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 286

Centre for AI, Trust and Governance at the University of Sydney

Submitter group: Universities.

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.

Responds to each term of reference, arguing for public sector AI testbeds, stronger competition and interoperability, extension of the Security of Critical Infrastructure Act to digital and AI infrastructure, targeted rather than wholesale regulatory reform, a distinction between individual and ecosystem-level information harms, equitable data centre planning, and a Global AI and Digital Trust Index.

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. Proposes that the Office of AI work with researchers on public sector AI testbeds, such as public sector LLMs, to test responsible uses.

    We propose that the Office of AI in the Department of Prime Minster and Cabinet work closely with researchers in developing AI “testbeds” such as public sector LLMs through which ethical and responsible uses of AI tools and data can be implemented, tested and evaluated

  2. Argues that openness and data-sharing policies alone will not promote Australian innovation without attention to competition and interoperability.

    Policies premised on openness and data-sharing alone will not, by themselves, promote Australian innovation unless matched by attention to competition, interoperability and proactive measures to promote innovation and restrict the emergence of market monopolies

  3. States that content delivery networks and AI inference services sit outside the explicit scope of the SOCI Act.

    Content delivery networks and AI inference services sit outside the explicit scope of the Security of Critical Infrastructure Act 2018 (SOCI Act).

  4. Says even proponents of copyright reform rarely argue for unconditional scraping and training rights, preferring exceptions with fairness criteria or lawful-access conditions.

    Notably, even proponents of legislative reform rarely argue for an unconditional right to scrape and train. Preferred exceptions are almost always framed around fairness criteria or lawful-access conditions, which narrows the effective range of live options considerably.

  5. Argues that AI-enabled harms should be split between individually harmful content, a platform responsibility, and ecosystem-level information degradation, a shared responsibility.

    distinguish individually harmful content as a platform responsibility amenable to content moderation, from ecosystem-level degradation of information quality as a shared responsibility engaging platforms, government, media and civil society

  6. Says it supports Dr Rebecca Johnson's call for sovereign evaluation capability: independently producing, obtaining, interrogating and contesting evidence about AI systems Australia depends on but did not build.

    This submission supports the call by independent researcher Dr. Rebecca Johnson for Australia to develop a "sovereign evaluation capability". This refers to the capacity to independently produce, obtain, interrogate and contest evidence about AI systems Australia depends on but did not build.

Recommendations

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

  1. Have the Office of AI (PM&C) work with researchers to develop public sector AI testbeds, such as public sector LLMs, to implement, test and evaluate responsible AI uses.

    We propose that the Office of AI in the Department of Prime Minster and Cabinet work closely with researchers in developing AI “testbeds” such as public sector LLMs through which ethical and responsible uses of AI tools and data can be implemented, tested and evaluated

  2. Develop media and cultural industry investment and robust information/cultural policies alongside regulation of online and AI-generated content.

    This means that investment in media and cultural content industries and robust information and cultural policies need to be developed in tandem with measures to regulate online and AI-generated content

  3. Consider the South Korean AI for All initiative as an example of an equitable and inclusive AI strategy that is also cognisant of sovereign capability.

    The South Korean AI for All initiative provides an important example for Australian policy makers to consider for what equitable and inclusive AI strategies that are also cognisant of sovereign capability can look like

  4. Integrate AI infrastructure into metropolitan and precinct planning from the outset, following NSW's Connecting with Country framework and meaningfully engaging affected communities and Traditional Owners.

    A particular opportunity cost which needs to be addressed is that land allocated to data centres may constrain housing delivery or other public priorities unless planning frameworks integrate AI infrastructure at the metropolitan and precinct level from the outset, in line with the NSW Government’s Connecting with Country framework and with meaningful engagement with affected communities and Traditional Owners

  5. Sustain multidisciplinary research investment for comparative-advantage sectors such as financial services, health/medical research and advanced manufacturing.

    Australia has a number of sectors where it has comparative advantage. These include financial services, health and medical research, and advanced manufacturing. Realising AI’s potential in these sectors requires sustained, multidisciplinary research investment.

  6. Develop sovereign evaluation capability to independently produce, obtain, interrogate and contest evidence about AI systems Australia depends on but did not build.

    This submission supports the call by independent researcher Dr. Rebecca Johnson for Australia to develop a "sovereign evaluation capability". This refers to the capacity to independently produce, obtain, interrogate and contest evidence about AI systems Australia depends on but did not build.

  7. Consider embedded independent expertise in the Office of AI to oversee evidence requirements, observability of high-consequence and agentic systems, and material-change re-evaluation triggers.

    Such evidentiary requirements, observability for high-consequence and agentic systems, and material-change triggers for re- evaluation, could be overseen through embedded independent expertise in the Office of AI in Department of Prime Minster and Cabinet

  8. Invest in human-centred design, creative experimentation, critical thinking and aesthetic literacy across education and professional development.

