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 180

Professor Enrico Coiera

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.

Open in the interactive guide

Summary

Read the original submission for the full text.

Professor Enrico Coiera, who leads the Australian Alliance for AI in Healthcare, argues for a whole-of-sector national plan for AI in healthcare, algorithmic sovereignty, an integrated regulatory approach balancing patient safety against overregulation, workforce training and investment, and attaches the Alliance's roadmap of 28 recommendations across eight priority areas.

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 currently has no cohesive national approach to AI in healthcare, no integrated ethical or regulatory framework and no workforce strategy.

    Surprisingly, we currently have no cohesive national approach to maximise the benefits of AI in healthcare through better health outcomes and efficiencies, and no integrated ethical or regulatory framework to protect consumers when technology is unsafe or threatens their legitimate interests such as their privacy, and no workforce strategy.

  2. Argues that Australia will always require a sovereign capability because AI is a platform capability whose algorithms need to be localised and maintained.

    we will always require a sovereign capability given that AI is not a commodity. Rather, AI is a platform capability, and any AI algorithm we use will need to be localised and maintained — we will need local training sets for machine learning, and local testing to ensure AI systems meet our national standards for safety and quality.

  3. States that all Australians, including those in remote and regional areas and those managing chronic or rare illness, will benefit significantly from AI in healthcare.

    All Australians, including those in remote and regional areas and those managing chronic illness and rare disease, will benefit significantly.

  4. Describes a well-discussed tension between protecting patients from harm and avoiding overregulation of a still emerging industry.

    There is a well discussed tension between the clear mandate to ensure patients are not exposed to harm because of poor technology development or use, and the need to avoid overregulating a still emerging industry.

  5. Argues that other nations’ huge AI investments leave Australia facing a ‘run to catch up’ rather than readily achieving global leadership.

    Given the huge investments other nation states have already made in AI, and are planning to make in the future, Australia is already facing a “run to catch up” situation, not one in which it can easily contemplate global leadership.

  6. Argues that Frontier models embedded in healthcare infrastructure create a possible single point of failure if turned off or interfered with by other nations or industry.

    It is likely that in the next few years we will see Frontier models embedded into the foundations of healthcare’s digital infrastructure. This creates a unique and possibly single point failure mode if these models are turned off or otherwise interfered with by other nations, or by industry whim.

Recommendations

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

  1. Develop a whole-of-sector national plan to adopt AI safely and effectively across healthcare.

    The first and most important policy recommendation in the Roadmap is for national leadership and the development of a whole of sector (not just whole of government) national plan to safely and effectively adopt AI across healthcare.

  2. Set specific national AI targets and strategies involving stakeholders across healthcare, beyond government and agencies.

    If Australia truly wishes to exploit the potential for AI, it will need specific targets and strategies to enable these, and the solution must involve multiple players across the health sector, not just Government and its Agencies.

  3. Provide local AI training datasets and testing against national safety and quality standards, while taking advantage of leading technologies from any country.

    While Australia should always take advantage of leading technologies, no matter which nation they might come from, we will always require a sovereign capability given that AI is not a commodity. Rather, AI is a platform capability, and any AI algorithm we use will need to be localised and maintained — we will need local training sets for machine learning, and local testing to ensure AI systems meet our national standards for safety and quality.

  4. In mission-critical settings, sometimes own the entire AI process from algorithm creation to deployment.

    In mission critical settings, we will sometimes need to own the entire process from algorithm creation to deployment.

  5. Invest in developing Australia's own AI capabilities, including frontier models, possibly in partnership with a hyperscaler.

    We will need to invest in developing our own AI capabilities, including Frontier models, possibly in partnership with a ‘hyperscaler’ company.

  6. Ensure the national AI in healthcare strategy secures sovereign control over the use of any mission critical models in digital infrastructure.

    our national AI in healthcare strategy will need to ensure we have sovereign control over the use of any mission critical models that find their way into our digital infrastructure.

  7. Bring regulatory and standards organisations together to develop a clear approach to safe and ethical AI use in healthcare.

    We need as a nation to bring these different organisations together and develop a clear approach to ensuring we have safety and ethical use firmly embedded in any use of AI in the healthcare system.

