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 18

Human Technology Institute, University of Technology Sydney

Submitter group: Universities.

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 four reforms are needed to build community trust in AI: a training-data transparency requirement in the forthcoming Australian Standards for AI, Fair Work Act amendments on worker surveillance, a law governing government automated decision making, and passage of amended privacy reforms with a resourced OAIC.

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. Recommends requiring companies to publish a high-level summary of AI training data when a model is made available in Australia.

    A TDT requirement would require companies to publish a high-level summary of the data used for training an AI model when it is made available in Australia.

  2. Argues that existing laws on worker surveillance are complex and do not provide adequate protections for workers.

    This problem is not addressed by existing laws, which are complex and confusing for employers and workers alike. Current law also fails to provide adequate protections for workers subject to surveillance.

  3. Recommends that a government automated decision-making law have Parliament decide which categories of decision may be automated.

    Parliament should determine which types of decision may be automated. The determination whether to allow automation for a category of government decision making (such as the availability of services under the National Disability Insurance Scheme) should be made by Parliament, given that this enables effective democratic oversight.

  4. Argues that poorly designed and regulated automated decision making is already harming large numbers of vulnerable Australians.

    Real harm is being experienced now – by large numbers of vulnerable Australians – as a result of poorly designed and regulated ADM.

  5. States that it supports the Exposure Draft Privacy Amendment (Personal Data Protection) Bill 2026, subject to amendments.

    HTI supports the Exposure Draft Privacy Amendment (Personal Data Protection) Bill 2026, subject to some amendments outlined below.

  6. Argues that privacy reforms should be packaged with increased OAIC funding.

    New reforms should be packaged with increased funding for the OAIC so that they can continue to fulfil their remit.

Recommendations

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

  1. Include a training-data transparency requirement in the Australian Standards for AI, with public input through the upcoming public consultation on the Standards.

    The Australian Government should include a training-data transparency requirement in the Australian Standards for AI, which are currently being developed. In developing this requirement, the Government should provide an opportunity for public input via its upcoming public consultation process on the Standards.

  2. Require companies to publish a high-level training-data summary when an AI model is made available in Australia.

    A TDT requirement would require companies to publish a high-level summary of the data used for training an AI model when it is made available in Australia.

  3. Model the training-data transparency requirement on comparable laws already operating in jurisdictions such as California.

    HTI proposes that these Standards should include a training data transparency requirement, modelled on similar laws already operating in jurisdictions like California.

  4. Introduce a requirement in the Fair Work Act that any worker monitoring be 'fair and reasonable', setting out the relevant factors.

    introducing a requirement in the Fair Work Act that any worker monitoring be ‘fair and reasonable’ outlining the relevant factors that apply to this context

  5. Establish a dedicated complaints stream in the Fair Work Commission to hear privacy-related worker complaints.

    establishing a dedicated complaints stream in the Fair Work Commission to hear privacy-related worker complaints

  6. Amend the Fair Work Act so enterprise bargaining agreements and Modern Awards can build on the proposed 'fair and reasonable' protections.

    amending the Fair Work Act to ensure that enterprise bargaining agreements and Modern Awards can build on the protections in the proposed ‘fair and reasonable’ requirement

  7. Include a default consultation clause for new surveillance in EBAs and modern awards; allow tailored negotiated clauses without undermining the fair-and-reasonable minimum protections.

    If the proposed ‘fair and reasonable’ requirement were included in the Fair Work Act, it would be logical to also include a provision that applies to consultation on worker surveillance in enterprise bargaining agreements (EBAs) and modern awards. A model consultation clause for EBAs and awards could provide for consultation where an employer proposes to introduce new surveillance tools or methods. The model consultation clause would serve as the default, with unions and employers also having the opportunity to negotiate tailored worker surveillance clauses articulating what is fair and reasonable in particular industries and workplaces. Negotiated worker surveillance clauses would be restricted from undermining the minimum protections outlined in the fair and reasonable test.

  8. Support regulators such as the Fair Work Ombudsman to issue practical guidance on applying new and existing laws to worker surveillance.

    supporting regulators such as the Fair Work Ombudsman to issue practical guidance on how to apply new and existing laws in the worker surveillance context

  9. Coordinate with states and territories to harmonise worker-surveillance law and reduce regulatory overlap.

    The reform process should harmonise Australian law and reduce areas of regulatory overlap – which will require coordination with states and territories.

  10. Introduce a Bill implementing Robodebt Royal Commission recommendation 17 to regulate government ADM, with safeguards for effective, lawful and accountable operation.

    The Australian Government should fulfill its commitment to implement recommendation 17 of the Robodebt Royal Commission by introducing a Bill to regulate automated decision making in government. The Bill should contain the necessary safeguards to ensure that ADM operates effectively, lawfully and accountably by requiring:

  11. Require Parliament, rather than a minister or unelected official, to decide which categories of government decisions may be automated.

    Parliament should determine which types of decision may be automated. The determination whether to allow automation for a category of government decision making (such as the availability of services under the National Disability Insurance Scheme) should be made by Parliament, given that this enables effective democratic oversight. This responsibility should not be delegated to a Minister or an unelected official.

