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 569

Mr Peter Leech

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.

Argues that AI policy should expand beneficial capability while preserving freedom, personal agency and clear institutional responsibility. Proposes assessing regulation against evidence, foregone benefits and less restrictive alternatives, and comments on the Committee's 12 October 2026 letter, assurance and incident reporting, and explainable, correctable government decisions.

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 realising AI's benefits requires reliable tools, skills and room to question, and that people should be able to use AI lawfully and exercise judgement about its outputs.

    Realising these benefits requires reliable, affordable tools, skills, accessible design and evaluation in practice. It also requires room to experiment, learn, question and disagree. Freedom and meaningful agency are part of the opportunity: people should be able to use AI lawfully and exercise judgement about its outputs.

  2. Argues that duties should be allocated according to each actor's role, knowledge, control and capacity, with adults, developers and public agencies each responsible for their own decisions.

    Adults should remain responsible for their choices and conduct. Developers, providers, deployers and public agencies should remain responsible for their own decisions and risks they can reasonably foresee and influence. Preventive duties should reflect role, knowledge, foreseeability, control and capacity; responsibility for an adverse outcome also depends on causal contribution and the applicable law.

  3. Argues that assessment of AI regulation must include foregone benefits, and that the effects of restrictions need evidence just as claims of harm do.

    Assessment must include foregone benefits. Restrictions affecting research, assistive tools or small-business adoption may delay valuable activity, impose costs or narrow lawful inquiry. Those effects need evidence, just as claims of harm do.

  4. Argues that serious uncertain risks can justify precaution, but that uncertainty does not automatically justify prohibition, and responses should be targeted, monitored and revisable.

    Serious uncertain risks can justify precaution before injury. Uncertainty does not automatically justify prohibition or inaction: responses should be targeted, monitored and revisable.

  5. Recommends that consequential government decisions have usable reasons, auditable decision paths, accessible review and plans to remedy errors.

    For consequential government decisions, I recommend usable reasons, auditable decision paths, accessible review by someone able to change the outcome, and plans to remedy errors.

  6. Says that, while supporting proportionate assurance and incident reporting in principle, the Committee's letter proposals on assurance, frontier-lab reporting and information use need clear thresholds and safeguards.

    For developers and providers, I support proportionate assurance and useful incident reporting in principle. The letter’s 1(c) assurance proposals, 1(d–e) frontier-lab reporting and evaluation, and 1(f) information-use protections need clear thresholds, evaluator independence, access boundaries and confidentiality safeguards.

Recommendations

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

  1. Expand beneficial capability and enable responsible activity while preserving freedom, meaningful agency and people’s ability to learn, question, disagree and use AI lawfully.

    Expand beneficial capability while protecting freedom and meaningful agency. Enable useful, responsible activity and preserve people’s ability to learn, question, disagree and use AI lawfully.

  2. Preserve personal and specify institutional responsibility, allocating duties according to each actor’s choices, risks and control, without automatic provider liability or blanket immunity.

    Preserve personal responsibility and specify institutional responsibility. Allocate duties to the choices, risks and control of each actor, without automatic provider liability or blanket immunity.

  3. Calibrate policy to evidence, comparing benefits, harms and foregone opportunities; test necessity, proportionality and less restrictive alternatives while allowing justified precaution under uncertainty.

    Calibrate policy to evidence and compare benefits, harms and foregone opportunities. Test necessity, proportionality and less restrictive alternatives, while allowing justified precaution under uncertainty.

  4. Allocate preventive duties by role, knowledge, foreseeability, control and capacity; assess adverse-outcome responsibility against causal contribution and applicable law, without automatic provider liability or blanket immunity.

    Preventive duties should reflect role, knowledge, foreseeability, control and capacity; responsibility for an adverse outcome also depends on causal contribution and the applicable law. This neither makes a provider automatically responsible for every misuse nor grants blanket immunity.

  5. Do not transfer responsibility by consent checkbox to people unable to understand/control risk; keep agencies answerable for decisions using external suppliers.

