Submission 86
Western Australian AI Hub
Submitter group: Other.
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.
Proposes eight reforms grouped under human capability, sovereign and competitive markets, rights and accountability, and security and sustainable infrastructure. Topics include a sovereign educational commons, AI education and assessment reform, procurement settings, data provenance, public-sector automated decision-making safeguards, deception duties, risk-based security and a national coordination gateway.
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.
Argues that durable AI benefits depend on skills, contestable markets, accountable institutions, lawful data governance and infrastructure secure during disruption.
Durable benefits depend on a skilled workforce, contestable markets, accountable public institutions, trusted institutions, lawful data governance, and infrastructure that remains secure during operational disruption.
Recommends safeguarding core human capabilities while developing critical digital agency through stage-targeted learning and assessment across primary, secondary and tertiary education.
School education must safeguard core human capabilities while developing critical digital agency through stage-targeted learning and assessment frameworks across primary, secondary, and tertiary sectors.
Recommends proportionate disclosure of model and data provenance, hosting, beneficial ownership and exit arrangements in Commonwealth procurement rules and grants.
The Commonwealth Procurement Rules and grant programmes should require proportionate disclosure of model and data provenance, hosting and support locations, beneficial ownership, subcontractors, security controls, accessibility, incident notification, audit rights, data portability and termination assistance.
Recommends Commonwealth registers of material AI and automated decision-making systems, subject to necessary security and privacy protections.
Commonwealth entities should establish and maintain a public register of material AI and automated decision-making (ADM) systems (Western Australian AI Hub, 2026h), subject to necessary security and privacy protections.
Recommends risk-proportionate obligations with explicit protections for satire, artistic expression, journalism and legitimate political communication.
Obligations should remain proportionate to risk, ensuring explicit protections for satire, artistic expression, journalism and legitimate political communication.
Recommends a National AI Coordination Gateway connecting policy, governance, safety, regulatory, cyber and intelligence functions across existing government bodies.
A single National AI Coordination Gateway should connect policy, governance, safety, regulatory, cyber and intelligence functions across existing government bodies.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Pilot an optional publicly governed Sovereign Educational Commons for vetted AI services, open resources and literacy tools, with privacy, risk-proportionate Australian tenancy, accessibility, appropriate age assurance, affirmative-consent limits on secondary advertising/training use and independent pre-rollout evaluation.
Pilot an optional Sovereign Educational Commons: a publicly governed gateway to vetted AI services, open educational resources, and foundational AI literacy tools (Western Australian AI Hub, 2026c, 2026f). The pilot should enforce privacy-by-design principles, risk-proportionate Australian data tenancy requirements, accessibility standards, age assurance where appropriate, and an explicit prohibition on secondary use of participant data for advertising or model training without affirmative consent. Independent evaluation should precede any national rollout.
Fund a grant stream for SMEs, family businesses, regional providers, Aboriginal Community Controlled Organisations and educational providers to access advice, redesign work and upskill.
a grant stream should support small and medium-sized enterprises (SMEs), family businesses, regional service providers, Aboriginal Community Controlled Organisations (ACCOs) and educational providers to access trusted advice, redesign work and upskill workforces
Keep grant application processes streamlined, scale co-contributions to organisational capacity and require funded intermediaries to declare conflicts of interest.
Application processes should remain streamlined, co-contributions should scale with organisational capacity, and funded intermediaries should be required to declare conflicts of interest.
Adopt stage-targeted AI learning and assessment frameworks across primary, secondary and tertiary education.
School education must safeguard core human capabilities while developing critical digital agency through stage-targeted learning and assessment frameworks across primary, secondary, and tertiary sectors.
Keep AI-free early-learning reserves in primary education (K-6) and hold core mathematics and language offline through Year 4.
Primary education (K–6) should keep AI-free early-learning reserves so handwriting, tactile literacy and social development can form. Hold core mathematics and language offline through Year 4.
For secondary education, use a gradual-release framework and self-regulated digital-tool use, alongside age-tailored digital literacy and Socratic discussion addressing metacognitive risks without treating AI as inherently detrimental.
