## A Fair Go for the AI Age: Postcode-Level Sovereign Compute as Shared National Infrastructure Draft submission to the Senate Environment and Communications References Committee Inquiry: Artificial intelligence and data centres Submitter: Luke Nathan Hayes, private citizen, Minjerribah / North Stradbroke Island, Queensland Date: 18 August 2026 Status of proposal: Concept-stage and unfunded. Strange but True is my existing sole-trader entity. Ready SET Co. is a proposed cooperative model. The Dunwich site discussed below is a proposed pilot location. Nothing in this submission should be read as claiming government, council, Quandamooka, university, industry or community endorsement that has not actually been obtained. ### Submission purpose and relationship to Submission #2 This submission is intended to directly extend the case I made in Submission #2, Luke Nathan Hayes, to the federal inquiry into Local Government Funding and Fiscal Sustainability. That earlier submission argued that local government is operating with an “Incomplete Ledger”: councils are expected to maintain physical infrastructure, community wellbeing, environmental assets, social cohesion and resilience while much of the labour and value that creates those outcomes is poorly represented in conventional economic accounting. It proposed a Braided Economy and Community-Hour or C-Hour; community sovereign wealth mechanisms; sovereign digital twins; an L0-L2 data and governance architecture; gamified civic participation; and a Minjerribah pilot incorporating First Nations data sovereignty. fileciteturn0file7 This submission asks the next physical-infrastructure question: Where should the intelligence that operates those future civic systems actually live, who should own it, and who should be capable of understanding and changing it? The Senate's current inquiry provides an appropriate place to ask that question. Its terms of reference explicitly cover the effectiveness of Australia's regulatory frameworks for data-centre growth, existing and future dealings with global AI companies, impacts on communities, industries, water and energy, and “any other related matters”. Submissions currently close on 1 September 2026. [\[1\]](https://www.aph.gov.au/Parliamentary_Business/Committees/Senate/Environment_and_Communications/AIdatacentres48P) The Albanese Government has simultaneously announced that it intends to legislate Australian Standards for AI, building on its Data Centre Expectations, using Australian values as a benchmark and seeking to strengthen sovereignty and ensure community benefit. The Office of AI was established in the Department of the Prime Minister and Cabinet on 15 July 2026. [\[2\]](https://www.pm.gov.au/media/ai-australias-interests) My proposal is therefore straightforward: Australia should not build its AI future with all of its eggs in one basket. We should deliberately balance centralised data centres, sovereign national compute, decentralised postcode-level compute, neighbourhood edge compute and self-sovereign personal AI. At least one serious layer of advanced compute should be physically distributed through Australian communities rather than concentrated almost entirely in a small number of large data centres. This is not a rejection of data centres. It is an argument for architectural diversity, competition, resilience and democratic access. Nor is the alignment between this proposal and current Australian policy an accidental discovery. I have deliberately been developing these ideas across multiple scales: Redlands, Queensland, Australia, Quandamooka Country and other Native Nations, Oceania, international civic networks, the United Kingdom, United Nations frameworks, treaty relationships, sister-city relationships and ultimately a global interoperable system. The different pieces are intended to fit together rather than being isolated proposals. The Australian Government's own National AI Plan now describes compute and digital connectivity as critical infrastructure, recognises distributed processing as part of the digital backbone, calls for stronger domestic capability and global partnerships, and says the benefits of AI should extend across metropolitan, suburban, regional and remote communities. [\[3\]](https://www.industry.gov.au/publications/national-ai-plan/introduction) Its Data Centre Expectations likewise seek national-security and data-sovereignty benefits, local skills, jobs and innovation from major infrastructure providers while expressly sitting outside small-scale edge and on-site enterprise data centres. [\[4\]](https://www.industry.gov.au/publications/expectations-data-centres-and-ai-infrastructure-developers) There is already an important Australian technical precedent for the basic architecture I am proposing. In May 2026 CSIRO launched Vetra at its Queensland Centre for Advanced Technologies at Pullenvale. CSIRO describes Vetra as compact, modular high-performance computing placed close to robots and sensors, calls it “sovereign, trusted AI computing at the edge”, and says cloud-only approaches are not sufficient for some safety-critical and time-sensitive applications. Crucially, CSIRO describes Vetra as working with central supercomputers as an integrated edge-core-cloud system and says the model can be replicated elsewhere. [\[5\]](https://www.csiro.au/en/news/All/News/2026/May/Vetra-AI-infrastructure) That is close to the architectural principle I am advocating, extended from a national research facility into Australian civic life: personal → neighbourhood → postcode/local → bioregional → state → federal → Native Nation → Oceania → global → eventually off-world interoperability. My earlier decentralised-compute work already describes the same broad L0-L3 progression: sovereign individual and household computing, neighbourhood meshes and disaster kiosks, rack-scale local or bioregional intelligence, and large national/global compute. fileciteturn0file9 The Sovereign Space Builder specification independently develops the same architecture around offline-first