  12. Generally prohibit ADM for discretionary elements or where it may narrow discretion; exceptional automation of discretion requires specific justification and greater oversight.

    Generally, decisions involving discretion or the exercise of evaluative judgment should not be automated. The ADM legal framework should make clear that the default position is that ADM should not be used for any discretionary element of the decision-making process, nor should it be used where there is a likelihood that it will narrow the exercise of discretion. While there may be exceptional circumstances, which could justify automating discretion (for example, by enabling favourable discretionary decisions to be made via ADM), this requires specific justification and greater oversight.

  13. Preserve accessible, appropriate review of automated decisions, including senior-official internal review, external merits review and judicial review, without reducing existing review rights.

    People should be able to seek review of automated decisions. The law should enable a person who wishes to challenge an automated decision to do so using accessible and appropriate review processes. The use of ADM should not have the effect of denying or reducing review rights for people affected by government decisions. If a decision would be reviewable if made by a human, it should also be reviewable if made through automation. This should include: - internal review by a more senior official, which is the most common, cost-effective and accessible pathway to address inaccurate or unfair decisions36 - extend to external merits review (by a tribunal) - judicial review.

  14. Give anyone affected by ADM reasons; provide a plain-language legal and factual explanation when a decision affects legal or human rights or otherwise significantly impacts them.

    Anyone affected by an automated decision should be given reasons for the decision. When a decision is made affecting a person’s legal or human rights, or otherwise significantly impacting them, they should be provided with a plain language explanation of the legal and factual basis for the decision.

  15. Before adopting ADM, assess affected people’s human rights, particularly vulnerable or marginalised groups; do not implement systems with unjustified human rights impacts.

    The Government should give proactive consideration to human rights, including impacts on vulnerable and marginalised communities. Before adopting ADM, government should be required to proactively assess the human rights of those likely to be affected, including giving particular consideration to groups experiencing vulnerability or marginalisation. ADM systems that have an unjustified impact on human rights should not be implemented.

  16. Pass the exposure draft of the Privacy Amendment (Personal Data Protection) Bill 2026 with HTI's proposed amendments (its Appendix B).

    The exposure draft of the Privacy Amendment (Personal Data Protection) Bill 2026 should be passed with HTI’s proposed amendments (see Appendix B) to ensure that the reform makes good on the Australian Government’s prior commitments to privacy law reform.

  17. Amend the Privacy Bill to require the Privacy Commissioner to develop a code on biometric and other surveillance technologies, modelled on s 26GC of the current Privacy Act.

    a requirement that the Privacy Commissioner develop a code on biometric and other surveillance technologies, modelled on s 26GC in the current Privacy Act

  18. Amend the Privacy Bill to include a direct right of action where a regulated entity breaches the Australian Privacy Principles.

    a direct right of action where a regulated entity breaches the Australian Privacy Principles

  19. In the medium term, implement HTI’s Facial recognition technology: Towards a model law recommendations, balancing safeguards against FRT harms with positive innovation.

    HTI has previously outlined a model law for FRT in Facial recognition technology: Towards a model law.41 In the short term, HTI recommends that the exposure draft be amended to empower the Privacy Commissioner to develop a code under the Privacy Act for biometric and surveillance technologies, with a view to addressing the risks associated with smart glasses and other similar products and services. In the medium term, HTI recommends that the Government implement the recommendations from this report, which would introduce necessary safeguards, and set an appropriate balance that addresses the risks of harm associated with FRT, while encouraging positive innovation with FRT.

  20. Implement outstanding Privacy Act reform commitments before the next Federal election.

    The Australian Government should implement outstanding Privacy Act reform commitments prior to the next Federal election.

  21. Repeal the small business exemption in the Privacy Act, which it says excludes approximately 94% of businesses from compliance.

    the small business exemption which excludes approximately 94% of businesses in Australia from compliance with the Privacy Act.43 This anomalous exemption leaves consumers exposed and unprotected, and should be repealed.

  22. Resource the OAIC adequately so new and existing privacy protections are understood and enforced.

    The Australian Government should make provision to resource the OAIC adequately to ensure that new and existing privacy protections are appropriately understood and enforced.

  23. Restore APP 3.2(b)’s reasonable-necessity standard for organisations, and reasonable-necessity or direct-relationship standard for agencies, instead of merely requiring a relationship to functions or activities.

    This differs from the Government's original proposed wording which was “whether the collection, use or reliance by APP entities disclosure is reasonably necessary for the functions and activities of the on consent to avoid their privacy obligations, especially where such consent is often not genuine e ensure that APP entities give greater weight to the privacy rights of individuals before handling their personal information. organisation or is reasonably necessary or directly related for the functions and activities of the agency” (emphasis added).

  24. Redraft APP 3.2(e) to consider whether consent was actually given, rather than merely whether genuine choice was offered.

    Second, proposed new APP 3.2(e) asks whether the individual is ‘provided with genuine choice’ in relation to the collection, use or disclosure of information. HTI recommends that proposed APP 3.2(e) be redrafted to require consideration of whether the individual has given consent.