  8. Significantly change and invest in the training provided to the health sector workforce to integrate AI.

    In Australia this will require a significant change to, and investment in, the training we provide to the health sector workforce.

  9. Make the healthcare AI investment plan targeted but proportionately ambitious, given Australia’s modest starting point.

    Given our relatively modest starting point, this will need to be a clever, targeted but proportionately ambitious plan.

  10. Give AI, as a core platform technology, investment consideration similar to the Medical Research Future Fund genomics mission, in healthcare and across other industry sectors.

    By way of benchmark, Australia’s Medical Research Future Fund (MRFF) is investing more than half a billion dollars into a national genomics mission over a decade. We believe AI, which is a core platform technology, deserves similar consideration, in healthcare, and indeed across other industry sectors.

  11. Develop and routinely refresh an ambitious, well-resourced National AI in Healthcare strategy led by government with broad community and stakeholder engagement.

    Develop and routinely refresh an ambitious and well- resourced National AI in Healthcare strategy led by government but based upon broad engagement with the community and key stakeholders.

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

  12. Focus the national healthcare AI strategy on federal issues while supporting developments such as state strategies.

    The national strategy should focus on federal issues but be supportive of developments elsewhere e.g. State level strategies.

  13. Establish and resource a National Health AI Expert Group to advise government on policy and regulatory proposals for national adoption of AI in healthcare.

    Establish and resource a National Health AI Expert Group to advise government as it develops policy and regulatory proposals for a coherent and strategic approach to national adoption of AI in healthcare.

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

  14. Include consumers, Aboriginal and Torres Strait Islander people, industry, health services, academia and relevant government agencies in the expert advisory committee.

    The Expert Advisory committee should include representation from consumers, Aboriginal and Torres Strait Islander people, industry, health services, academia, and government agencies such as the TGA, ACSQHC, Ahpra, and ADHA.

  15. Consider a consumer panel feeding into the high-level committee, and align the strategy with the National Agreement on Closing the Gap.

    Aconsumer panel or sub-committee feeding into the high-level committee may help capture a more representative community perspective. The Strategy should align with the National Agreement on Closing the Gap.>

  16. Require healthcare organisations using AI clinically to have robust clinical governance, with access to national tools and guidelines and independent review of governance quality, for example through accreditation.

    HCOs that use AI for clinical purposes must have robust clinical governance processes that prioritise the safe, effective and ethical use of AI. 3.1. HCOs must have access to national tools and guidelines that support effective AI clinical governance at an organisational level. 3.2. The quality of a HCO’s AI clinical governance processes must be independently reviewed, e.g. as part of existing organisational accreditation processes run by ACSQHC.

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

  17. Where possible, build AI governance on existing digital-health and patient-safety arrangements and collaborate with related agencies; make data storage governance follow Indigenous sovereignty, governance and related national policies.

    Where possible, AI governance should build on existing digital health and patient safety governance arrangements and collaborate with agencies that have related missions, e.g. the AI Safety Institute. Data storage governance should adhere to the principles of Indigenous Data Sovereignty, Indigenous Data Governance, and existing related national policies.” ®

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

  18. Use governance approaches suited to different AI technologies and applications.

    Governance requires recognition that different technologies (e.g. GenAl vs Deep Learning) and different applications (e.g. SaMD vs scribes) will require different approaches.

  19. Extend governance from procurement through implementation and monitoring of clinical outcomes, and develop national healthcare AI governance accreditation standards to augment organisational standards.

    AI governance should extend from AI procurement, through to implementation and post-implementation performance monitoring that includes clinical outcomes. National Healthcare AI governance accreditation standards will need to be developed to augment existing organisational standards.

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

  20. Include AI safety incident reporting and guidance favouring technologies meeting appropriate standards, such as medical-device approval.

    Governance should include AI safety incident reporting (Recommendation 5) and guidance to preferentially adopt technologies that meet appropriate standards such as medical device regulatory approval.

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

  21. Move to actionable national ethical guidelines, tools and standards of care reflecting value-based care across clinical and consumer AI.

    Embedding ethics into clinical governance will require a move from developing AI ethics principles and frameworks to the use of actionable national ethical guidelines, tools, and standards of care that reflect value-based care spanning clinical and consumer AI.