    This requires sustained investment in human- centred design, creative experimentation, critical thinking and aesthetic literacy across education and professional development

  9. Apply public-sector transparency reporting covering independent monitoring, evolving policies/interventions, harmful-outcome risk assessments, data quality and researcher access, including granular data through regulated APIs.

    Reporting should cover platform (or, by direct analogy, agency) contributions to independent monitoring, changes to policies and interventions in response to evolving conditions. Reporting must include risk assessment processes for identifying harmful outcomes, and data quality and accessibility for independent researchers. Specifically, this includes access to granular data via regulated APIs enabling independent qualitative and quantitative analysis

  10. Transfer resource-use/performance reporting, inclusive governance and consultation models to Commonwealth AI use.

    The University of Sydney’s submission on data centre governance likewise recommends public reporting frameworks for resource use and performance. As well as inclusive governance and consultation models, as mechanisms for maintaining public trust in large-scale digital infrastructure (University of Sydney, 2026). Both models are directly transferable to Commonwealth AI use.

  11. Establish a dedicated, consent-based regime for Indigenous Cultural and Intellectual Property, addressing culturally inauthentic AI material produced without free, prior and informed consent.

    It is apparent to all stakeholders that Indigenous Cultural and Intellectual Property (ICIP) is inadequately protected by conventional copyright, which is individually owned, time-limited, and protects only expression rather than underlying knowledge or cultural motifs. AI's capacity to generate culturally inauthentic material without free, prior and informed consent is a distinct harm both to cultural rights and to opportunities for First Nations creatives, and requiring a dedicated, consent-based regime.

  12. Provide real-time cross-agency visibility into emerging AI risks for the AI Safety Institute, Cyber and Infrastructure Security Centre and intelligence community, beyond periodic audits.

    A comparable model of real-time, cross-agency visibility (rather than periodic compliance audits alone) would strengthen the ability of the Australian AI Safety Institute, the Cyber and Infrastructure Security Centre and the broader intelligence community to identify and respond to emerging AI-enabled risks. This includes AI- enabled cyber intrusions and AI-generated disinformation associated with foreign interference.

  13. Have the Office of AI work with independent researchers to develop a Global AI and Digital Trust Index tracking Australians' trust in AI and digital technologies over time and benchmarking against international evidence.

    We would propose that the Office of AI work with independent researchers in the development of a Global AI and Digital Trust Index that can reliably and independently identify changes in Australians’ attitudes to trust in AI and other digital technologies over time and can benchmark these against comparable international evidence.

  14. Frame national strategy and public-sector adoption around evidence of durable adoption and equitable productivity benefits, including explicability, trust, data quality, interoperability and workforce capability.

    National AI strategy and public sector AI adoption needs to be framed around clear evidence of what actually drives durable adoption and equitably distributed productivity benefits. This includes explicability of processes and outcomes, trust, data quality, interoperability and workforce capability. There are opportunities public sector agencies to act as “testbeds” for safer AI with transparent risk-assessment processes, data provenance and assessment grounded in independent research access, including the development of public sector LLMs.

  15. Use public-sector testbeds with transparent risk assessments, data provenance, independent research access and public-sector LLMs.

    There are opportunities public sector agencies to act as “testbeds” for safer AI with transparent risk-assessment processes, data provenance and assessment grounded in independent research access, including the development of public sector LLMs.

  16. Strengthen competition, data portability and interoperability in AI markets as a precondition for SME and family-business adoption, recognising the differential compliance burden of any AI regulatory model on smaller firms.

    Policy settings should act to strengthen competition, data portability and interoperability in AI markets as a precondition for SME and family-business adoption, with recognition of the differential compliance obligations of any AI regulatory model on smaller firms.

  17. Develop a national sovereign AI compute strategy and support fine-tuning of open-weight models.

    Australia should also develop a national sovereign AI compute strategy and support fine-tuning of open-weight models in order to advance AI agency and sovereign evaluation capability.

  18. Give the Security of Critical Infrastructure Act 2018 explicit digital-infrastructure coverage, naming content delivery networks, AI services and managed service providers, following the EU NIS2 model.

    Give the Security of Critical Infrastructure Act 2018 explicit digital-infrastructure coverage. This includes naming content delivery networks, AI services and managed service providers, following the EU NIS2 model.

  19. Define critical infrastructure assets for the space technology sector, including ground-segment security obligations.

    Define critical infrastructure assets for the space technology sector, including ground- segment security obligations.

  20. Give the AI Safety Institute and the Cyber and Infrastructure Security Centre express statutory information-gathering and coordination powers, resourced comparably to peer regulators.

    This will also involve giving the Australian AI Safety Institute and the Cyber and Infrastructure Security Centre express statutory information-gathering and coordination powers, resourced on a footing comparable to peer regulators and international frameworks.

