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 203

ARC Centre of Excellence for Automated Decision-Making and Society

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 for enforceable public-sector ADM law, sovereignty through local governance control, coordinated legal reform, liability clarification, data-centre regulation and independent AI monitoring. Recommends digital inclusion and productivity evaluation with worker input, and creator support beyond copyright. The authors state that their views do not necessarily reflect ADM+S’s official position.

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. States that Australia governs public sector AI use through voluntary standards and policies that are not legally enforceable, which it says are no replacement for law.

    Australia governs public sector AI use through voluntary standards and ‘mandatory’ policies that are not legally enforceable and cannot be properly monitored: they are no replacement for law.

  2. Argues that Australia can pursue AI sovereignty, understood as local governance control rather than ownership of the whole AI stack, alongside local AI capacity.

    Australia can and should pursue AI sovereignty (understood as local governance control, not ownership of the whole AI stack), and AI capacity (local technology, infrastructure and skills) at a range of levels and through multiple mechanisms.

  3. Says it has recommended clarifying how AI value-chain responsibility passes between actors or is shared, and coordination on co-created risks.

    We have recommended measures to clarify to stakeholders across the AI value chain how responsibility passes between them or is shared, and how to coordinate to address co-created risks.

  4. Argues that Australia can build an independent AI monitoring ecosystem by giving more actors legal protection and data access.

    Australia can build an ecosystem to monitor AI by giving more actors legal protection and data access.

  5. Says affiliated scholars argue copyright is ill-suited to creator concerns, most professional creators do not own it, and individual dataset contributions’ marginal value approaches zero.

    ADMS4S affiliated scholars have shown that copyright is not an apt tool to address creator concerns about AI.?? Most professional artists and creators don’t own copyright (which is frequently owned by intermediaries); the marginal value of any individual’s contribution to a dataset quickly approaches zero

  6. Argues that AI can benefit regions only if the existing digital and emerging AI divide is addressed, citing that 1 in 5 Australians are digitally excluded or highly excluded.

    AI can benefit Australia’s regions only if we proactively address the existing digital and emerging AI divide. 1 in 5 Australians are digitally ‘excluded’/‘highly excluded’

Recommendations

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

  1. Design methods for evaluating AI productivity gains as part of AI adoption, with workers having input into those methods.

    we suggest (a) designing methods for evaluating gains should be part of AI adoption, and (b) workers should have input into methods for evaluating gains and losses

  2. Coordinate national efforts to address digital and AI divides, including literacy programs and recognising internet access as an essential service in planning new developments.

    we recommend coordinated national efforts to address digital and AI divides:® including supporting literacy programs and ensuring that internet access is recognised and included as an essential service when planning new developments.

  3. Establish an effective, up-to-date regulatory framework setting legal baselines for all companies, with greater transparency and record keeping.

    An effective, up-to-date regulatory framework setting the legal baselines for all companies, and greater transparency and record keeping, are all needed

  4. Pursue public AI infrastructure, sovereign sub-frontier and small models, and Australian systems or harnesses with transparency and governability, including on open-weight models.

    Technical tools and investments: public AI infrastructure; sovereign (sub-frontier) and small models; Australian systems/harnesses built with transparency and governability on sub-frontier models including open weights models

  5. Counter lock-in through interoperability and model/provider switching; consider extending infrastructure regulation to long-term price stability and access, minimising unilateral product changes, and transparency.

    Legal mechanisms to counter lock-in by proactively forcing interoperability and the ability to switch (models and providers). Laws that regulate critical infrastructure already exist, and could be extended.” Infrastructure regulation has long ensured privately owned infrastructures are managed in service of public interest, and could include requirements for long-term price stability and access, obligations that minimise the risk of unilaterally changing product configurations, and transparency obligations.

  6. Work on representing Australian knowledge, values and contexts in AI systems.

    Work is needed on representing Australian knowledge, values, and contexts in AI systems.

  7. Develop sovereign capacity to collect and store unique Australian ecological data for conservation AI, respecting traditional owners’ cultural rights.

    ADM+S research with environmental data stewards is also identifying the need for sovereign capacity to collect and store unique Australian ecological/ecosystem data to develop AI capacity for conservation of our unique species and landscapes, and with regard for traditional owners’ cultural rights.*

  8. Develop an exemplary framework for public sector automated decision-making, which the submission says could be tested, assessed and refined over time.

    The opportunity: Australia has sufficient evidence and expertise to develop an exemplary framework which could be tested, assessed and refined over time. Repeated delays in addressing the real risks to Australians from ill- designed or even illegal public sector ADM systems undermine public trust.

  9. Convene current and future Bill-drafting teams publicly across government to explain interactions with unfair-trading law and ongoing relevance to advancing and agentic AI, timetable priorities, and identify government-recognised gaps.