Secondary education (7-12) should focus on self-regulated use of digital tools and a gradual-release framework, not avoidance (Archer and Hughes, 2011; Cleary and Zimmerman, 2012; Panadero, 2017; Pintrich, 2004; Zimmerman, 2000, 2002). In parallel, teach age-tailored digital literacy and Socratic discussion models (OECD, 2026a, 2026b; Western Australian AI Hub, 2026f) that address metacognitive risks without treating digital tools, including AI tools, as inherently detrimental.
Treat two-lane assessment as a bridge and move toward programmatic assessment with oral defences, live demonstrations and unseen tasks.
Treat two-lane assessment as a bridge, not the destination (Lodge et al., 2025). Move toward programmatic assessment with built-in “defend” checkpoints: oral defences, live demonstrations and unseen tasks
Have publicly funded educational platforms embed intentional pedagogical friction so AI cannot serve as a silent and passive answer engine.
Publicly funded platforms should embed intentional pedagogical friction so AI cannot serve as a silent and passive answer engine
In tertiary and VET, prioritise programmatic assessment, authentic student engagement, academic integrity, structured organisational capabilities and cross-sector credit for flexible lifelong learning.
Tertiary and VET should prioritise programmatic assessment, authentic student engagement, academic integrity practices, structured organisational capabilities (Western Australian AI Hub, 2026j, 2026k) and cross-sector credit to establish flexible, lifelong learning pathways
Do not treat probabilistic AI-text detectors as sole evidence of misconduct.
Probabilistic AI-text detectors should not be sole evidence of misconduct, consistent with sector practice including Curtin University’s deactivation of automated detection features
Require proportionate disclosure in Commonwealth Procurement Rules and grant programmes of model and data provenance, hosting, ownership, subcontractors, security, audit rights and exit assistance; for essential services, test migration plans and continuity before contract award.
The Commonwealth Procurement Rules and grant programmes should require proportionate disclosure of model and data provenance, hosting and support locations, beneficial ownership, subcontractors, security controls, accessibility, incident notification, audit rights, data portability and termination assistance. For essential services, agencies should test migration plans and continuity before contract award.
Prioritise procurement and public-compute vendors demonstrating organisational readiness, governance controls and transparent Human-AI collaboration standards, allowing direct applications and publishing selection criteria, decisions and conflicts.
Procurement gateways and public compute programmes should prioritise vendors that demonstrate verified organisational readiness, robust governance controls, and transparent Human-AI collaboration standards (Western Australian AI Hub, 2026k), allowing direct applications while publishing selection criteria, decisions and conflicts.
Treasury should consult on a temporary, capped productivity incentive covering AI/robotics training, workflow redesign, cybersecurity, AI evaluation, research collaboration and domestic AI-related IP commercialisation.
The Treasury should consult on a temporary, capped AI Diffusion Incentive for expenditure that demonstrably improves Australian productivity: AI and robotics workforce training, workflow redesign, cybersecurity, AI evaluation, AI research collaboration and domestic AI-related IP commercialisation across key economic sectors such as agriculture, resources, health, advanced manufacturing, financial services and defence
Independently cost and review any incentive, requiring public benefit alongside the R&D Tax Incentive; assess where IP, skilled work, decision rights, tax value and operational capability remain, rather than Australian ownership alone.
Any Australian measure should be independently costed, periodically reviewed, and conditioned on a public-benefit test, in addition to the Research and Development Tax Incentive. Eligibility should not depend on Australian ownership alone; it should assess where IP, skilled work, decision rights, tax value and operational capability remain.
Mandate public-funding and procurement data-provenance registers detailing sources, legal bases, permissions, restrictions, retention and downstream reuse rights.
Public funding and procurement frameworks should mandate documented data-provenance registers detailing data sources, legal bases, usage permissions, restrictions, retention policies and downstream reuse rights (Western Australian AI Hub, 2026e, 2026h).
Prioritise narrow, domain-specific models trained on high-quality Australian datasets for local industry and government applications.
Alongside general data sovereignty, policy must prioritise narrow, domain-specific AI models trained on high-quality Australian datasets to deliver accurate, context-aware outcomes for local industry and government applications.
Support workable copyright licensing models while preserving fair dealing for research, study, criticism, review and reporting.
Policy should support workable licensing models while preserving fair dealing for research, study, criticism, review and reporting.