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

  22. Make ethics guidelines specific enough for professional and organisational standards of care, potentially with expert input, and place ownership and stewardship with an accrediting body such as ACSQHC.

    Ethics guidelines must be sufficiently specific to inform the development of professional and organisational standards of care. Development of such guidelines may best occur with expert input, but ownership and stewardship should sit with an accrediting body such as the ACSQHC.

  23. Update ethical guidelines and tools regularly and promptly as technologies, uses and norms evolve.

    Ethics guidelines and tools will require timely and regular updating as new challenges emerge with new technologies, uses, and norms.

  24. Respect consumers’ conscientious objection to AI where a choice is possible.

    Where a choice is possible, conscientious objection to the use of AI by consumers should be respected.

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

  25. Immediately prioritise and resource effective post-market surveillance of AI in routine use so new risks are rapidly detected and communicated, including to consumers.

    Effective post-market surveillance of AI in routine use requires immediate prioritisation and resourcing so that new AI risks are rapidly detected and communicated to all relevant parties, including consumers.

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

  26. Have ADHA, regulators and accreditation organisations jointly examine AI monitoring responsibilities and enhance missions and capabilities as appropriate.

    ADHA, regulators such as the TGA, and accreditation organisations such as the ACSQHC should jointly examine their potential responsibilities for monitoring of AI in use and enhance their mission and capabilities as appropriate.

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

  27. Harmonise and integrate patient safety reporting nationally, and include routine post-market AI safety monitoring and regulator reporting in clinical governance.

    Harmonising and integrating patient safety reporting should be a national priority. Formal clinical governance processes should include routine post-market monitoring for AI-related safety incidents, and their reporting to regulators.

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

  28. Give generative AI products additional post-market scrutiny, move towards real-time risk monitoring as adoption scales, and adapt pre-market assessment, for example through predetermined change control programs.

    Generative AI products, such as AI scribes, require additional scrutiny because safety assessment will necessarily rely more heavily on post-market mechanisms than other forms of AI.? Given that Generative AI is now being adopted at scale, it will soon require real-time monitoring for risks and harms, e.g. of live patient records, and transactions such as e-prescriptions. Pre-market assessment will continue to be important but will need to adapt, e.g. using predetermined change control programs.

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

  29. Include national cybersecurity standards compliance in accreditation standards for AI deployment in healthcare organisations.

    Accreditation standards for AI deployment in healthcare organisations should include a requirement for national cybersecurity standards compliance.

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

  30. Explicitly address AI-specific privacy risks not covered by cybersecurity standards in healthcare organisations' clinical governance, training and procurement.

    AI-specific privacy risks in healthcare that are not covered in national cybersecurity standards (such as patient re-identification from machine learning models), should be explicitly addressed in HCO clinical governance, training, and procurement processes.

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

  31. Establish guidance for managing when Australian patient data is subject to access or disclosure obligations to international jurisdictions.

    Establish guidance for managing when Australian patient data is subject to access or disclosure obligations to international jurisdictions.

  32. Build sufficient technological independence, including local expertise, infrastructure and governance, to implement AI in healthcare and reduce strategic reliance on international providers.

    9. Australia should build sufficient technological independence—including local expertise, infrastructure, and governance—to safely and effectively implement Alin healthcare and reduce our strategic reliance on international technology providers.

  33. Balance international providers’ expertise and resources with locally governed independent capability available in crises; prioritise sovereign capability in national AI policy.

    We must balance reliance on international providers (who could provide cutting edge expertise and resources) with the need for local independent capability (which is under local governance and available in times of crisis). + Development of national AI policy that ensures sovereign capability is developed to avoid the risks of over-reliance on international providers in geopolitically unstable times is a priority (see Industry recommendations).

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

  34. Make AI localisation capabilities available so Australian healthcare organisations can use AI calibrated to local populations, services and health needs.

    10. AI localisation capabilities must be available to ensure Australian HCOs can use AI that is calibrated to the characteristics of specific local populations, services, and health needs.

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

  35. Give Australian companies and healthcare organisations access to expertise, datasets and technology for localisation; accelerate data governance and ensure the NHMRC data plan supports quality improvement and AI localisation alongside research.

