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 551

Design Institute of Australia

Submitter group: Industry bodies.

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 for recognising design in AI commercialisation, testing its contribution through pilots, procurement design requirements and economic measurement. Recommends computing access, skills, data-centre engagement and design-IP reform. Says AI-specific economic returns remain unquantified, blanket training-consent rules need assessment, and individual-work transparency must be workable and protect confidentiality. It makes no substantive recommendations on data-centre energy, water or cost allocation.

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 Australia can attract computing and AI training investment without capturing a commensurate share of the economic value.

    Australia can attract major investments in computing infrastructure and frontier AI training without capturing a commensurate share of the resulting economic value.

  2. The Design Institute of Australia recommends that design be explicitly recognised in the Australian AI innovation supply chain alongside research, engineering and development.

    The Design Institute of Australia (DIA) recommends that design be explicitly recognised within the Australian AI innovation supply chain, alongside research, engineering and development.

  3. States that Australian evidence shows design is economically significant but does not yet quantify returns from design-led AI commercialisation.

    Australian evidence establishes that design is economically significant, but does not yet quantify the incremental return from design-led AI commercialisation.

  4. Argues designers need a way to find out whether their work has been used to train AI.

    Designers need a way to find out whether their work has been used to train AI. DIA proposes that transparency about AI training data work at three levels:

  5. Argues that the Commonwealth, as the country's largest buyer, can set the standard for the wider economy through how it buys AI.

    The Commonwealth is the country’s largest buyer, and how it buys AI can set the standard for the wider economy.

  6. Says communities should be consulted early on data centre siting, using design methods such as co-design.

    Communities should be consulted early, while decisions about location and design can still change. Design methods such as co-design, clear and accessible communication, mapping who is affected, and testing options with prototypes help communities understand the choices and who will bear the impacts.

Recommendations

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

  1. Recognise design, including the design of how AI systems behave, as a core part of turning AI research and infrastructure into products and services, alongside research, engineering and development.

    Recognise design, including the design of how AI systems behave, as a core part of turning AI research and infrastructure into products and services, alongside research, engineering and development.

  2. Recognise designing AI systems, designing with AI and designing AI-enabled organisations as complementary practices; treat this as a proposed policy framing rather than an existing OSCA taxonomy.

    We propose that Government recognise three complementary dimensions of practice: designing AI systems (including interaction, delegation and oversight); designing with AI (including augmented ideation, prototyping and production); and designing AI-enabled organisations (including services, workflows, decision processes and operating models). This is a proposed policy framing, not an existing OSCA occupational taxonomy.

  3. Use the OSCA Minor Group 242 design occupations for reporting and eligibility guidance, noting Service Designer (242134) is proposed in the OSCA 2027 draft and not yet adopted.

    Use the OSCA Minor Group 242 design occupations for reporting and eligibility guidance, noting that Service Designer (242134) is proposed in the OSCA 2027 draft and not yet adopted.

  4. Maintain OSCA mapping through revisions without excluding genuine multidisciplinary capability outside the codes; allow qualifications, experience, multidisciplinary practice and appropriate accreditation as evidence without monopolising participation.

    Mapping to OSCA should be maintained through future revisions and should not exclude genuine multidisciplinary capabilities that fall outside these codes. The DIA supports using occupation codes to track training, staffing and professional development without making classification a monopoly on participation. Professional qualifications, demonstrated experience, multidisciplinary practice and appropriate accreditation may all evidence capability.

  5. Fund design-led AI commercialisation partnerships pairing designers with Australian researchers, developers, manufacturers and small businesses, first as pilots with published selection rules, independent assessment and measured results, expanding only where they deliver.

    Fund design-led AI commercialisation partnerships that pair skilled designers with Australian researchers, developers, manufacturers and small businesses. Run them first as pilots with published selection rules, independent assessment and measured results, and expand them only where they deliver.

  6. Open design partnerships to digital, industrial, strategic and service design; AI behaviour and interaction design; accessibility and safety; and service redesign, selecting demonstrated economic/community benefit rather than fixed discipline shares.