Require publicly funded Indigenous-data/knowledge AI projects to follow ICIP protocols, community-led governance and equitable benefit-sharing; require free, prior, informed consent for affected lands/resources/heritage/data, implemented through community-designated representative bodies rather than standardised consultation.
Publicly funded AI projects involving Indigenous data or cultural knowledge should adhere to Indigenous Cultural and Intellectual Property (ICIP) protocols, community-led governance structures and equitable benefit-sharing arrangements. Free, prior, and informed consent should be a core requirement for funding and procurement where projects affect Indigenous lands, resources, cultural heritage, or community data, consistent with the UN Declaration on the Rights of Indigenous Peoples (United Nations, 2007). Implementation should occur through representative bodies designated by affected communities rather than standardised consultation procedures (Barrowcliffe et al., 2025).
Establish a public register of material AI and automated decision-making systems in Commonwealth entities, subject to security and privacy protections.
Commonwealth entities should establish and maintain a public register of material AI and automated decision-making (ADM) systems (Western Australian AI Hub, 2026h), subject to necessary security and privacy protections.
Before high-impact agency deployments, publish algorithmic/human-rights assessments, confirm authority, assess vulnerable-group impacts and assign human responsibility, audit and incident response; provide notice, explanation, human review, administrative-law remedies and non-digital channels.
Before deploying in high-impact settings, agencies should complete and publish algorithmic and human rights impact assessments, confirm legal authority, evaluate impacts on vulnerable groups, assign human administrative responsibility, and establish ongoing audit and incident-response mechanisms (Western Australian AI Hub, 2026j). Affected people should receive notice, an explanation of material factors, human review and administrative-law remedies. Non-digital channels must remain open.
Enact high-risk synthetic-media and conversational-AI duties: expedited reporting/takedown for impersonation, fraud and non-consensual content; interaction disclosure; companion-service safety and protection of children, elderly and vulnerable people; evidence preservation and accessible human support.
Enact targeted statutory duties addressing high-risk synthetic media and conversational AI systems. Key measures should include expedited reporting and takedown mechanisms for AI-enabled impersonation, fraud and non-consensual synthetic content; clear disclosure when users interact with automated systems; safety-by-design standards and child, elderly, and vulnerable people’s protection duties for synthetic companion services (Western Australian AI Hub, 2026f, 2026g); evidence preservation requirements; and accessible human support options.
Keep obligations on synthetic media and conversational AI proportionate to risk, with explicit protections for satire, artistic expression, journalism and legitimate political communication.
Obligations should remain proportionate to risk, ensuring explicit protections for satire, artistic expression, journalism and legitimate political communication.
Enforce material deception and provenance verification in addition to automated content detection.
Enforce material deception and provenance verification (Western Australian AI Hub, 2026h, 2026i), in addition to automated content detection.
Require labels on materially deceptive synthetic audiovisual campaign content, a public repository of digital political advertising and rapid correction of false election information.
Reform should require labels on materially deceptive synthetic audiovisual campaign content, a public repository of digital political advertising, and rapid correction of false election information.
Classify government/critical-infrastructure AI by sensitivity, impact, autonomy, connectivity, supplier exposure and reversibility: controlled enterprise use; public tools without sensitive/personal/classified/commercially confidential data and independently verified outputs; prohibited/unapproved systems failing security/legal thresholds or under government direction.
Government and critical-infrastructure operators should classify AI use by information sensitivity, decision impact, autonomy, connectivity, supplier exposure and reversibility: ● Protected enterprise use: assessed systems operating under contractual, identity, logging, data-tenancy, encryption, testing and incident-response controls (Western Australian AI Hub, 2026k). ● Low-risk public-tool use: no sensitive, personal, security-classified or commercially confidential information; outputs independently verified. ● Prohibited or unapproved use: systems that fail defined security or legal thresholds, or are subject to a specific government direction.
Use vendor jurisdiction as a risk factor, assessing foreign legal duties, beneficial ownership, telemetry, updates, supply-chain dependencies, verified performance and governance risk matrices rather than automatically disqualifying vendors.
Vendor jurisdiction should inform risk assessments rather than act as an absolute disqualifier; risk evaluations should examine foreign legal obligations, ultimate beneficial ownership, telemetry practices, update mechanisms, supply-chain dependencies, verified technical performance, and structured governance risk matrices (Western Australian AI Hub, 2026j).
