Submission 419
Mrs Ivy Cook
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.
Summary
Read the original submission for the full text.
Argues that Australia should pursue broad participation in AI benefits while preserving human agency. Recommends treating AI literacy as national infrastructure, an AI Champions program, SME adoption support, a federated sovereign research network and commercialisation pathway, meaningful human oversight, governance of delegated AI agency, and a two-speed approach to AI governance.
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.
Argues that if AI becomes a significant multiplier of individual and organisational capability, existing digital inequalities may become AI inequalities.
If AI becomes a significant multiplier of individual and organisational capability, existing digital inequalities may therefore become AI inequalities.
Argues that productivity should be measured by the additional capability AI gives people, not only by the human labour it removes.
We should ask not only how many hours of human labour a system removes, but also how much additional capability it gives to a person or organisation.
Argues that sovereign AI capability means retaining meaningful choices when technology, markets or access conditions change, not producing everything domestically.
Sovereign AI capability is not the ability to produce everything domestically. It is the ability to retain meaningful choices when technology, markets or access conditions change.
Argues that a nominal human in the loop is insufficient if the person merely rubber-stamps an automated recommendation.
A person positioned at the end of an automated process is not necessarily exercising meaningful oversight. If workloads, system design or institutional expectations make rejection of the automated recommendation unrealistic, the human may function merely as a rubber stamp.
Argues that governance should address delegated AI agency: what authority has been delegated to an AI system, beyond whether its output is accurate or fair.
The policy question is no longer only whether an AI output is accurate or fair. It also becomes: what authority has been delegated to the system?
Argues it would be unwise either to assume that AI will leave existing occupations untouched or to assume that widespread technological unemployment is inevitable.
It would be unwise either to assume that AI will leave existing occupations untouched or to assume that widespread technological unemployment is inevitable.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Treat equitable AI access and literacy as infrastructure; provide education through schools, universities, libraries, TAFE, community and online services, prioritising financial, geographic, accessibility and technological barriers; consider capable-service access alongside education.
Treat equitable access to AI capability and AI literacy as national infrastructure. Establish accessible AI education through schools, universities, libraries, TAFE, community institutions and online services, with particular attention to people facing financial, geographic, accessibility or technological barriers. Practical access to capable AI services should be considered alongside education.
Create a national AI Champions program based on a multiplier model, supporting people with domain expertise and practical AI capability to demonstrate responsible AI collaboration within industries, professions and organisations.
Create a national AI Champions program based on a multiplier model. Support people with both domain expertise and practical AI capability to become trusted demonstrators within industries, professions and organisations. Their role should be to show colleagues what useful, responsible AI collaboration looks like in the actual context of their work.
Build on existing initiatives to accelerate responsible adoption, especially SMEs, with targeted, time-limited grants, co-funding or tax incentives supporting experimentation, workforce capability and process redesign, beyond software purchases.
Accelerate responsible organisational adoption, particularly among SMEs. Build upon existing initiatives with targeted and time-limited grants, co-funding or tax incentives that support experimentation, workforce capability, process redesign and responsible implementation rather than merely subsidising software purchases.
Establish a federated coordinating AI research institute/network linking universities, CSIRO and other organisations, with shared programs, joint appointments, fellowships, talent attraction/retention, international collaboration and advanced compute for independent strategic capability.
Build a federated sovereign AI research capability. Establish a national coordinating institute or network connecting leading researchers across Australian universities, CSIRO and other research organisations. It should support shared research programs, joint appointments and fellowships, attraction and retention of talent, international collaboration, and meaningful access to advanced compute. Its purpose should be to ensure Australia retains the ability to independently understand, evaluate and contribute to strategically important AI capabilities.
Connect research to startups and scale-ups via industry pathways, tax/early funding, compute, procurement and international commercialisation support, retaining Australian companies, IP, jobs and economic value.
Build a sovereign AI commercialisation and scale-up pathway. Connect Australian research and innovation to startups and scale-ups through research-to-industry pathways, appropriate tax and early-stage funding support, access to compute, procurement opportunities and assistance with international commercialisation. Australia should seek not only to produce excellent research, but to retain more of the companies, intellectual property, employment and economic value that can grow from it.
Strengthen education and workforce capabilities in critical thinking, judgement, strategy, synthesis, creativity, communication and AI collaboration, with AI literacy beyond product operation.
Prepare Australians for changing work by strengthening higher-order human capabilities. Education and workforce development should increasingly emphasise critical thinking, judgement, strategy, synthesis, creativity, communication and the ability to collaborate effectively with AI. AI literacy should mean more than learning how to operate particular products.
Preserve meaningful human participation for significant consequences; prevent rubber-stamping and give high-impact uses genuine capacity to understand, question, challenge and, where appropriate, override or review AI-supported decisions.
Preserve meaningful human participation and judgement where consequences are significant. A nominal “human in the loop’ is insufficient if a person merely rubber- stamps an automated recommendation. High-impact uses should provide genuine capacity to understand, question, challenge and, where appropriate, override or review AI-supported decisions.