    These partnerships should be open to all relevant kinds of design work, including: digital, industrial, strategic and service design; designing how AI systems behave and interact with people; making AI products and services accessible and safe to use; and redesigning services to make use of AI. Projects should be chosen for the benefit they can show for Australia’s economy or community, not by setting aside a fixed share for particular design disciplines.

  7. When negotiating developer or infrastructure-provider contributions, publish clear selection rules and keep applications simple so sole traders and small studios can access support.

    Where Government negotiates these contributions, it should publish clear selection rules. Programs should be simple to apply for, so that support reaches sole traders and small studios and is not taken up mainly by large, established firms.

  8. If government reserves research and innovation compute, provide designers and creative businesses an open-criteria application pathway rather than an automatic entitlement.

    If Government sets aside computing power for research and innovation (known as “reserved compute”), designers and creative businesses should have a clear, fair way to apply for it as part of the wider national program. This should be a pathway with open criteria, not an automatic entitlement.

  9. Assess reserved-compute applications on significance, genuine compute need, deliverability, additionality and expected Australian benefit.

    Applications should be assessed on: how significant the project is; whether it genuinely needs the computing power; whether it can realistically be delivered; whether the support makes a real difference, rather than funding work that would happen anyway; and the benefit it is expected to bring to Australia.

  10. Offer affordable AI tools/models and technical support, plus computing credits, shared facilities and expert help, with appropriate procurement and security safeguards.

    Not every design business needs large-scale computing. Many studios would benefit more from affordable access to AI tools and models, plus technical support. Government should offer a mix of options, such as computing credits, shared facilities and expert help, with appropriate procurement and security safeguards.

  11. Include design research and design-led commercialisation as eligible purposes for public-interest research funding, subject to demonstrated public benefit and independent evaluation.

    Include design research and design-led commercialisation as eligible purposes for public- interest research funding, subject to demonstrated public benefit and independent evaluation.

  12. Open university, TAFE and RTO partnerships to design schools and cross-disciplinary AI programs.

    Open university, TAFE and RTO partnerships to design schools and cross-disciplinary AI programs.

  13. Measure results, not just activity: track progress from prototype to real-world use, adoption, user outcomes, Australian-owned IP and exports, using OSCA codes for workforce reporting.

    Measure results, not just activity: track progress from prototype to real-world use, adoption, user outcomes, Australian-owned IP and exports, using OSCA codes for workforce reporting.

  14. Record project baselines and track speed to market, sustained adoption, task success including disability access, and compliance rework.

    Programs should record a starting point for each project and then track results such as: how many prototypes progress to real-world use, and how quickly they reach market; how many people adopt the product or service and keep using it; whether users can complete tasks successfully, including people with disability; how much rework is needed to meet regulatory or compliance requirements; and the intellectual property held in Australia, and revenue and exports where these can be measured.

  15. Where practicable, independently evaluate results against similar projects without designers or projects selected by clear published criteria.

    Where practicable, results should be independently evaluated and compared with similar projects that did not involve designers, or with projects chosen using clear, published criteria.

  16. In program terms, specify input/output ownership, lawful third-party use and licensing, retention and deletion, confidentiality/security, and whether information may train or improve models.

    The terms of any program should make clear: who owns the work designers put in and the results they get out; how any third-party material can be lawfully used and licensed; how long information is kept, and when it is deleted; how confidentiality and security are protected; and whether designers’ information can be used to train or improve AI models.

  17. Protect unpublished design work particularly and make program terms work across cloud services and information sensitivity levels.

    Unpublished design work needs particular protection, because early disclosure can affect a designer’s ability to register and protect it. These terms should work across different cloud services and for information of differing sensitivity.

  18. Fund AI upskilling for design occupations that reaches sole traders and small studios as well as large employers, and evaluate it by the skills and work outcomes it delivers.

    Fund AI upskilling for design occupations that reaches sole traders and small studios as well as large employers, and evaluate it by the skills and work outcomes it delivers.