Keep the AI security classification model technology-neutral and subject to regular review.
This proposed security model should remain technology-neutral and subject to regular review.
Fold ASD-guided controls (executive accountability, rapid patching, least privilege, segmentation, logging, tested backups, defence in depth) into programmes under the Security of Critical Infrastructure Act 2018.
ASD guidance stresses executive accountability, rapid patching, least privilege, segmentation, logging, tested backups and defence in depth (Australian Signals Directorate, 2026). Fold those controls into programmes under the Security of Critical Infrastructure Act 2018 (Cth).
Require defence systems to maintain human command and lifecycle ethical governance.
Defence systems must maintain human command and lifecycle ethical governance
Create a single National AI Coordination Gateway connecting policy, governance, safety, regulatory, cyber and intelligence functions across existing government bodies.
A single National AI Coordination Gateway should connect policy, governance, safety, regulatory, cyber and intelligence functions across existing government bodies.
Have the Gateway maintain a responsibility map, shared risk taxonomy, governance readiness diagnostics, standardised procurement clauses and regulatory referral processes.
A single National AI Coordination Gateway should connect policy, governance, safety, regulatory, cyber and intelligence functions across existing government bodies. It should maintain a clear responsibility map, shared risk taxonomy, governance readiness diagnostic frameworks (Western Australian AI Hub, 2026k), standardised procurement clauses and regulatory referral processes.
Have the Gateway facilitate regulator, AISI and national-security information-sharing to monitor emerging capabilities, safety risks and supply-chain vulnerabilities.
The Gateway should facilitate structured information-sharing across statutory regulators, the Australian AI Safety Institute (AISI) and national security agencies to monitor emerging system capabilities, safety risks and supply-chain vulnerabilities (Department of Industry, Science and Resources, 2026).
Have the AISI provide technical evaluation and risk analysis while statutory regulators retain their enforcement powers.
The AISI should provide technical evaluation and risk analysis, while statutory regulators retain their respective enforcement powers.
Run joint, time-bound assessments of major compute proposals across planning, energy, water, environment, security and regional impact.
The same gateway should run joint, time-bound assessments of major compute proposals across planning, energy, water, environment, security and regional impact.
Before approval, mandate facility registration in national supply-side datasets and disclosure of power/water use, emissions, backup generation, grid integration, PUE and regional benefits quantified by standardised social licence scoring.
Infrastructure proponents and public sector compute procurement should mandate facility-level registration on national supply-side datasets (Western Australian AI Hub, 2026e), disclosing projected power and water usage, emissions profiles, backup generation plans, grid integration strategies, Power Usage Effectiveness (PUE), and regional benefit commitments quantified using standardised social licence scoring (Western Australian AI Hub, 2026b) before approval.
Allow fast pathways for major compute proposals only where complete and meeting published tests, never as a bypass of environmental assessment or Indigenous consultation.
Fast pathways only for complete proposals that meet published tests, never as a bypass of environmental assessment or Indigenous consultation.
Within six months, publish a responsibility map, design the Commons pilot, issue procurement/register templates and consult on public-sector safeguards, involving PM&C, Finance, DTA, DISR and the Attorney-General’s Department.
In the first six months, publish a responsibility map, design the Commons pilot, issue model procurement and register templates, and open consultation on public-sector decision safeguards (PM&C, Finance, Digital Transformation Agency, DISR, Attorney-General’s Department).
Within 18 months, launch pilots, legislate high-impact safeguards, establish Indigenous data conditions and cost the incentive, involving Treasury, NIAA, states and territories.
Within 18 months, launch pilots, legislate high-impact safeguards, lock in Indigenous data conditions and cost the diffusion incentive (Treasury, National Indigenous Australians Agency, states and territories).
By 36 months, evaluate and scale successful measures, finish cross-jurisdiction infrastructure pathways and publish annual uptake, concentration, domestic IP/work, incident and energy/water indicators.
By 36 months, evaluate, scale what works, finish cross-jurisdiction infrastructure pathways, and publish annual indicators on SME and regional uptake, procurement concentration, domestic intellectual property and skilled work, incidents, and energy and water commitments.