Submission 213
Australian Federation of Disability Organisations (AFDO)
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.
Argues the Terms of Reference omit people with disability and asks the Committee to name them as a stakeholder group. Drawing on the Robodebt Royal Commission, it addresses workforce effects, government AI accountability, gaps in existing law, and structural AI bias and harm affecting people with disability.
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 the Terms of Reference do not name people with disability as stakeholders in any of their fourteen paragraphs.
The ToR do not name people with disability as stakeholders at any point across their fourteen paragraphs.
Argues AI-driven labour market change may fall disproportionately on workers with disability, given existing employment concentration in routine roles.
People with disability already experience substantially lower workforce participation and higher underemployment than the general population. Displacement from routine or process-based roles is likely to fall disproportionately on people with disability, given existing employment concentration of people with disability in those role types
Argues Robodebt Royal Commission Recommendation 17.1, calling for a consistent legal framework for government automation, equally applies to Commonwealth AI use.
The Royal Commission's Recommendation 17.1 called for legislative reform establishing a consistent legal framework for government automation. This recommendation equally applies to Commonwealth AI use.
Argues that algorithmic and AI harm to people with disability should be treated as a distinct risk category under paragraph (k), separate from fraud and criminal misuse.
We submit the Committee should treat algorithmic and AI harm to people with disability as a distinct risk category under paragraph (k), separate from fraud and criminal misuse.
Argues that models are trained for the statistical centre of population data, while people with disability are disproportionately represented in the tails.
AI models are trained and validated to perform well against the statistical centre of a population's data. People with disability are disproportionately represented in the tails of that distribution,
Argues transparency alone is insufficient and that systems should be tested and reported against disability-specific accuracy and error-rate benchmarks before deployment.
Disclosure that an AI system is in use does not by itself correct for a model that structurally underperforms for users with disability. The Committee must distinguish procedural transparency from a substantive requirement that systems be tested and reported against disability-specific accuracy and error-rate benchmarks before deployment.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Name people with disability as a stakeholder group under paragraphs (h) and (k) and invite evidence on AI and automated decision-making effects, including in the NDIS and aged care assessment tools.
We recommend the Committee explicitly includes people with disability as a named stakeholder group under paragraphs (h) and (k) and invites evidence on AI (and automated decision-making) and its potential effects on people with disability, including in disability-specific programs, such as the NDIS and aged care assessment tools.
Specifically examine the effect of AI-driven labour market change on workers with disability.
We submit the Committee should specifically examine: the effect of AI-driven labour market change on workers with disability.
Examine whether AI adoption is increasing or reducing employers' willingness to provide workplace adjustments, and any change in employer incentives to employ or retain workers with disability.
whether AI adoption is increasing or reducing employers’ willingness to provide workplace adjustments. AFDO is concerned that some AI-enabled tools may be seen as an adequate substitute for adjustments. This possibility, and any consequential changes in employer incentives to employ or retain workers with disability, needs to be carefully considered and included in the Committee’s recommendations
Examine the adequacy of education and vocational training systems, including adult and vocational retraining, in supporting people with disability into AI-adjacent roles.
the adequacy of education and vocational training systems, including adult and vocational retraining pathways, in supporting people with disability to move into AI-adjacent roles, rather than assuming displaced workers with disability will be reabsorbed into roles elsewhere in the economy.
Establish a consistent government-AI legal framework under Robodebt Recommendation 17.1, with affected individuals’ review pathways, plain-language public AI disclosure, and business rules/algorithms available for independent expert scrutiny.
The Royal Commission's Recommendation 17.1 called for legislative reform establishing a consistent legal framework for government automation. This recommendation equally applies to Commonwealth AI use. This framework must include a clear path for affected individuals to seek review, plain-language public disclosure that AI is used, and business rules and algorithms are made available for independent expert scrutiny.
Ensure government AI use addresses actual, perceived or threatened bias and is accessible and usable by all, including people with disability.
It is also essential that any use of AI by government addresses issues of actual, perceived, or threatened bias in the operation or use, and accessibility and usability by all, including people with disability.
Accompany Commonwealth AI adoption with a body holding responsibility and all necessary power to monitor and audit AI use for fairness, bias and client usability (Robodebt Recommendation 17.2).
Consistent with Recommendation 17.2 of the Robodebt Royal Commission, any adoption of AI by the Commonwealth must be accompanied by a Commonwealth body that has responsibility, and all necessary power, to monitor and audit the use of AI for fairness, bias and client usability.
Test current/proposed Commonwealth AI use across Services Australia, DSS, Health, Disability and Ageing, the NDIA and their programs/processes against Robodebt Recommendations 17.1 and 17.2.
must test current and proposed Commonwealth AI use (including by Services Australia, the Department of Social Services, the Department of Health, Disability and Ageing, and the National Disability Insurance Agency and within any programs or processes those agencies have responsibility for) against Recommendations 17.1 and 17.2.