  19. Cover AI fluency, limitations, evaluation, IP/confidentiality, systems thinking, agent supervision, accessible interaction, safety and sector-specific tools in publicly supported workforce programs; reach sole traders, established practitioners and multidisciplinary teams.

    Publicly supported workforce programs should cover AI fluency, model limitations, evaluation, IP and confidentiality, systems thinking, agent supervision, accessible interaction, safety and sector- specific tools. Training should extend beyond incumbent graduates and large employers to sole traders, established practitioners and multidisciplinary teams. Funding should be evaluated by skill adoption and work outcomes, not enrolment counts alone.

  20. Commission a regular, government-funded national design economy study, comparable to the UK Design Council's reports and using 2026 Census data.

    Commission a regular, government-funded national design economy study, comparable to the UK Design Council’s Design Economy reports and using 2026 Census data, so that design’s economic contribution and the opportunity it represents are visible to government.

  21. Distinguish specialist-industry and embedded design employment in the recurring government study; publish methodology, uncertainty and time-series comparability.

    A recurring Australian study, commissioned by government, should therefore distinguish design employment in specialist industries from designers embedded across the economy, and publish its methodology, uncertainty and time-series comparability.

  22. Use appropriate 2026 Census data, comparability-tested recoded 2021 data and surveys of embedded designers/firms; publish uncertainties and avoid headcount alone as economic-impact evidence.

    A regular design economy study should use 2026 Census occupational data where appropriate, compare with recoded 2021 data subject to comparability tests, and supplement it with surveys of embedded designers and firms. Government should publish uncertainties and avoid treating headcount as a sufficient indicator of economic impact.

  23. Ask Jobs and Skills Australia to move its occupation reporting to OSCA so design employment is measured consistently.

    Ask Jobs and Skills Australia to move its occupation reporting to OSCA, so that design employment is measured consistently.

  24. For Commonwealth AI-enabled services, require service/human-centred design capability, real-user testing and monitoring, building on the Digital Service Standard and permitting small firms and equivalent-experience teams.

    When the Commonwealth buys AI-enabled services, require suppliers to show service and human-centred design capability, testing with real users, and ongoing monitoring, building on the Digital Service Standard and open to small firms and teams with equivalent experience.

  25. Match supplier real-user research and testing requirements to service risk.

    DIA recommends that when government buys AI-enabled services, suppliers be required to show: they have suitable service design and human-centred design capability; they will research and test the service with real users, to a level that matches its risk; and they will monitor how the service performs once it is in use.

  26. For services significantly affecting lives, record user effects, human-review escalation, expected performance and post-launch feedback, building on existing digital service rules.

    For services that have a significant effect on people’s lives, agencies should also record how the service affects users, how problems are escalated to a person for review, what performance is expected, and how feedback is collected after launch. These requirements should build on existing digital service rules, not create a separate system.

  27. Where needed, include interior, industrial, graphic/visual communication, UI/UX, service and strategic design skills in data-centre plans for fit-out, wayfinding, operator tools, ergonomics and staff wellbeing; offer graduate placements/cadetships and OSCA reporting.

    Include design skills in data-centre workforce plans where a project needs them, including interior, industrial, graphic and visual communication, UI/UX, and service and strategic designers, covering fit-out, wayfinding, operator tools, ergonomics and staff wellbeing. Offer graduate placements and cadetships, and report these roles using OSCA codes.

  28. Engage communities early with service/strategic designers and public commitments; involve strategic designers, architects, urban designers and landscape professionals in siting.

    Engage communities early, drawing on service and strategic designers, and keep a public record of the commitments made. In siting decisions, involve strategic designers, architects, urban designers and landscape professionals, and weigh design quality, heritage, visual amenity, clustering and the potential to reuse suitable industrial buildings.

  29. Respect relevant First Nations consultation and cultural heritage obligations in data-centre engagement, beyond general design methods.

    Engagement with First Nations peoples must meet the relevant consultation and cultural heritage obligations, not rely on general design methods alone.

  30. Weigh land-use efficiency, design quality, heritage, surroundings and clustering in siting; consider existing-building reuse where technically suitable, with energy/water/cooling requirements, resilience and cost determining suitability.

