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 327

Professionals Australia

Submitter group: Unions.

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.

Professionals Australia, a union of professional employees, proposes a risk-based National AI Practitioner Registration Scheme (which it asks to help design) and mandatory workplace Labour Impact Assessments, with just-transition obligations, professional sign-off of safety-critical AI outputs and rebuilt in-house Commonwealth capability. It says these would support innovation while protecting safety, rights and job quality.

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. Argues that AI practitioner registration would identify responsibility for AI-specific functions and prevent accountability disappearing between parties.

    AI practitioner registration would identify responsibility for AI-specific functions and prevent accountability disappearing between a vendor, deployer, consultant and end user.

  2. Argues that employers should complete a Labour Impact Assessment before introducing AI that may significantly affect workers.

    Before introducing AI that may significantly affect workers, employers should complete a workplace Labour Impact Assessment.

  3. Argues that workers who reasonably decline to certify work they cannot verify should be protected from adverse action.

    A worker who reasonably declines to certify work they cannot verify should be protected from adverse action.

  4. Argues that employers should provide consultation, paid training, income protection and redeployment where AI changes or displaces work.

    Where AI changes or displaces work, employers should provide early consultation, paid training, career-transition assistance, appropriate income protection and redeployment before redundancy.

  5. Argues that AI replacing entry-level tasks may weaken the pathways through which graduates build professional capability.

    If AI replaces entry-level tasks, it may weaken the pathways through which graduates acquire skills and Australia develops future professional capability.

  6. Argues that technical expertise should be treated as sovereign capability rather than purchased project by project.

    Engineering, scientific, data and AI expertise should be treated as sovereign capability, not merely purchased when a project arises.

Recommendations

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

  1. Establish a risk-based National AI Practitioner Registration Scheme, developed in partnership with Professionals Australia (the submitter), for people performing defined high-risk AI functions.

    The Australian Government establish a risk-based National AI Practitioner Registration Scheme, developed in partnership with Professionals Australia, for people performing defined high-risk AI functions.

  2. Require employers to undertake and regularly review Labour Impact Assessments with affected workers and their representatives before introducing AI that may materially affect workers.

    Employers be required to undertake and regularly review a workplace Labour Impact Assessment, with the participation of affected workers and their representatives, before introducing AI that may materially affect workers.

  3. Require employers introducing AI to provide paid training, preserve professional development pathways, prioritise redeployment and support workers whose jobs materially change or disappear.

    Employers introducing AI be required to provide paid training, preserve professional development pathways, prioritise redeployment and support workers whose jobs materially change or disappear.

  4. Require suitably qualified professional review and sign-off for engineering, scientific, clinical or other safety-critical AI outputs; protect professionals declining to certify work they cannot reasonably verify.

    AI-generated outputs used in engineering, scientific, clinical or other safety-critical work require review and sign-off by a suitably qualified professional, who must be protected if they decline to certify work they cannot reasonably verify.

  5. Rebuild and maintain in-house Commonwealth engineering, scientific, data and AI capability.

    The Commonwealth rebuild and maintain in-house engineering, scientific, data and AI capability as part of Australia's sovereign capability.

  6. Attach registration to material responsibility for high-risk architecture, training, deployment approval, validation, assurance, monitoring, incident response and independent audit; exclude routine general-purpose tool use.

    Registration should attach to defined high-risk functions, not routine use of a general-purpose tool. Covered functions should include material responsibility for system architecture, model training, deployment approval, validation, assurance, monitoring, incident response and independent audit.

  7. Cover high-risk employment/worker management, safety-critical engineering, health/clinical decisions, essential public services, policing/compliance, critical infrastructure and decisions materially affecting legal rights or livelihood.

    High-risk uses should include employment and worker management, safety-critical engineering, health and clinical decisions, essential public services, policing and compliance, critical infrastructure, and decisions materially affecting legal rights or livelihood.

  8. Require relevant qualifications or demonstrated competence for safe, lawful and ethical high-risk AI work, including risk management for workers, customers and the public.

    Relevant qualifications or demonstrated competence to develop, Entry standards train, deploy or oversee high-risk AI safely, lawfully and ethically, including managing risks to workers, customers and the public

    A program corrected obvious text-recognition errors in this quote for display.

  9. Require continuing professional development on changing models, failures, bias, security, automation bias, over-reliance, human oversight and applicable law, without limiting it to those topics.

    Mandatory continuing professional development covering, but not limited to, changing models, known failure modes, bias, security, automation bias, over-reliance, effective human oversight and applicable law

  10. Enforce ethical duties of competence, honesty and public interest, records of significant decisions/risks/safeguards, and serious-risk reporting to someone authorised to act.

    An enforceable code requiring practitioners to act competently, honestly and in the public interest; keep clear records of significant decisions, risks and safeguards; and report serious risks to someone with authority to act.

  11. Keep auditable records naming registered practitioners responsible for high-risk design, deployment, assurance and monitoring decisions.

    An auditable record of the registered practitioners responsible for Named accountability high-risk design, deployment, assurance and monitoring decisions.

  12. Require risk-based audit, incident reporting and periodic revalidation, supported by access to systems, data and documentation.

    Risk-based audit, incident reporting and periodic revalidation, Independent assurance . P supported by access to systems, data and documentation.

  13. Provide powers to investigate, impose conditions, suspend/cancel registration and refer serious breaches for civil or criminal action.

    Powers to investigate, impose conditions, suspend or cancel Enforcement . A . — a . registration and refer serious breaches for civil or criminal action.