operation, local custody, L0 Home, L1 Mesh, L2 Bioregion and L3 Planetary Commons, with access tiers, revocation, peer-to-peer synchronisation and Indigenous data sovereignty. fileciteturn0file14 The policy proposition is not that one layer replaces another. It is that Australia needs the whole stack. ### A national compute ecology with a postcode layer I intentionally use the term postcode-level compute. A postcode is not a claim about the correct statistical boundary, ecological boundary, council boundary or boundary of Country. My existing work already uses island, neighbourhood, council, watershed, disaster zone and bioregion when those concepts are appropriate. “Postcode” does something different. It is ordinary Australian language. People understand the idea that something exists in my postcode. It gives a human-scale picture of distributed infrastructure without requiring every citizen to learn a new bureaucratic expression. The architecture I ask the Committee to consider is approximately this: | Sovereignty domain | Proposed compute role | Governing principle | | --- | --- | --- | | Personal / household - L0 | Personal AI, live digital twins, living memoirs, health and ageing support, private RAG, household sensors, small local models | My data, my keys, my model preferences, my permissions | | Neighbourhood - L1 | Mesh networking, disaster kiosks, shared idle compute, local communications, schools, maker projects | Explicit sharing, local resilience, degraded-mode operation | | Postcode / local - L2 | One serious rack or suitable cluster serving local development, digital twins, simulations, inference, fine-tuning, citizen science and education | Shared national infrastructure physically accessible to communities | | Bioregion / island / watershed / LGA | Cross-postcode environmental and infrastructure models, emergency planning, water, transport, biodiversity and civic simulations | Govern according to the geography and problem being modelled | | State and territory | Large training workloads, emergency coordination, grid and infrastructure modelling, research and service integration | State capability interoperable with local and Commonwealth layers | | Commonwealth | National sovereign models, frontier research compute, national-security workloads, model testing and cross-jurisdictional coordination | Australian strategic independence without technological isolation | | First Nations / Native Nations | Nation-governed models, Country digital twins, language, culture, ICIP and community datasets | A sovereignty plane crossing other geographic levels, not a subordinate rung | | Oceania | Federated climate, health, disaster, culture, language, ocean and scientific models | Regional interoperability without compulsory centralisation in Australia | | Global and off-world | Scientific models, planetary digital twins, international research, frontier models and eventually orbital compute | Open protocols, provenance, consent and negotiated interoperability | This is a federated sovereignty model, not a pyramid in which everything inevitably flows upwards. Sometimes the correct operation will be entirely on a person's device. Sometimes several neighbours will collaborate. Sometimes a postcode rack will do it. Sometimes Redland City, Queensland, the Commonwealth, a university supercomputer or a hyperscale commercial provider will be the sensible place for a workload. Sometimes an Australian researcher should call a frontier model overseas because it is simply the best tool available. The architecture should make those choices possible. The current danger is not merely that foreign technology companies are foreign. The danger is dependency without credible alternatives. A country can enthusiastically participate in the global AI economy while still asking whether its people can retain critical data locally, change suppliers, inspect important systems, continue essential functions during network disruption, develop its own models, and train enough Australians to understand the stack rather than merely consume it. Australia is already entering major collaborations with international providers. Microsoft has an Australian Government MOU covering cloud infrastructure, AI capability, safety and workforce development, while the Commonwealth has welcomed a proposed $7 billion OpenAI-NEXTDC infrastructure partnership in Western Sydney. [\[6\]](https://www.industry.gov.au/news/australian-government-has-signed-memorandum-understanding-mou-tech-giant-microsoft) Those investments can contribute enormous capability. They should be one part of the system, not the system. CSIRO's Vetra is important precisely because it demonstrates that this need not be framed as “cloud versus edge”. CSIRO is using edge, core and cloud together. [\[5\]](https://www.csiro.au/en/news/All/News/2026/May/Vetra-AI-infrastructure) That is the same balance I am proposing at national civic scale. I therefore propose that Australia explore treating local AI capability as shared national infrastructure, in much the same conceptual family as libraries, schools, roads, telecommunications, emergency infrastructure and research facilities. A postcode rack would not merely rent compute cycles to residents. Its more valuable output could be capability transfer. A Year 10 student should be able to discover that there is serious AI hardware in their community. A teacher should be able to build a project around it. A local developer should be able to experiment on it. A TAFE student should be able to learn its physical and software stack. A university researcher should be able to federate a project through it. A Traditional Owner organisation should be able to host or connect a sovereign model under its own permissions. A council planner should be able to simulate a proposal rather than just read a consultant's PDF. A citizen scientist should be able to plug a sensor network into a local digital twin. A community should be able to understand that “AI” is not magic occurring somewhere in Virginia, California or a