    Siting decisions should weigh efficient use of land, the quality of the building’s design, heritage, how the facility looks in its surroundings, and the combined effect of several facilities clustering in one area. Where it is technically suitable, existing industrial buildings should be considered for reuse as data centres. Many large data centres have high energy, water and cooling demands that existing buildings cannot meet, so these practical requirements, along with resilience and cost, should decide whether reuse is suitable.

  31. Record publicly the alternative sites and competing land uses considered in site assessments.

    Site assessments should openly record what alternatives and competing land uses were considered.

  32. Introduce three levels of training-data transparency: a public summary of main content types and sources, detailed records for regulators, and a process for rights holders to ask whether a work was used.

    Introduce three levels of training-data transparency: a public summary of the main types and sources of training content; detailed records available to regulators; and a clear process for designers and other rights holders to ask whether a particular work was used.

  33. Make fuller training-data records available for lawful audit when required and give rights holders a process to exercise rights over particular training works.

    2. Detailed records. Developers keep fuller records of where their training data came from, available to regulators or for lawful audit when required. 3. A way for creators to ask. Designers and other rights holders have a clear process to ask whether a particular work was used, and to assert their rights if it was.

  34. Ensure individual-work search or disclosure requirements are technically workable and protect privacy, commercial confidentiality and security.

    Any requirement to search or disclose individual works would need to be technically workable and include protections for privacy, commercial confidentiality and security.

  35. Confirm existing duties, registration, professional standards and insurance obligations continue for AI-assisted design within regulatory environments; build appropriate human oversight into AI systems acting on people’s behalf.

    Confirm that, where design work is undertaken within regulatory environments, existing legal duties, registration rules, professional standards and insurance obligations continue to apply to AI-assisted work, and build appropriate human oversight into AI systems that act on people’s behalf.

  36. Focus regulation on work risks and legal responsibility, distinguishing voluntary accreditation from statutory registration across jurisdictions.

    Regulation should focus on the risks of the work and on who is legally responsible for it, recognising that voluntary professional accreditation is not the same as statutory registration, which varies between states and territories.

  37. Incorporate agentic-system authority divisions, permission boundaries, risk-appropriate logging, override, escalation and recourse into design education and public procurement evaluation; make requirements technology-neutral and proportionate to failure consequences.

    For agentic systems, good design also includes a clear division of authority, permission boundaries, logging appropriate to risk, meaningful override, escalation and recourse. Requirements should be technology-neutral and proportionate to the consequences of failure. DIA proposes incorporating these capabilities into design education and public AI procurement evaluation.

  38. Develop fair and workable options for licensing designers' work for AI training and paying them, after assessing legal scope, cost, enforceability and the risk of training moving offshore.

    Develop fair and workable options for licensing designers’ work for AI training and paying them for it, after assessing legal scope, cost, enforceability and the risk of training moving offshore.

  39. Assess blanket consent-before-training rules further; distinguish copyrighted from other material, licence terms from legislation, research from commercial training, and major-model training in Australia from later adaptation/use.

    A blanket rule requiring creators’ consent before any training needs further assessment. Any reform should clearly distinguish between: material protected by copyright and material that is not; terms agreed in licences and new rules set by law; research and commercial training; and training of major AI models in Australia and later adaptation or use of those models.

  40. In training-IP reform, consider smaller Australian developers’ access and consistency with international approaches.

    Reforms should also consider the cost of compliance, whether rules can be enforced, access for smaller Australian developers, consistency with international approaches, and the risk that AI training simply moves offshore without improving protection for Australian creators.

  41. Bring the Designs Act 2003 and its overlap with copyright within the Attorney-General’s copyright and AI process.

    Include design expertise in the Attorney-General’s copyright and AI process, and bring the Designs Act 2003 and its overlap with copyright within its scope.

  42. Include industrial, digital and spatial design expertise, including CAD, 3D models and renders, in the Attorney-General’s copyright and AI process.

    DIA asks that the Attorney- General’s copyright and AI process include people with expertise in industrial, digital and spatial design, and in CAD files, 3D models and renders.