  14. Protect workers against adverse action for raising risks, declining unsafe deployment or refusing to certify work that cannot reasonably be verified.

    Protection from adverse action for raising risks, declining unsafe Worker protection deployment or refusing to certify work that cannot reasonably be verified.

  15. Complement existing professional registration/licensing and retain regulated professionals’ accountability under their governing frameworks.

    The scheme should complement existing registration and licensing. An engineer, health practitioner or other regulated professional would remain accountable under their governing framework.

  16. Give professionals responsible for AI-assisted sign-off the information, time and authority to verify outputs.

    Competent human review must be genuine. A professional responsible for signing off an AI- assisted output must have the information, time and authority needed to verify it.

  17. Undertake Labour Impact Assessments early enough for workers and representatives to influence proposals.

    Before introducing AI that may significantly affect workers, employers should complete a workplace Labour Impact Assessment. It should be undertaken early enough for workers and their representatives to influence the proposal.

  18. Require assessments where AI may materially affect employment, duties, classifications, staffing, workload, performance standards, monitoring, safety, professional responsibility, skills or career development.

    A workplace Labour Impact Assessment should be completed where a proposed AI system may materially affect employment, duties, classifications, staffing, workload, performance standards, monitoring, safety, professional responsibility, skills or career development.

  19. Review assessments after material upgrades, new uses, serious incidents, unanticipated harm or cumulative effects of smaller deployments.

    The assessment should also be reviewed after a material system upgrade, expansion to a new use, a serious incident, evidence of unanticipated harm, or cumulative effects from several smaller deployments.

  20. Require employers to involve workers/representatives before decisions, explain proposed AI/limitations/impacts, consider alternatives/safeguards/training/redeployment, share findings and workers’ influence, and review/respond after implementation.

    Employers should: • involve affected workers and their representatives before final decisions are made; • explain the proposed use of AI, its limitations and its expected workforce impacts; • consider alternatives, safeguards, training and redeployment options; • share the findings and explain how workers’ views affected the final proposal; and • review the assessment after implementation and respond to emerging impacts

  21. Assess changes to tasks, duties, staffing, classifications and job security; examine work intensification, performance targets and physical/psychosocial risks.

    Work and jobs: changes to tasks, duties, staffing, classifications, and job security. • Workload and safety: work intensification, performance targets, and physical and psychosocial risks.

  22. Assess training, redeployment, graduate development and career progression.

    Skills and careers: training, redeployment, graduate development and career progression.

  23. Assess employee data collection/use/storage/sharing, monitoring necessity/proportionality and how workers learn about decisions affecting them.

    Data and privacy: how employee data will be collected, used, stored and shared; whether monitoring is necessary and proportionate; and how workers will be informed about decisions affecting them.

  24. Assess disadvantage by gender, gender identity, sexuality, race, disability, age, caring responsibilities, socioeconomic background, employment status or location, including intersecting disadvantages.

    Equality and inclusion: whether the system may disadvantage people because of gender, gender identity, sexuality, race, disability, age, caring responsibilities, socioeconomic background, employment status or location, including where disadvantages intersect.

  25. Assess human oversight, professional autonomy, decision review, worker concerns and how productivity gains improve work or services.

    Accountability and benefits: human oversight, professional autonomy, review of decisions, worker concerns and how productivity gains will improve work or services .

  26. Include graduate employment, supervised learning, critical skills and career progression in Labour Impact Assessments.

    Labour Impact Assessments should therefore consider graduate employment, supervised learning, critical skills and career progression.

  27. Where AI changes/displaces work, require early consultation, paid training, career-transition assistance, appropriate income protection and redeployment before redundancy; workers should not bear the cost of technology-driven change.

    Where AI changes or displaces work, employers should provide early consultation, paid training, career-transition assistance, appropriate income protection and redeployment before redundancy. Workers should not bear the cost of technology-driven change.

  28. Have the Commonwealth maintain graduate and early-career pathways and report on whether agencies and regulators have the skills to assess high-risk systems, infrastructure demands and vendor claims.

    The Commonwealth should rebuild in- house capability, maintain graduate and early-career pathways, and report on whether agencies and regulators have the skills needed to assess high-risk systems, infrastructure demands and vendor claims.

  29. Require public procurement and funding to include Australian professional participation, supervised development opportunities and knowledge transfer, extending through major contractors and supply chains.

    Public procurement and funding should reinforce this capability by requiring Australian professional participation, supervised development opportunities and knowledge transfer. These requirements should extend through major contractors and supply chains so that publicly supported AI investment builds lasting Australian expertise.

  30. Give Professionals Australia a central scheme-design role covering scope, entry standards, continuing development, ethical obligations and accountability, consulting regulators, professional bodies, unions and technical experts.

    The Australian Government should work with Professionals Australia to develop and implement the National AI Practitioner Registration Scheme. Professionals Australia should have a central role in designing the scheme’s scope, entry standards, continuing professional development requirements, ethical obligations and accountability arrangements, in consultation with relevant regulators, professional bodies, unions and technical experts.

  31. Develop a national Labour Impact Assessment framework with unions, employers and regulators; require workplace processes with affected workers/representatives before significant AI and as impacts emerge, and review the framework after two years.

    The Australian Government should develop a practical national framework for workplace Labour Impact Assessments in consultation with unions, employers and relevant regulators. Each workplace should establish a process for completing and reviewing assessments with affected workers and their representatives before significant AI is introduced and as its impacts emerge. The national framework should be reviewed after two years.

Read the original

In some quotes a program corrected obvious text-recognition errors for display.

Open in the interactive guide