Seek evidence on and consider whether Commonwealth AI use currently affecting disability support, income support or aged care eligibility is consistent with the disclosure and review-pathway requirements of Recommendation 17.1.
whether any Commonwealth use of AI (whether as part of or to assist decision- making or otherwise) currently affecting disability support, income support, or aged care eligibility is consistent with the disclosure and review-pathway requirements of Recommendation 17.1
Seek evidence on Commonwealth compliance with Robodebt Recommendation 17.2, including whether any proposed body’s powers and responsibilities would extend to government AI systems or use.
the status of the Commonwealth’s efforts to comply with Recommendation 17.2, and whether the powers and responsibilities of any body proposed by the Commonwealth would extend to AI systems or use within government, and
Seek evidence on and consider whether affected individuals, including people with disability, have an accessible, supported pathway to challenge the use of AI, not merely a formally available one.
whether affected individuals, including people with disability, have an accessible, supported pathway to challenge the use of AI, not merely a formally available one.
Explicitly capture AI, including ADM, in Commonwealth welfare and disability administration as a test case for regulatory adequacy under paragraph (i).
We submit that paragraph (i) needs to explicitly capture AI, including in the context of ADM, in Commonwealth welfare and disability administration as a test case for regulatory adequacy.
Have regard to the Automated Decision-Making (Safeguards and Transparency) Bill 2026 and the AI Kill Switch and Data Centre Control Bill 2026 in its deliberations and recommendations.
We also urge the Committee to have regard in its deliberations and recommendations to: 1. the Automated Decision-Making (Safeguards and Transparency) Bill 2026, introduced into the House of Representatives on 7 September 2026 by the independent member for Curtin, and 2. the AI Kill Switch and Data Centre Control Bill 2026, introduced into the House of Representatives on 7 September 2026 by the independent member for Calare
Examine the absence of a legislated, consistent framework governing when and how AI may be used in decisions, particularly those affecting income support, disability support and aged care eligibility.
the absence of a legislated, consistent framework governing when and how AI may be used in decisions, particularly those affecting income support, disability support, and aged care eligibility, as identified by Recommendation 17.1
Examine the adequacy of current review and appeal mechanisms, including additional barriers people with disability may face in understanding or accessing them.
the adequacy of current review and appeal mechanisms, including any additional barriers that people with disability may face in understanding or accessing those mechanisms, and
Examine whether forthcoming Privacy Act reforms on substantially automated decisions are sufficient on their own, or whether AI-specific or disability-specific procedural protections are also required.
whether forthcoming Privacy Act reforms addressing "substantially automated decisions" are sufficient on their own, or whether AI-specific or disability-specific procedural protections are also required.
Treat algorithmic and AI harm to people with disability as a distinct risk category under paragraph (k), separate from fraud and criminal misuse.
We submit the Committee should treat algorithmic and AI harm to people with disability as a distinct risk category under paragraph (k), separate from fraud and criminal misuse.
Seek evidence on the differential impact of scams and deepfakes on people with disability, including those with cognitive disability or limited digital literacy support.
the differential impact of scams and deepfakes on people with disability, including those with cognitive disability or limited digital literacy support, who may face heightened vulnerability to AI-enabled fraud, and
Seek evidence on government/private AI decisions about disability eligibility, support or access, including cumulative psychological and financial harm documented by the Robodebt Royal Commission.
harm arising from government and private-sector AI systems making or informing decisions about eligibility, support, or access for people with disability, including the cumulative psychological and financial harm documented by the Robodebt Royal Commission.
Distinguish procedural transparency from a substantive requirement that systems be tested and reported against disability-specific accuracy and error-rate benchmarks before deployment.
The Committee must distinguish procedural transparency from a substantive requirement that systems be tested and reported against disability-specific accuracy and error-rate benchmarks before deployment.
Examine whether any Commonwealth body requires and assesses disability-disaggregated accuracy or error-rate testing or data before deploying AI systems.
We recommend the Committee seeks evidence on and considers; in the context of any recommendations it may make whether: a. any Commonwealth body currently requires and assesses disability- disaggregated accuracy or error-rate testing or data prior to deploying AI systems,
A program corrected obvious text-recognition errors in this quote for display.
Examine whether the monitoring and audit function proposed under Robodebt Recommendation 17.2 would have the mandate and technical capability to assess disability-disaggregated accuracy or error-rate testing.
We recommend the Committee seeks evidence on and considers; in the context of any recommendations it may make whether: a. any Commonwealth body currently requires and assesses disability- disaggregated accuracy or error-rate testing or data prior to deploying AI systems, and b. whether the monitoring and audit function proposed under Recommendation 17.2 of the Robodebt Royal Commission would have the mandate and technical capability to assess this.
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.