  21. Commission a coordinated gap analysis across privacy, critical infrastructure, copyright and content-standards regimes as they apply to AI, rather than pursuing wholesale reinvention.

    Resolve targeted regulatory and copyright gaps rather than pursuing wholesale reinvention. Commission a coordinated gap analysis across privacy, critical infrastructure, copyright and content-standards regimes as they apply to AI.

  22. Resource the OAIC commensurate with ASIC and the ACCC if outcomes-based regulatory models are to be relied upon.

    Resource the Office of the Australian Information Commissioner commensurate with ASIC and the ACCC if outcomes- based regulatory models are to be relied upon.

  23. Give close attention to the rights of First Nations creatives, and alignment with UNDRIP, in any revision of copyright laws.

    Any proposed revisions to Australian copyright laws give close attention to the rights of First Nations creatives, and alignment with the United Nations Declaration on the Rights of Indigenous Peoples (UNDRIP)

  24. Ground AI-harms policy in public epistemic rights, distinguishing individually harmful content as platforms' moderation responsibility from ecosystem-level information disorder as a shared platform, government, media and civil-society responsibility.

    Distinguish, in AI-related harms policy, between individually harmful content (a platform content-moderation responsibility) and ecosystem-level information disorder (a shared responsibility across platforms, government, media and civil society), grounded in the public’s epistemic rights.

  25. Give the Australian Electoral Commission legislative, not merely advisory, powers to address AI-generated deepfakes in election campaigns.

    Measures should be undertaken to empower the Australian Electoral Commission legislative, not merely advisory, powers to address AI-generated deepfakes in election campaigns.

  26. Integrate AI infrastructure development into metropolitan and regional strategic planning, with mandatory cumulative-impact assessment.

    We support measures to integrate AI infrastructure development into metropolitan and regional strategic planning, with mandatory cumulative- impact assessment.

  27. Support consortium/public–private models such as waste-heat reuse and shared energy infrastructure, with genuine community and Traditional Owner engagement and direct benefits for hosting communities.

    We also support consortium and public–private partnership models (for example, waste-heat reuse and shared energy infrastructure) so that hosting communities, alongside genuinely engaged local communities and Traditional Owners and capture direct benefit.

  28. Build critical, creative and security-aware capacities into education and training alongside technical AI literacy, with skills development for AI incorporating sovereign evaluation capability and skills in AI safety, trust and ethics.

    Build critical, creative and security-aware capacities into education and training alongside technical AI literacy, Skills development for AI needs to incorporate sovereign evaluation capability as well as skills and capabilities in AI safety trust and ethics.

  29. Mandate real-time, location-based carbon accounting and transparent reporting.

    Mandate realtime, Require water-efficient Define certification Support integrated 3 Fa RT Cea location-based carbon Cooling, reuse, and security, and sovereignty enenyy-cata benter 5 transmission, and accounting and next-generation thermal requirements for business models and g equipment upgrade transparent reporting technologies in strategic AI coordinated policy and The University of Sydney ‘standards for AI loads. approvals infrastructure. approval pathways

    A program corrected obvious text-recognition errors in this quote for display.

  30. Require water-efficient cooling, reuse and next-generation thermal technologies in approvals.

    Mandate realtime, Require water-efficient Define certification Support integrated 3 Fa RT Cea location-based carbon Cooling, reuse, and security, and sovereignty enenyy-cata benter 5 transmission, and accounting and next-generation thermal requirements for business models and g equipment upgrade transparent reporting technologies in strategic AI coordinated policy and The University of Sydney ‘standards for AI loads. approvals infrastructure. approval pathways

    A program corrected obvious text-recognition errors in this quote for display.

  31. Define certification, security and sovereignty requirements for strategic AI infrastructure.

    Mandate realtime, Require water-efficient Define certification Support integrated 3 Fa RT Cea location-based carbon Cooling, reuse, and security, and sovereignty enenyy-cata benter 5 transmission, and accounting and next-generation thermal requirements for business models and g equipment upgrade transparent reporting technologies in strategic AI coordinated policy and The University of Sydney ‘standards for AI loads. approvals infrastructure. approval pathways

    A program corrected obvious text-recognition errors in this quote for display.

  32. Support integrated business models and coordinated policy and approval pathways.

    Mandate realtime, Require water-efficient Define certification Support integrated 3 Fa RT Cea location-based carbon Cooling, reuse, and security, and sovereignty enenyy-cata benter 5 transmission, and accounting and next-generation thermal requirements for business models and g equipment upgrade transparent reporting technologies in strategic AI coordinated policy and The University of Sydney ‘standards for AI loads. approvals infrastructure. approval pathways

    A program corrected obvious text-recognition errors in this quote for display.

Read the original

In some quotes a program corrected obvious text-recognition errors for display.

Open in the interactive guide