    Australian companies and HCOs should have access to the best-practice expertise, high-quality datasets, and technologies needed to support model localisation. The Australian Institute of Health and Welfare (AIHW) is already a trusted source of health and welfare data, and making data available for AI development, including by industry, will require different and accelerated approaches to data governance. The new NHMRC National Data Integration, Sharing and Access Plan should ensure it supports not just data use for research but also data use for quality improvement, including the localisation of AI models.

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

  36. Invest in privacy-preserving data ecosystems, such as federated learning, enabling AI development without necessarily sharing raw patient data across institutions.

    11. Australia should invest in privacy-preserving data ecosystems (e.g. federated learning) that enable AI development without necessarily sharing raw patient data across institutions.

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

  37. Universally adopt data interoperability and prioritise and reward data sharing across public/private sectors and state/federal jurisdictions; consider Indigenous Data Sovereignty and Governance alongside national sovereignty.

    The development of data ecosystems will be enhanced by universal adoption of data interoperability and policy that prioritises and rewards data sharing across public and private sectors, and State and Federal jurisdictions. + Indigenous Data Sovereignty and Indigenous Data Governance must be considered alongside national data sovereignty concerns.

  38. With federal government support, develop, maintain and provide national access to locally governed generative AI optimised to local healthcare needs, possibly developed by industry.

    12. With Federal government support, Australia should develop, maintain, and provide national access to locally governed generative AI capabilities, optimised to local healthcare needs. Such national generative AI capabilities may well be developed by industry, with government engagement.

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

  39. Develop, disseminate and routinely refresh national best-practice guidance for clinical implementation of AI, such as use cases, standards compliance, workflows, calibration, transparency and explainability.

    Develop, disseminate, and routinely refresh national best-practice guidance for clinical implementation of AI (e.g. high priority use cases, technical standards compliance, workflows, local calibration, transparency, and explainability).

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

  40. Leverage AI-advanced state and territory health systems’ expertise and experience in implementation guidance.

    This should leverage the expertise and experience of AI-advanced State and Territory health systems.

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

  41. Develop, disseminate and routinely refresh State and Federal procurement guidance to help health services identify proven AI technology.

    Develop, disseminate, and routinely refresh procurement guidance at State and Federal levels to support health services to identify proven AI technology for implementation in healthcare.

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

  42. Have ADHA and MTAA collaborate with jurisdictional e-health agencies and industry to develop and maintain healthcare AI procurement guidance, supporting local industry and sovereign capability where possible.

    Organisations such as ADHA and MTAA should collaborate with e-health agencies from the jurisdictions and industry to develop and maintain these healthcare AI procurement guidelines. Procurement guidelines should, where possible, support local industry and contribute to the development of sovereign capabilities in healthcare AI, its implementation, and operation.

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

  43. Identify key national healthcare challenges where AI can enhance outcomes and create national beacon sites to address them before broader national translation.

    Identify key national healthcare challenges where AI can significantly enhance outcomes and effectiveness and create national beacon sites where these challenges can be first addressed before broader national translation.

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

  44. Have beacon projects strive to be full services embedded in routine care rather than pilots, with state and federal leadership and resourcing.

    Beacon projects should strive to be ‘full’ AI service implementations embedded in routine care, rather than pilots. Potentially spanning from primary through to tertiary care, they will require leadership and resourcing from State and Federal Governments.

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

  45. Fully fund process, economic and clinical-outcome evaluations of beacon projects, and include translation into other healthcare organisations in project KPIs.

    Beacon projects must include fully funded evaluations, covering process, economic, and clinical outcome assessments. KPIs for beacon projects should include evidence of translation of the new AI service into other HCOs.

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

  46. Support national consensus on foundational AI clinical competencies, professional scopes and codes of conduct to underpin patient safety, service quality and practitioner credentialling.

    To prepare the sector for the increased use of AI, we will need to support the creation of national consensus on foundational clinical competencies, scopes of professional practice, and codes of professional conduct to use AI, and provide a basis for patient safety, service quality, and practitioner credentialling.

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

  47. Have AI governance standards encompass workforce training requirements so healthcare organisations can deliver safe, secure and effective AI services.

