand is anticipated to increase. The Academy proposes an investment of $200 million per year over 10 years to support sector planning, upgrade facilities, and coordinate investment in Tier-0 capability.

  5. Argues that data sovereignty depends on effective control over who can access and use data, and that onshore storage alone may not provide sovereignty.

    Data sovereignty depends on maintaining effective control over who can access, process, combine, disclose and reuse the data, including its use for AI model training, fine-tuning and inference. Onshore storage alone may not provide sovereignty if these governance and control arrangements are weak.

  6. Says the adoption of AI is currently impeded by a shortage of AI skills in the workforce and insufficient understanding of AI across sectors.

    The adoption of AI is currently impeded due to a shortage of AI skills in the Australian workforce and insufficient understanding of AI across sectors.

Recommendations

Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.

  1. Invest in national AI capability to build resilience and shape the development and adoption of AI technologies.

    Australia must invest in national AI capability to build resilience and shape development and adoption of AI technologies, as other countries have done.

  2. Have the Office for AI, National AI Centre and DISR's Technology and Digital Policy division support national coordination of AI uptake in research as a priority.

    The Office for AI, National AI Centre and the Department of Industry, Science and Resources Technology and Digital Policy division should support national coordination of uptake of AI in research as a priority.

  3. Develop a 10-year national high-performance computing strategy and investment plan for AI/data-intensive research; invest $200 million annually over 10 years in sector planning, facility upgrades and coordinated Tier-0 (regional) capability.

    The Australian Government should develop a 10-year national high-performance computing strategy and investment plan to secure computing capability for AI and data-intensive research. The Academy proposes an investment of $200 million per year over 10 years to support sector planning, upgrade facilities, and coordinate investment in Tier-0 (regional) capability.

  4. Upgrade and expand NCI and Pawsey to meet growing data-intensive research needs.

    Australia’s national (Tier-1) HPC facilities (NCI and Pawsey) need to be upgraded and expanded to ensure they are equipped to meet the growing demands of data-intensive research needs.

  5. Make dedicated investment to scale GPU capacity for AI-driven research.

    Dedicated investment is needed to scale up GPU capacity for AI-driven research.

  6. Include planning for AI and high-performance computing needs, particularly GPUs, in the 2026 National Research Infrastructure Roadmap.

    The 2026 National Research Infrastructure Roadmap should include planning for AI and high-performance computing needs, particularly GPUs.

  7. Ensure diversification of onshore large-scale computing infrastructure and partnerships with other nations to secure access to frontier models and high-performance computing amid geopolitical uncertainty.

    Australia needs to ensure diversification of onshore large-scale computing infrastructure and partnerships with other nations to secure access to frontier models and high-performance computing amid geopolitical uncertainty.

  8. Invest in addressing the AI skills gap through measures to uplift AI literacy within Australia's research workforce.

    The Australian Government should invest in addressing Australia’s AI skills gap by implementing measures to uplift AI literacy within Australia’s research workforce.

  9. Develop the national AI talent pipeline and support training to improve AI literacy and skills across the existing research workforce.

    Australia needs to develop the national AI talent pipeline, and also needs to support AI training for researchers to improve AI literacy and skills across the existing research workforce.

  10. Consider researcher-literacy education and funding options, including Academy teaching resources, ATSE’s STELR program, and workforce/diversity/equity programs such as Superstars of STEM targeted towards computing or AI researchers.

    Uplifting researcher AI literacy could be achieved through education initiatives or funding for programs that support Australia’s scientific workforce. This may be through: • Development of teaching resources (e.g. the Australian Academy of Science’s education resources and the Australian Academy of Technological Sciences and Engineering STELR program) • Support for workforce, diversity and equity programs (e.g. Science & Technology Australia’s Superstars of STEM program – this could be targeted towards supporting researchers working in computing and/or AI)

  11. Consider scholarships/cadetships, organisational partnerships, industry PhD funding and researcher data-literacy programs, potentially delivered with the National AI Centre or Office of AI to target greatest need.

    • Scholarships and cadetships • Partnerships with organisations • Funding industry PhDs, and • Data literacy programs for researchers. These could be delivered in partnership with the National AI Centre or Office of AI to target the areas of greatest need.

  12. Retain visibility over what research data types commercial data centres store, with agreements guaranteeing safe data stewardship and appropriate custodianship, rights and control over research data.

    Government should retain visibility over what research data types are stored by commercial data centres, with agreements guaranteeing safe data stewardship and appropriate custodianship, rights and control over research data.

  13. Invest in privacy-preserving and secure data-sharing infrastructure, identity/access controls, provenance and auditable governance.

    investment is also needed in privacy-preserving and secure data-sharing infrastructure, identity and access controls, provenance, and auditable governance.

  14. Have government and commercial arrangements for AI infrastructure uphold Indigenous data governance authority.

    Government and commercial arrangements for AI infrastructure should therefore uphold Indigenous data governance authority.

  15. Require data centre and AI developers to contribute to trusted public data and knowledge infrastructure such as scientific data repositories, metadata standards, provenance mechanisms and long-term stewardship.

    Data centre and AI developers should be required to contribute to trusted public data and knowledge infrastructure, including scientific data repositories, metadata standards, provenance mechanisms and long-term stewardship of publicly accessible information resources that underpin AI development and scientific discovery.

  16. Where AI developers benefit from publicly funded data/knowledge, ensure contributions sustain those resources, including repositories, provenance, attribution, licensing and responsible-access frameworks.

    Where AI developers benefit from publicly funded data and knowledge resources, policy settings for AI and data centres should ensure that they contribute to the long-term sustaining of those resources, including through support for data repositories, provenance, attribution, licensing, and responsible access frameworks.

  17. Invest in open science practices that enable high-quality, well-governed data for AI and AI-driven research, including curation of nationally significant datasets such as research collections.

    Investment is required to support open science practices that enable high-quality, well-governed, well-described and fit-for-purpose data for developing AI systems and AI-driven research, including curation of nationally significant datasets such as Australia’s research collections.

  18. Coordinate investment in digitising nationally significant datasets, including Australia’s research collections.

    To maximise the value of data for scientific discovery, including AI-driven research, Australia needs coordinated investment in digitisation of nationally-significant datasets, such as Australia’s research collections.

  19. Implement robust mechanisms to ensure scientific outputs (e.g. research publications and data) are appropriately attributed.

    The Australian Government should implement robust mechanisms to ensure scientific outputs (e.g. research publications and data) are appropriately attributed.

  20. Ask the Committee to consider ownership and use of scientific publications and research datasets alongside creative, cultural and media content.

    The Terms of Reference consider intellectual property and copyright laws in relation to Australian creative, cultural and media content in the training of AI models, but the Committee should also consider ownership of and use of scientific publications and research datasets.

  21. Preserve attribution and provenance for research data used in AI, building on DOI links to datasets/software/analyses; open-access deposition should not diminish originators’ recognition for subsequent use.

    Robust mechanisms are needed to preserve attribution and provenance when research data are used in AI systems. Existing initiatives linking DOIs to datasets, software and analyses provide a useful foundation. Importantly, open-access deposition should not diminish the ability of data originators to receive appropriate recognition for subsequent use of their work.

  22. Make data centre expectations and AI standards enforceable through transparent assessment criteria such as measurable reporting obligations, independent assurance and clear consequences for misaligned actions.

    Australia’s data centre expectations and AI standards must be made enforceable through transparent assessment criteria such as measurable reporting obligations, independent assurance, and clear consequences for misaligned actions.

Read the original

Open in the interactive guide