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 187

National Open Science Taskforce (NOST)

Submitter group: Civil society.

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.

Members of NOST’s steering committee argue that Australian AI sovereignty depends on public-interest knowledge infrastructure, diverse and open models, Indigenous-led data governance and creators’ rights retention. They recommend coordinated research infrastructure, documented, open-weight, versioned AI capabilities and models, and research-sector consultation on legislation.

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 AI development has relied heavily on publicly funded knowledge and tools, and that AI should be public infrastructure.

    AI currently is a combination of public and private infrastructure and its development thus far has relied heavily on publicly funded knowledge and tools. AI should be public infrastructure.

  2. Argues that Australian research infrastructure is fragmented and increasingly integrated with commercial systems outside Australian control.

    Yet these systems are fragmented, unevenly funded and increasingly integrated with commercial publisher, platform and AI-provider systems whose terms, data practices and future costs are outside Australian control.

  3. Proposes reducing reliance on closed foreign-operated LLMs through partnerships with multiple providers and smaller specialised models.

    Reduced reliance on closed Large Language Models, including those operated by foreign multinationals, by developing strategic partnerships with multiple providers and also training smaller, specialised models.

  4. Argues that data sovereignty should extend beyond domestic hosting or ownership to rights, obligations, provenance and community authority.

    Data sovereignty should not be understood only as domestic hosting or Australian ownership. The Committee’s consideration of Australian data and domain-specific models should recognise that data is governed by rights, obligations, provenance, privacy, confidentiality, cultural protocols and community authority.

  5. Argues that separate AI licensing of scholarly collections risks publicly funded research being paid for repeatedly.

    Where publishers separately license the same scholarly collections to AI providers, there is a real risk that publicly funded research is paid for repeatedly: first to produce it, then to access or publish it, then to use it in education, and again through AI-specific licensing arrangements from which neither the researchers nor the funding public benefit.

  6. Argues that large multinational publishers are entering AI licensing deals without needing to inform authors, removing authors' capacity to opt out.

    Due to one-sided terms, publishers - especially large multinational publishers - as rights holders, are entering into deals with AI companies without the need to inform authors, effectively removing any capacity to now opt out.

Recommendations

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

  1. Consult the research sector meaningfully on the planned AI legislation expected by early 2027.

    We welcome the intent to develop legislation on these issues by early 2027, which will require meaningful consultation with the research sector to fully explore risks and opportunities.

  2. Adopt a National Open Science Policy or Action Plan, rights retention and linked repositories enabling public research to be deposited, preserved, verified, discovered and reused through trusted public-interest infrastructure.

    NOST has previously called for a National Open Science Policy or Action Plan, effective rights retention and linked national repositories. These measures are now also essential components of national AI capability. Publicly funded research should be able to be deposited, preserved, verified, discovered and reused through trusted infrastructure governed in the public interest.

  3. Provide new funding to create more interoperable, robust infrastructure for Australia's research outputs.

    However, new funding is required to create a more interoperable, robust infrastructure for Australia's research outputs.

  4. Create accountability, transparency and reporting requirements for large commercial AI providers.

    Policy that creates accountability, transparency, and reporting requirements for large, commercial AI providers.

  5. Reform policy so researchers can undertake non-commercial activities with certainty and fewer cost and licensing barriers.

    Policy reform to enable researchers to undertake non-commercial activities with certainty and fewer cost and licensing barriers

  6. Invest in skills so researchers and students can use AI effectively.

    Investment in skills to ensure researchers and students have the capabilities to use AI effectively

  7. Invest in open research infrastructure connecting the Australian research landscape, fostering creative reuse and providing appropriate AI controls.

    Additional investment in open research infrastructure to connect the Australian research landscape and foster opportunities for creative reuse, and appropriate controls for AI

  8. Reduce reliance on closed large language models, including foreign-operated ones, through strategic partnerships with multiple providers and training smaller, specialised models.

    Reduced reliance on closed Large Language Models, including those operated by foreign multinationals, by developing strategic partnerships with multiple providers and also training smaller, specialised models.

  9. Develop documented, open-weight, versioned Australian AI capabilities and models, as AIMOS recommends.

    We note and agree with the position of AIMOS in their written submission calling for Australian AI capabilities and models that are documented, open weight, and versioned.

  10. Recognise that data sovereignty extends beyond domestic hosting or ownership to rights, obligations, provenance, privacy, confidentiality, cultural protocols and community authority in the Committee’s consideration of Australian data and domain-specific models.

    Data sovereignty should not be understood only as domestic hosting or Australian ownership. The Committee’s consideration of Australian data and domain-specific models should recognise that data is governed by rights, obligations, provenance, privacy, confidentiality, cultural protocols and community authority.

  11. Where relevant, support Indigenous-led governance, resourced Indigenous-led AI and data initiatives, culturally appropriate access conditions and meaningful mechanisms for remedy in AI policy.

    When relevant, AI policy should therefore support Indigenous-led governance, resourced Indigenous-led AI and data initiatives, culturally appropriate access conditions, and meaningful mechanisms for remedy.

  12. Ensure any copyright policy response protects creators' ability to retain their rights in future, including the choice to make work openly available on their own terms.

    Any policy response should protect creators' ability to retain their rights in future, including the choice to make their work openly available on their own terms

  13. Enable researchers and institutions to examine and improve the quality and reliability of knowledge resources used in AI systems, including through access to correction and retraction information.

    Policy should also enable researchers and institutions to examine and improve the quality and reliability of the knowledge resources used in AI systems, including through access to correction and retraction information.

  14. Establish more robust mechanisms for provenance and attribution and stronger controls on how content inputted to LLMs can be used.

    there is an urgent need to establish more robust mechanisms for provenance, attribution, and stronger controls about how content inputted to LLMs can be used.

Read the original

Open in the interactive guide