Submission 310
Australian Lawyers Alliance (ALA)
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.
Summary
Read the original submission for the full text.
Sets out the Australian Lawyers Alliance's position that AI must remain a tool used within the law, without diluting duties or obscuring responsibility. Recommends notice, reasons, registers and meaningful human review for AI in government and claims decisions, a risk-based national framework, evidence preservation, and safeguards for children and vulnerable people.
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.
Says individuals should be told in plain language when AI has materially influenced a decision.
The ALA strongly believes that it should be communicated to an individual, in plain language, when AI has materially influenced a decision.
Argues that legal responsibility for public functions should remain with an identifiable agency or office-holder despite outsourcing.
Legal responsibility for the exercise of a public function must remain with an identifiable department, agency or statutory office-holder. Outsourcing must not diminish the transparency, accountability or reviewability of that function.
Recommends Commonwealth entities and contractors maintain an accessible public register of material AI uses in administrative or service-delivery decisions.
Commonwealth departments, agencies and contracted service providers should maintain an accessible public register identifying each material use of AI in administrative or service-delivery decision-making.
Says existing laws apply to discrete aspects of AI use but remain fragmented and predominantly reactive, often requiring the affected person to identify the AI use and a legal wrong after harm.
Existing laws governing privacy, discrimination, consumer protection, administrative decision- making, negligence and evidence apply to discrete aspects of AI use. However, those laws remain fragmented and predominantly reactive, often requiring an affected person to identify both the use of AI and an actionable legal wrong after harm has occurred.
Argues that privacy law should protect inferred personal information and give rights to contest material inferences.
Privacy law should expressly protect inferred and derived personal information and provide individuals with effective rights to access, correct and contest any material inference used in decision-making.
Argues that where AI is likely to affect a child, the child's best interests should be a primary consideration.
Where an AI system is likely to affect a child, the best interests of the child should be a primary consideration in its design, deployment and regulation.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Make minimum safeguards enforceable and consistently applied to agencies, contractors and third-party systems where AI materially affects rights.
Where AI materially affects rights, however, minimum safeguards should be enforceable and apply consistently to agencies, contractors and third-party systems.
Tell individuals in plain language when AI has materially influenced a decision, with notice and reasons identifying the system's role, decisive information and inferences, responsible entity and avenues for correction and review.
it should be communicated to an individual, in plain language, when AI has materially influenced a decision. In these cases, there should be both notice and reasons identifying the system's role, the decisive information and inferences, the responsible entity and the available avenues for correction and review.
Do not allow confidentiality or IP claims to withhold information needed to challenge an adverse decision.
Claims of commercial confidentiality or intellectual property rights should not be relied upon to withhold information necessary for an affected person to understand the basis of, and effectively challenge, an adverse decision.
Maintain an accessible public register of each material use of AI in Commonwealth administrative or service-delivery decision-making, including by contracted providers.
Commonwealth departments, agencies and contracted service providers should maintain an accessible public register identifying each material use of AI in administrative or service-delivery decision-making.
Publish impact assessments for AI systems in high-risk contexts, including the NDIS and aged care, covering legality, data quality, discriminatory impacts, accessibility, monitoring and review. Include avenues for redress.
AI systems used in high-risk contexts, including the National Disability Insurance Scheme and aged care services, should also be subject to published impact assessments addressing legality, data quality, discriminatory impacts, accessibility, ongoing monitoring and avenues for review and redress.
Entitle people affected by a significant AI-assisted decision to prompt reconsideration by a qualified human decision-maker with all relevant material and authority to decide independently of the AI output.
A person affected by a significant AI- assisted decision should be entitled to prompt reconsideration by an appropriately qualified human decision-maker who has access to all relevant material, is authorised to reach an independent determination and is not bound by the AI-generated output.
Keep legal responsibility for public functions with an identifiable agency or office-holder, without outsourcing diminishing transparency, accountability or reviewability.
Legal responsibility for the exercise of a public function must remain with an identifiable department, agency or statutory office-holder. Outsourcing must not diminish the transparency, accountability or reviewability of that function.
Procurement and contractual arrangements should require preservation and disclosure of records necessary for audit, merits review and judicial review.
Procurement and contractual arrangements should require the preservation and disclosure of records necessary for audit, merits review and judicial review.
Embed AI safeguards at the design and procurement stages, test them before deployment and maintain them throughout operation.
safeguards should therefore be embedded at the design and procurement stages, tested before deployment and maintained throughout the operation of the system.
Supplement, rather than displace, existing rights and remedies with a nationally consistent, risk-based AI framework.
A nationally consistent, risk-based framework should supplement, rather than displace, existing rights, duties and remedies.