  43. Examine how CAD files, 3D models and digital or virtual designs are protected, including whether they can be registered and what remedies apply when they are misused.

    Examine how CAD files, 3D models and digital or virtual designs are protected, including whether they can be registered and what remedies apply when they are misused.

  44. Examine commercially manufactured designs, small studios’ practical enforcement and designers’ authorship recognition.

    The process should look at: whether digital or virtual designs can be registered and protected; how designs that have been manufactured and sold are treated; whether small studios can realistically enforce their rights; recognition of designers as authors of their work, and remedies when it is misused; and whether a short-term, low-cost protection for unregistered designs, as exists in the UK and EU, would fill a genuine gap.

  45. Examine the case for a short-term, low-cost protection for unregistered designs, as exists in the UK and EU.

    Examine the case for a short-term, low-cost protection for unregistered designs, as exists in the UK and EU.

  46. Consult on a standard digital tag for design registrations and files that tells AI developers whether a work can be used for training and on what terms.

    Consult on a standard digital tag for design registrations and files that tells AI developers whether a work can be used for training and on what terms.

  47. Give any design-training digital tag a common standard, clear legal effect, authority-setting rules, legacy-work coverage and persistence through copying and sharing.

    To work, such a tag would need a common standard and clear legal effect. It would also need rules on who can set it, a way to cover older work, and a way to stay attached as data is copied and shared.

  48. Explore collective design licensing with fair governance, clear coverage and pricing, opt-out and competition safeguards.

    Collective licensing, where an organisation licenses work and collects payment on behalf of many creators, may help some designers. It would need fair governance, clear coverage and pricing, the option for designers to opt out, and safeguards for competition.

  49. Recognise design as a distinct creative practice rather than a form of visual art.

    Explore collective licensing for design, with fair governance and the option to opt out, and recognise design as a distinct creative practice rather than a form of visual art.

  50. Adopt national minimum standards and clear federal lead agencies; retain state/territory planning, environmental and professional responsibilities, with compatible information sharing where practicable.

    DIA supports a single set of national minimum standards, so that businesses and designers do not face conflicting or duplicated requirements from state to state. Each area of federal responsibility should have a clear lead agency, including how occupations are classified, how government buys AI, and transparency about data and intellectual property. States and territories should keep responsibility for planning, environmental and professional regulation, and share information in compatible formats where practicable.

  51. Distribute infrastructure, training, commercialisation and evaluation roles across organisations rather than expecting one provider; use independent program assessment and accountability for results.

    Government should not expect one organisation to provide the infrastructure, training, commercialisation support and evaluation. Independent bodies should assess programs and hold them accountable for results.

  52. Publish pilot outcomes in aggregate with safeguards for commercial confidentiality and personal information, and include sunset or review clauses so funding expands only where additionality and economic or public value are shown.

    DIA recommends that pilot outcomes be published in aggregate, with safeguards for commercial confidentiality and personal information. Programs should include sunset or review clauses so funding is expanded only where additionality and economic or public value are demonstrated.

  53. Within 0–6 months, identify owners, establish cross-sector advice, scope design-economy measurement and set pilot metrics; publish terms of reference, methodology and program criteria.

    0–6 months Identify policy owners; establish Terms of reference, cross-sector advisory input; scope methodology, program criteria design-economy measurement; set pilot metrics

  54. Within 6–18 months, run competitive translation pilots, map workforce reporting and consult on design-IP gaps; publish baseline/interim results and a legal-options paper.

    6–18 months Run competitive translation pilots; Baseline and interim pilot create workforce reporting mapping; results; legal-options paper consult on design-IP gaps

  55. Within 18–36 months, independently evaluate pilots, incorporate proven capability requirements and establish repeat reporting; publish cost-effectiveness evaluation, an outcome dashboard and repeatable statistics.

    18–36 months Commission independent pilot Cost-effectiveness evaluation, evaluation; incorporate proven outcome dashboard, capability requirements; establish repeatable statistics repeat reporting

Read the original

Open in the interactive guide