    This Committee could promote the future coherence and effectiveness of Australia’s legal framework for AI by coordinating a public cross-government discussion, bringing together in one room drafting teams responsible for current Bills and those in development to: Explain how these Bills (alongside the new unfair trading practices prohibition) interact; Explain any features of the Bills which, in the view of the drafters, ensure their ongoing relevance as AI technology advances, including agentic AI; 3. Develop a timetable for other priorities: data centre standards; workplace AI; public sector ADM, consumer protection (retail surveillance pricing/agentic commerce);° and 4. Outline any further gaps in the legal framework already identified by the government.

  10. Timetable data-centre standards, workplace AI, public-sector ADM and consumer protection covering retail surveillance pricing and agentic commerce.

    3. Develop a timetable for other priorities: data centre standards; workplace AI; public sector ADM, consumer protection (retail surveillance pricing/agentic commerce);°

  11. Use principles-based Digital Duty of Care replacing Codes/Basic Online Safety Expectations and privacy’s fair-and-reasonable test replacing consent/purpose limitations, provided effective legal protection for Australians remains high.

    We endorse the new Bills’ principles-based approach, i.e. the proposed Digital Duty of Care (and repeal of Codes of Practice and Basic Online Safety Expectations) and the ‘fair and reasonable’ test for privacy (and shift from consent/purpose limitations):3? provided effective legal protection for Australians remains high.

  12. Clarify liability across the AI value chain: how responsibility passes between or is shared among actors, and how to coordinate to address co-created risks.

    We have recommended measures to clarify to stakeholders across the AI value chain how responsibility passes between them or is shared, and how to coordinate to address co-created risks.

  13. Take coordinated legislative action for a consistent national data centre regulatory framework, including community participation and enforcement of the currently non-mandatory Expectations.

    Our research shows the need and opportunity for coordinated legislative action to provide a consistent regulatory framework across Australia, enable community participation®® and implement and enforce the Expectations of Data Centres and AI Infrastructure Developers, which are currently not mandatory.

  14. Give the impending data centre legislation retrospective effect, at least to the date the Commonwealth's expectations were published.

    Given the Commonwealth’s expectations are already public, the impending legislation should have retrospective effect, at least to the date the expectations were published.

  15. Consider creator living wages, data dividends or data-work payments funded by firms exploiting Australian culture, with periodic, stable, ongoing, sufficient compensation and production funding managed by a transparent public entity.

    For creators, more resilient solutions lie outside copyright: such as artist living wages, data dividend, or payment for data work, funded by payments from companies that profit from exploiting Australian cultural outputs, with periodic, stable, ongoing, and sufficient funding to compensate creators and support further production, managed by a transparent public entity.

  16. Address misinformation through digital, media and AI literacy and regulatory transparency measures.

    There are opportunities to address misinformation through efforts to improve digital, media, and AI literacies among audiences, and regulatory transparency measures.

  17. Consider investment in science-informed, responsible AI mental-health support systems, duty-of-care obligations and monitoring.

    With Australians increasingly using chatbots for mental health support, there is an argument for investment in good/responsible systems informed by science, duty of care obligations and monitoring.

  18. Extend legislated data access across AI systems, not only for online safety.

    Data access should be extended across AI systems, not only for online safety: data and system access is necessary to monitor for responsible AI, and AI safety as AI is embedded across a wide range of critical societal systems (see the discussion of sovereignty/TOR(e) above). Without legislatively guaranteed data access, researchers face platform retaliation and the withdrawal, restriction, or restructuring of data retrieval tools.°°

  19. Provide legal protections for researchers, including copyright and IP exceptions and protection for citizen data donation.

    Rights to research/legal protection for researchers: the Digital Duty of Care Bill protects research ‘sock puppets’ (false identities); other necessary protections include copyright and IP exceptions and protection for citizen data donation.

  20. Create direct private rights of action and/or rights of action for civil society to expand who can act against online harms.

    Direct private rights of action and/or rights of action for civil society would expand the set of actors able to take action against online harms.

  21. Provide peer review of AI systems by the Australian AI Safety Institute and independent researchers, balancing provider confidentiality against public interest in inspectability and testing.

    Peer review of AI systems by the Australian AI Safety Institute, and independent researchers, balancing confidentiality concerns of providers with public interests in inspectability, testing and validation of AI platforms and models to identify and manage risks.

  22. Reconsider the previously proposed mandatory guardrails, including documentation, risk management, testing and monitoring obligations for high-impact AI.

    The government should reconsider the past proposed mandatory guardrails for AI, including basic documentation, risk management, testing and monitoring obligations for high-impact applications of AI: creating a ‘paper trail’ to facilitate enforcement.

Read the original

Open in the interactive guide