    16. AI governance standards should encompass workforce training requirements, ensuring that HCOs can deliver safe, secure, and effective services using AI.

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

  48. Make safe, secure and effective use of AI a component of clinical training and digital health training programs.

    17. Safe, secure, and effective use of AI should be a component of clinical training and digital health training programs.

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

  49. Resource a national Chief AI Officer training program to develop and upskill executive leaders to drive procurement, governance and implementation of healthcare organisation AI.

    18. Anational Chief AI Officer training program should be resourced to produce a new generation of executive leaders, and upskill existing leaders such as Chief Information Officers, to drive the procurement, governance, and implementation of HCO AI.

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

  50. Use workforce training requirements to develop a marketplace of appropriately targeted programs, including university and professional-organisation provision.

    Training requirements should drive the development of the educational marketplace, e.g. universities and professional organisations such as the AIDH can be providers of appropriately targeted educational programs.

  51. Make clinicians aware of AI’s potential disruption of cognitive processes and train them to maintain high-quality decisions.

    Clinicians should be made aware that use of AI may disrupt current cognitive processes (such as the act of sense-making whilst writing patient notes), and be provided with training to ensure high quality decision making continues.

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

  52. Inform and engage consumers in decisions about AI development, use and implications to foster public confidence.

    Consumers must also be informed and engaged in decisions about the development and use of AI and its implications for them, to foster public confidence.

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

  53. Upskill consumers to safely and effectively adopt healthcare AI products and services.

    19. Upskill consumers to safely and effectively adopt healthcare AI products and services.

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

  54. Give consumers accessible, targeted education on health AI use, safety and clinicians’ obligations, including consent for tools such as digital scribes.

    Consumers should have access to appropriate, accessible, and targeted educational material that covers how to use AI for health purposes, AI safety, and the professional obligations of clinicians who use AI e.g. how patient consent to use AI like digital scribes should be obtained.

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

  55. Consider teaching safe, effective consumer AI use through services such as 1800 Medicare, My Health Record and the Medicare app, and provide information on discussing AI-generated information with health professionals.

    Training in the safe and effective use of AI by consumers might best occur as a part of using real services such as the 1800 Medicare helpline, My Health Record, or the 1800 Medicare app. + Consumers should have access to information on how to best use AI-generated information e.g. discussing such information during a clinical consultation with a doctor or other health professional.

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

  56. Create subgroup-specific educational materials and work with Aboriginal and Torres Strait Islander community-controlled, CALD and disability sectors to determine effective literacy strategies.

    Specific educational material should be created for different subgroups, e.g. age specific groups and individuals from Culturally and Linguistically Diverse (CALD) communities. Partnership with the Aboriginal and Torres Strait Islander community-controlled, CALD, and disability sectors will determine effective strategies to build AI literacy for these communities.

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

  57. Undertake a national consumer educational campaign to build trust in healthcare organisations and address health misinformation and disinformation encountered through publicly available AI.

    Undertake a national consumer educational campaign to build trust in HCOs, and address health misinformation and disinformation consumers encounter when using publicly available AI.

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

  58. Embed consumers in the educational campaign from conception to implementation; sustain it over a lengthy period with agility to counter emerging threats.

    Consumers will need to be embedded in the campaign from conception to implementation. As AI capabilities are rapidly evolving, new (mis)uses for AI are always emerging. Hence, the campaign should extend over a lengthy period and be agile to counter new threats.

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

  59. Target the campaign at online health misinformation and consider 1800 Medicare with clinical peak bodies as its lead.

    The campaign should seek to counter the specific risks of online healthcare misinformation and disinformation. 1800 Medicare in partnership with clinical peak bodies may be an appropriate lead for such a campaign.

  60. Mandate and resource inclusive, safe and accessible AI-driven consumer health services for disadvantaged populations, including disability, CALD and Aboriginal and Torres Strait Islander communities, based on deep engagement.

    21. AI-driven consumer health services should support disadvantaged populations. A mandate and resources are needed to support inclusive, safe, and accessible access to AI-driven online consumer resources by, for example, people living with disabilities, CALD communities, and Aboriginal and Torres Strait Islander communities, based on deep engagement with these populations.