Treat AI systems that make, recommend or materially influence decisions on rights, entitlements, treatment, compensation or essential services as high-risk, with mandatory pre-deployment assessment and testing and continuing oversight, monitoring, record-keeping and review.
an AI system should be treated as high-risk where it makes, recommends or materially influences a decision affecting a person’s rights, entitlements, treatment, compensation or access to an essential service. Such systems should be subject to mandatory pre-deployment assessment and testing, together with continuing obligations concerning human oversight, monitoring, record-keeping and review.
Keep the entity that uses AI to exercise a function directly accountable to the affected person, without avoiding responsibility through third-party providers or supply chains.
the entity that uses AI to exercise a function or make a decision should remain directly accountable and should not be permitted to avoid responsibility by relying on a third- party provider or complex supply chain.
Allocate legal and regulatory responsibility to developers, deployers and decision-makers according to their control and capacity to prevent harm.
Legal and regulatory responsibility should attach to developers, deployers and decision-makers according to the control they exercise and their capacity to prevent harm.
Expressly protect inferred and derived personal information under privacy law, with effective rights to access, correct and contest material inferences used in decisions.
Privacy law should expressly protect inferred and derived personal information and provide individuals with effective rights to access, correct and contest any material inference used in decision-making.
Do not repurpose information provided for treatment, a claim or a statutory process to train or test AI without a clear lawful basis, demonstrated necessity and proportionate safeguards.
Information provided for treatment, a claim or a statutory process 3 Royal Commission into the Robodebt Scheme, Report (7 July 2023), https://robodebt.royalcommission.gov.au/publications/report. 7 should not be repurposed to train or test an AI system without a clear lawful basis, demonstrated necessity and proportionate safeguards.
Require insurers, statutory schemes and claims managers to give timely notice when AI has materially influenced a decision, with enough information to understand and challenge that influence.
An insurer, statutory scheme or claims manager should provide timely notice when AI has materially influenced a decision, together with sufficient information to enable the affected person to understand and challenge that influence.
Require adverse determinations on liability, eligibility, treatment, benefits, impairment or quantum to be made by a human decision-maker who has considered individual circumstances and can depart from the AI output.
Any adverse determination concerning liability, eligibility, treatment, benefits, impairment or quantum should be made by a human decision-maker who has considered the person’s individual circumstances and has authority to depart from the AI-generated output.
Require claims administrators using AI to preserve inputs, outputs, system versions, decision logs and records of human intervention for the life of the claim and any applicable review, appeal or litigation period.
Entities using AI in claims administration should preserve all relevant inputs, outputs, system versions, decision logs and records of human intervention for the life of the claim and the duration of any applicable review, appeal or litigation period.
Attach appropriate consequences, including adverse inferences where appropriate, to a failure without reasonable excuse to preserve or disclose information uniquely within an entity's control.
A failure, without reasonable excuse, to preserve or disclose information uniquely within an entity’s control should attract appropriate consequences under ordinary evidentiary and procedural principles, including the drawing of adverse inferences where appropriate.
Prevent outsourcing, proprietary restrictions or technical opacity from defeating rights of disclosure and review.
Rights of disclosure and review should not be defeated by outsourcing, proprietary restrictions or technical opacity.
Require meaningful disclosure and preserve meaningful access to human review.
An effective regulatory framework should therefore require meaningful disclosure and preserve meaningful access to human review.
Prevent entities gaining an evidentiary advantage from failing to preserve, disclose or explain systems within their control.
An entity should not obtain an evidentiary advantage from its own failure to preserve, disclose or explain the operation of a system within its control.
Keep an entity relying on an AI-generated output responsible for explaining and substantiating its decision.
Further, an entity that relies upon an AI-generated output should remain responsible for explaining and substantiating its decision.
Support consistent court and tribunal guidance on preparing, filing and using AI material, covering accuracy, relevance, provenance, authentication, reliability and appropriate expert evidence.
The ALA supports the development by courts and tribunals of consistent guidance concerning the preparation, filing and use of AI-generated material. That guidance should address obligations to verify accuracy and relevance, as well as questions of provenance, authentication and reliability, and the appropriate use of expert evidence.
Require an institutional party relying on a proprietary fraud-detection or authenticity system to challenge evidence to disclose and substantiate the material basis of its assessment.
Where an institutional party relies upon a proprietary fraud-detection or authenticity system to challenge evidence, it should be required to disclose and substantiate the material basis of that assessment.
Preserve and where necessary supplement existing rights through accessible, timely, affordable and practically enforceable remedies and redress.
The regulatory framework should preserve and, where necessary, supplement existing rights by providing accessible pathways to effective remedies and redress. 32. Those remedies must be timely, affordable and practically enforceable.