  25. Include the kind, sensitivity and amount of personal information as a standalone fair-and-reasonable factor or an explicit part of proportionality under APP 3.2(f).

    Thirdly, the proposed fair and reasonable test excludes one original factor from the AGD Report — a requirement for entities to consider ‘the kind, sensitivity and amount of information being collected, used or disclosed’. The amount and nature of personal data being handled are directly relevant to whether the APP entity is acting fairly and reasonably. For example, collecting more personal data increases the risk to the person’s privacy, and this should be justified. Hence, this consideration should be included either as a standalone fair and reasonable factor as originally intended, or as an explicit element of the proportionality factor (proposed APP 3.2(f)).

  26. Make redrafted APP 3.2(b), redrafted APP 3.2(e) and APP 3.2(f) essential requirements, rather than only holistic considerations.

    Finally, HTI notes that the proposed fair and reasonable requirements are currently articulated as non-essential indicia to be considered holistically — entities would not have to satisfy every legislated factor to meet their obligations. HTl recommends that some of the fair and reasonable factors be classified as essential indicia. This would ensure that entities take steps to comply with baseline considerations when handling personal information. HTI| recommends that proposed APP 3.2(b) (redrafted as above), APP 3.2(e) (redrafted as above) and APP 3.2(f) be treated as essential requirements.

  27. Retain current APP 6’s purpose limitation.

    Purpose limitation is a vital privacy protection, because it sets a default restriction on how personal information can be handled, based on an APP entity’s most relevant justification — its ‘primary purpose’. Handling personal information for a secondary purpose should require more specific justification — as the Privacy Act currently provides. Removing this protection would weaken the Privacy Act. Therefore, current APP 6 should be retained.

  28. Remove or narrow the new sensitive-information consent exceptions.

    The new exceptions to requirements to obtain consent for the collection of sensitive information should be removed, or at least more narrowly scoped.

  29. If retaining the publicly-available-document exception, exclude information behind a barrier to access.

    If this exception is to be retained, the definition of publicly available documents should be amended to exclude information where there is a ‘barrier to access’.

  30. Remove the sensitive-information consent exception for strictly necessary provision of explicitly requested goods or services.

    Proposed APP 4.3 would provide that consent is not required where collection is strictly necessary to provide or deliver a good or service explicitly requested by an individual, and where the good or service cannot be provided in a less privacy intrusive manner. In HTI’s view, it is no less important for individuals to know and consent to the use of their sensitive information, regardless of whether they've requested the relevant good/service or not. This exception should therefore be removed.

  31. If current APP 3.5 is removed, delete ‘collection’ from proposed APP 3.3(b), preserving protection for collection in permitted general situations.

    Proposed APP 3.3(b) provides that the new fair and reasonable requirement would not apply to PGSs. The Exposure Draft would also repeal APP 3.5, which would remove the requirement that collection of personal information be ‘lawful and fair’ — even in a PGS. If current APP 3.5 is removed, then proposed APP 3.3(b) should be amended to delete the word ‘collection’. This would have the effect of bringing the current protection across.

  32. Define ‘wrongdoing’ as serious misconduct and add a statutory note covering suspected exploitation of customer vulnerability or misuse of enduring power of attorney for personal gain.

    HTI recommends that the Bill define ‘wrongdoing’ as behaviour that amounts to serious misconduct, and it should also include a statutory note to make clear that wrongdoing in this context would include where an entity suspects a person is exploiting a customer’s vulnerability, or misusing enduring power of attorney for personal gain.

  33. Delete ‘for sensitive information’ from item 1, column 3 of the permitted-general-situations table so its exception is not widened.

    Finally, HT| recommends that in the PGS table, Item 1 column 3 should delete the words ‘for sensitive information’ so that the exception is not widened beyond its current articulation in the Privacy Act.

  34. Retain APP 5.2 notice-content requirements.

    HTl recommends that notice content current APP 5 for the requirements in APP 5.2 be retained.

  35. Narrow the ad-supported-services exception to direct-marketing opt-outs appropriately, or remove it if it cannot be narrowed.

    There may be cases where it is reasonable for ad-supported services to offer an alternative experience for those who opt out of direct marketing — for example, customers that do not opt out of direct marketing on a shopping site may receive personalised recommendations or feeds; customers who opt out would not have access to this feature. However, the Exposure Draft Bill contains a wider exception than this. Its breath would undermine the protective intent of the provisions. If the exception cannot be appropriately narrowed, it should be removed.

  36. Remove the restriction of the right to erasure to large digital platforms.

    The Exposure Draft would allow this right to be exercised only as against ‘large digital platforms’ (with gross annual revenue of $500 million or more, or with 2.5 million or more average monthly end users). There is no clear practical or principled justification for this restriction to the right to erasure. The qualification that the right only applies to ‘large digital platforms’ should be deleted.

  37. Grant the OAIC power to issue substantiation notices.

    HT recommends that the OAIC be granted the power to issue substantiation notices as part of this package.

Read the original

Open in the interactive guide