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

  61. Adopt an AI front door strategy using AI to enhance access to and navigation across services such as My Aged Care and 1800 Medicare.

    22. Adopt an AI front door strategy for consumer-facing health services: Use AI to enhance access to and navigation across and within critical services such as My Aged Care and 1800 Medicare.

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

  62. Develop a data and technology standards roadmap setting out AI-specific healthcare requirements beyond current digital health standards and how they will be adopted, embedded and sustained.

    23. Develop a data and technology standards roadmap that identifies AI-specific requirements for healthcare that extend beyond current digital health standards and clearly sets out how these standards will be adopted, embedded, and sustained across the health ecosystem.

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

  63. Include industry as a core partner in healthcare AI standards discussions.

    Industry (e.g. via MSIA) have created a self-regulatory pathway and so should be a core partner for these discussions.

  64. Develop a business case for government financial incentives and targeted support to help local industry with regulatory compliance.

    24. Develop a business case for government to provide financial incentives and targeted support for local industry to assist with regulatory compliance.

  65. Make AI an explicit focus of any National Health and Medical Research Strategy, generating evidence on AI safety, quality and effectiveness in health services.

    25. AI should be an explicit focus of any National Health and Medical Research Strategy, generating urgently needed evidence for AI safety, quality, and effectiveness in health services.

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

  66. Consider targeting AI research within existing funding programs in the short term, incorporate broader strategic considerations into NHMRC Strategy updates, and consider faster funding models, including targeted calls, pre-approved shortlists and rapid-response consortia.

    These gaps are a missed opportunity. They could be addressed in the short term by specific targeting of AI research within existing research funding programs. + Longer term, incorporation of broader strategic considerations into updates to the NHMRC Strategy will be essential. For example, the current pace of technology change means that traditional funding cycles are just too slow to answer urgent questions. Alternate funding models will need to be contemplated e.g. fast turnaround targeted calls, pre-approved short-lists, and funded consortia able to provide rapid evidence-based responses to urgent questions.

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

  67. Create dedicated AI review panels within the MRFF/NHMRC for expert multidisciplinary assessment of AI-related grant applications.

    26. Dedicated AI review panels should be created within the Medical Research Future Fund (MRFF)/National Health and Medical Research Council (NHMRC) to ensure expert multidisciplinary assessment of grant applications involving AI, across schemes such as Investigator and Partnership grants.

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

  68. Support research-sector retooling for safe and effective AI-enabled research, and ensure review panels can assess whether AI automation methods are appropriately declared and used in applications.

    To accelerate research, there will need to be research sector support to “re-tool” and to effectively and safely adopt new AI-enabled research tools. Review panels should have the expertise to assess whether such AI research automation methods are appropriately declared and used within grant applications.

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

  69. Allocate funding to one or more large-scale Centres of Research Excellence in Healthcare AI.

    27. Funding should be allocated to one or more large scale Centres of Research Excellence in Healthcare AI, e.g. with support potentially coming from the NHMRC, Australian Research Council (ARC), MRFF, DoHDA, and the States.

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

  70. Prioritise industry-academia AI adoption collaborations in existing schemes; make national synthetic or privacy-preserving datasets and localised LLM adoption early MRFF investment priorities.

    28. Funding and resources for industry-academia collaborations that progress the meaningful adoption of AI should be prioritised within existing schemes such as MTPConnect and NHMRC Partnership Grants. National resources (e.g. synthetic or privacy preserving data sets or adoption of localised LLMs) should be early foci for MRFF investment

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

  71. Fund large-scale infrastructure and foundational advances for sustainable sovereign AI innovation, alongside applications and tool development.

    R&D funding should not just focus on specific AI applications or AI tool development, but on large- scale infrastructure and foundational advances that enable sustainable and sovereign AI innovation.

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

  72. Prioritise research funding for populations with significant unmet needs or special circumstances, including Aboriginal and Torres Strait Islander-led research with ACCOs as lead researchers.

    Research funding should prioritise funding for populations with significant unmet needs or special circumstances, e.g. Aboriginal and Torres Strait Islander led research with support from Aboriginal Community-Controlled Organisations (ACCO) as lead researchers to produce high-impact, culturally grounded evidence to guide AI policy and practice.

    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