Do not deny effective redress merely because a system is technically complex or the necessary information is exclusively controlled by the deploying entity.
An affected person should not be denied effective redress merely because the operation of the relevant system is technically complex or information necessary to establish the claim remains within the exclusive control of the entity that deployed it.
Ensure any AI copyright framework makes rights practically enforceable through proportionate transparency and record-keeping obligations letting rights holders ascertain use of protected material, identify the responsible entity and pursue a remedy.
any framework must ensure that the rights it confers are capable of practical enforcement. This requires proportionate transparency and record-keeping obligations sufficient to enable rights holders to ascertain whether protected material has been used, identify the responsible entity and pursue an effective remedy.
Preserve lawful access to legislation, judgments, public records and other legal information in any copyright reform, including for legal advice, research and open justice.
Any reform should preserve lawful access to legislation, judgments, public records and other legal information, including their use for legal advice, representation, research and purposes associated with open justice.
Develop reforms affecting First Nations cultural and intellectual property or culturally sensitive material through direct, meaningful consultation with First Nations peoples, including relevant custodians and communities.
any reform affecting First Nations cultural and intellectual property or culturally sensitive material should be developed through direct and meaningful consultation with First Nations peoples, including the relevant custodians and communities.
Impose proportionate, risk-based obligations on platforms and service providers to preserve evidence and respond promptly to verified reports of harmful or unlawful content.
The ALA supports proportionate, risk-based obligations requiring relevant platforms and service providers to preserve evidence and respond promptly to verified reports of harmful or unlawful content.
Give affected persons clear and expeditious processes for takedown, correction and account recovery, including interim measures where there is a risk of serious or continuing harm.
Affected persons should have access to clear and expeditious processes for takedown, correction and account recovery, including interim measures where there is a risk of serious or continuing harm.
Where insurers, statutory authorities or institutions rely on AI fraud or authenticity assessments, require disclosure, an opportunity to respond, and genuine adverse determinations by appropriately qualified humans.
Where an insurer, statutory authority or other institution relies upon an AI- based fraud or authenticity assessment, procedural fairness requires disclosure of the material basis for that assessment and a reasonable opportunity for the affected person to respond. Any adverse conclusion should involve a genuine determination by an appropriately qualified human decision-maker.
Develop accessible guidance, specialist expertise and technical capability across the justice system to assess disputed digital material; avoid disproportionate individual costs, particularly where institutions or repeat litigants can better assess it.
Courts, tribunals, regulators and legal assistance services will require access to appropriate guidance, specialist expertise and technical capability to assess disputed digital material. The Australian Government should consider how that capability can be developed and made accessible across the justice system. The costs of responding to systemic verification risks should not be transferred disproportionately to individual litigants, particularly where an institution or repeat litigant is better placed to obtain and assess the relevant technical evidence.
Assess AI systems used in high-risk services for accessibility and discriminatory impacts before deployment, monitor them throughout operation, and maintain non-digital channels and supported decision-making.
AI systems used in high-risk services should therefore be assessed for accessibility and discriminatory impacts before deployment and monitored throughout their operation. Non-digital service channels and access to supported decision-making should also be maintained.
Make children's best interests primary where AI is likely to affect them, with age-appropriate design, privacy defaults, profiling and persuasive-design limits, accessible reporting and prompt action on harmful or abusive synthetic content.
Where an AI system is likely to affect a child, the best interests of the child should be a primary consideration in its design, deployment and regulation. Appropriate safeguards should include age-appropriate design, privacy-protective settings by default, limits on profiling and persuasive design, accessible reporting mechanisms and prompt action in response to harmful or abusive synthetic content involving children.
Make redress mechanisms accessible, trauma-informed and capable of prompt interim and final relief.
Redress mechanisms should be accessible, trauma-informed and capable of providing prompt interim and final relief.
Provide, where appropriate, evidence preservation, takedown, correction, account and identity restoration, and compensation for resulting loss or harm.
Available measures should include, where appropriate, evidence preservation, takedown, correction, account and identity restoration, and compensation for resulting loss or harm.
Minimise repeated disclosure and avoid requiring affected people to navigate multiple platforms, service providers and regulators to obtain urgent assistance.
Processes should minimise repeated disclosure and should not require an affected person to navigate multiple platforms, service providers and regulators to obtain urgent assistance.
Adopt a nationally consistent, enforceable and risk-based framework with sector-specific safeguards for government, insurance, statutory compensation and justice-related uses.
The ALA supports a nationally consistent, enforceable and risk-based framework, with sector- specific safeguards for government, insurance, statutory compensation and justice-related uses.