Govern delegated AI agency as well as AI models, addressing what an AI may access, infer and do, what needs human confirmation, how significant actions are recorded and reversed, and how responsibility is allocated.
Govern delegated AI agency as well as AI models. As systems gain access to tools, data, applications and the ability to act, policy should address what an AI may access, infer and do; what requires human confirmation; how significant actions are recorded and reversed; and how responsibility is allocated when humans and AI jointly produce outcomes.
Adopt two clocks: enduring legislation based on agency, accountability, fairness, transparency and accessibility; regularly or capability-triggered revised technical standards, guidance and capability assessments.
Adopt a “two clocks” model of AI governance. Enduring legislation should be grounded in durable principles such as human agency, accountability, fairness, transparency and accessibility. Faster-moving technical standards, guidance and capability assessments should be capable of regular or capability-triggered revision as AI develops.
Consider a relatively small coordinating research core, shared programs, fellowships, joint appointments, secondments and collaborative missions while allowing researchers to retain existing institutional affiliations.
A relatively small national coordinating core could bring together leading researchers through shared programs, fellowships, joint appointments, secondments and collaborative research missions while allowing researchers to retain their existing institutional affiliations.
Provide the sovereign research network either dedicated national research compute or dependable access to significant Australian computing resources, allocated by research merit and strategic need.
The network should therefore have either dedicated national research compute or dependable access to significant Australian computing resources allocated according to research merit and strategic need.
Build the AI commercialisation pathway on existing programs without duplication; consider targeted tax incentives, early-stage/co-investment mechanisms, research-to-industry fellowships, founder support and national compute access.
This should build upon, rather than duplicate, existing innovation programs. Targeted tax incentives, early-stage and co-investment mechanisms, research-to-industry fellowships, founder support and access to national compute could help promising companies cross the gap between technical development and commercial viability.
Give Commonwealth agencies and major Australian enterprises mechanisms to trial promising Australian AI technologies that meet operational needs and required standards of safety, security and value for money.
Commonwealth agencies and major Australian enterprises should have appropriate mechanisms to trial promising Australian AI technologies where these meet genuine operational needs and satisfy required standards of safety, security and value for money.
Give workers opportunities to gain AI capability before disruption reaches them.
Workers should have opportunities to gain AI capability before disruption reaches them.
Where productivity rises significantly, consider how gains are distributed, including wages, services, investment, reduced drudgery, job quality and shorter working time.
Where productivity increases significantly, attention should also be paid to how the gains are distributed. Productivity improvement can support higher wages, better services, new investment, reduced drudgery, improved job quality or shorter working time. It can also simply concentrate benefits.
Design education policy around collaborating with AI without outsourcing thought: formulating questions, interrogating assumptions, verifying claims, comparing alternatives, recognising uncertainty and taking responsibility for decisions.
Students and workers need to learn how to collaborate with AI without outsourcing thought to it. This includes learning how to formulate good questions, interrogate assumptions, verify claims, compare alternatives, recognise uncertainty and take responsibility for final decisions. There is an important distinction between using AI to avoid thinking and using AI to enable deeper thinking. Education policy should be designed around that distinction.
Retain layer-specific governance of data, models, applications and infrastructure; address risks where they originate, regulate outcomes rather than labels and require meaningful oversight for high-impact decisions.
My earlier submission to the South Australian Government’s AI Strategy consultation argued that AI governance should distinguish between the data, model, application and infrastructure layers, address risks at the layer where they originate, regulate outcomes rather than merely technological labels, and require meaningful human oversight for high-impact decisions. Those principles remain relevant, but the concept of human oversight requires further development.
Give consequential human participants genuine authority and sufficient information to exercise judgement.
For consequential applications, human participation should include genuine authority and sufficient information to exercise judgement.
Provide mechanisms for people affected by important AI-supported decisions to understand that automation was involved, seek explanation, challenge errors and obtain meaningful review.
People affected by important AI-supported decisions should also have appropriate mechanisms to understand that automation was involved, seek explanation, challenge errors and obtain meaningful review.
Apply different authority levels to different AI actions; let people know when AI acts for them, inspect significant actions, understand permissions and reverse actions where possible.
Different actions should require different levels of authority. People should be able to know when an AI is acting on their behalf, inspect significant actions, understand what permissions have been granted and reverse actions where reversal is possible.
Include privacy, contestability and responsibility in the enduring governance principles alongside human agency, accountability, transparency, fairness and accessibility.
The slow clock should contain principles that remain valuable even as technology changes: human agency, accountability, transparency, fairness, accessibility, privacy, contestability and responsibility.
Consider scheduled and capability-triggered review mechanisms where emerging AI capabilities materially alter the assumptions behind existing policy.
Governments should consider scheduled and capability-triggered review mechanisms where emerging AI capabilities materially alter the assumptions on which existing policy was based.