    A consent checkbox should not transfer responsibility to someone unable to understand or control the risk, and an agency should remain answerable for its decisions when it uses an external supplier.

  6. Begin policy with the intended outcome, the problem and the actor able to address it, comparing existing law, better enforcement, voluntary practice, technical design and new legislation alongside skills and practical support.

    Policy should start with the intended outcome, the problem and the actor able to address it. Is the gap in law, enforcement, information, design or remedies? Compare existing law, better enforcement, voluntary practice, technical design and new legislation, alongside skills and practical support.

  7. Examine harm evidence and emerging-risk pathways, severity, exposure and estimable likelihood; disclose assumptions/uncertainty in optimistic and alarming claims; do not treat speculation alone as proving restrictions necessary.

    Both optimistic and alarming claims should disclose assumptions and uncertainty. Examine the causes and extent of demonstrated harms; assess credible emerging risks through their plausible pathways, severity, exposure and likelihood where estimable. Speculative scenarios can inform preparedness but cannot alone establish a particular restriction’s necessity.

  8. Include evidence of foregone research, assistive-tool and small-business benefits; compare protection with privacy, security and distributional burdens, preferring less restrictive measures where comparably effective.

    Assessment must include foregone benefits. Restrictions affecting research, assistive tools or small-business adoption may delay valuable activity, impose costs or narrow lawful inquiry. Those effects need evidence, just as claims of harm do. Effective safeguards can improve reliability, confidence and uptake. Compare protection with burdens—including privacy, security and distributional effects—and prefer less restrictive measures where comparably effective.

  9. Test proposed legal restrictions for necessity, proportionality and rights effects, rather than treating legality alone as justification.

    Proposed legal restrictions must also be tested for necessity, proportionality and effects on rights; legality alone does not settle their justification.

  10. If future enforceable criteria invoke concepts such as culture and values, make clear their meaning, who determines them and protection for legitimate disagreement.

    It does not expressly propose censorship or takedown powers; its copyright discussion of culture and values does not establish them. If future enforceable criteria invoke those concepts, their meaning, who determines them and protection for legitimate disagreement should be clear.

  11. For consequential government decisions, provide usable reasons, auditable paths, accessible review able to change outcomes and error remedies; equip human oversight with information, competence, authority and resources. Keep remedies dependent on the decision/law and distinguish ordinary outputs from consequential decisions’ review.

    For consequential government decisions, I recommend usable reasons, auditable decision paths, accessible review by someone able to change the outcome, and plans to remedy errors. Human oversight needs information, competence, authority and resources. The Commonwealth Ombudsman’s guidance supports this approach; legal remedies depend on the decision and governing law. Ordinary model outputs need not all receive the same review process.

  12. For letter 1(c–f) assurance/frontier-lab reporting/evaluation/information-use proposals, specify thresholds, evaluator independence, access/confidentiality safeguards; avoid needless sensitive-information exposure or discharging other duties, and distinguish furnished-information protection from immunity for independently established misconduct.

    The letter’s 1(c) assurance proposals, 1(d–e) frontier-lab reporting and evaluation, and 1(f) information-use protections need clear thresholds, evaluator independence, access boundaries and confidentiality safeguards. Reporting should neither unnecessarily expose sensitive information nor discharge other responsibilities. Protection for furnished information must remain distinct from immunity for independently established misconduct.

  13. Have the review proposed in the letter's 1(k) assess capabilities realised, access, harm reduction, freedom and agency, burdens and effective remedies, with ineffective or disproportionate measures changed or withdrawn and inadequate protection strengthened.

    The review proposed in the letter’s 1(k) should assess capabilities realised, access, harm reduction, freedom and agency, burdens and effective remedies. Ineffective or disproportionate measures should be changed or withdrawn; inadequate protection should be strengthened.

  14. Make safeguards operational, evaluate gains and protection, make consequential government decisions explainable/correctable and assess whether governance achieves its objectives.

    Make safeguards operational and evaluate what people gain as well as what they are protected from. Make consequential government decisions explainable and correctable, and assess whether governance achieves its intended outcomes.

Read the original

Open in the interactive guide