giant warehouse in Western Sydney. It is computers, chips, models, data, cooling, power, networks, software, mathematics, human choices and governance. Australia's National AI Plan already identifies AI capability-building through schools, TAFEs, universities, community organisations and regional communities as part of spreading AI's benefits. [\[7\]](https://www.industry.gov.au/publications/national-ai-plan/spread-benefits) Postcode-level infrastructure could turn that ambition into something people can see, touch, test, break, fix and learn. That matters for national security as much as it matters for education. A society in which millions of people understand some part of its digital infrastructure is harder to fool, harder to monopolise, easier to mobilise during crises, and has a deeper pool of people capable of noticing when something has gone wrong. ### Personal sovereignty, Australian open models and values people can actually choose AI sovereignty begins before the postcode rack. It begins with the person. As AI systems become persistent companions, memory systems, agents and digital twins, a one-size-fits-all terms-and-conditions regime becomes increasingly absurd. The system may know a person's relationships, health history, fears, aspirations, politics, sexuality, finances, voice, face, movement, correspondence, memories and values. It may eventually mediate substantial parts of that person's relationship with the digital and physical world. We should therefore design for a future in which people can express their values and permissions before the software industry silently chooses them on their behalf. This is one of the original purposes of Aura of Intelligence. The public Aura material describes a user-owned architecture in which people build their own model, own their information and choose what they share rather than surrendering data as the price of participation. [\[8\]](https://auraofintelligence.github.io/) The Strange but True community-ledger prototype separates a relatively public local profile from a deeper private aura.md containing AI preferences, devices, wellbeing, world-building and other personal information. [\[9\]](https://auraofintelligence.github.io/strange-but-true/community-ledger.html) The Aura Genesis material develops the richer version: values, preferences, life story, family memories, AI dialogue, a digital memoir, a digital-twin draft and an explicit human-consent gate, with dementia and ageing among the proposed research pathways. [\[10\]](https://p4a.xyz/pages/aura-genesis.html) This is not an assertion that a completed clinical Aura system exists today. My uploaded clinical research plan sets out a proposed pathway for dementia-care research and validation rather than evidence of an existing approved product. fileciteturn0file2 But the governance concept can be useful long before the full clinical system exists. I propose an Australian choose-your-own permissions layer for AI. Instead of repeatedly clicking “I agree” beneath contracts almost nobody meaningfully negotiates, a person could maintain a machine-readable and plain-English sovereignty profile expressing matters such as: What may leave my device? Who may train on it? May my data be used for inference but not training? May a model learn from me temporarily? May derived representations be retained? May my voice be cloned? May my family see this? May researchers use this under de-identification? May it leave Australia? May it enter a commercial model? May it be shared during an emergency? What happens when I revoke a permission? That direction is not foreign to Australian privacy principles. OAIC guidance already says consent should be informed, voluntary, current and specific, should not be unnecessarily broad, should be communicated in plain English, and should be capable of being withdrawn through an easy and accessible process. Once consent is withdrawn, the previous consent cannot simply be relied on for future use or disclosure. [\[11\]](https://www.oaic.gov.au/privacy/your-privacy-rights/your-personal-information/consent-to-the-handling-of-personal-information) The technical and policy opportunity is to make those principles operational at AI speed. Revocation also needs a more sophisticated conception than “delete my entire digital life”. Where legally possible, I want a person to be capable of revoking another party's permission to use data while maintaining the person's own sovereign copy and history. Existing statutory retention, deletion and records obligations would still apply to organisations where relevant; the principle is that withdrawal from a service should not require a person to erase themselves. The values question sits above this permissions system. The Prime Minister has said Australia's forthcoming AI Standards will use “our values as the benchmark”. [\[2\]](https://www.pm.gov.au/media/ai-australias-interests) That raises a beautiful democratic question: What are our values? Rather than assuming the answer is permanently known by government, Silicon Valley, a standards committee or me, Australians could be periodically invited to describe their own values. I would like Australians to be able to nominate the core values and virtues they believe we inherited from the past, require in the present, and want to cultivate for the future-and, importantly, explain how they arrived at them. The aggregated results could inform a transparent democratic social model. Individual results could inform a person's own AI. Neither needs to overrule Australian law, and neither needs to force a majority's morality onto a minority. The point is almost the reverse: make the values layer visible instead of burying it inside model training and corporate terms of service. A person might then ask their digital twin to operate with different combinations of Australian community expectations, professional standards, cultural protocols, family rules and personally declared preferences. This is one reason I use Joyful Responsible Abundance as an organising question rather than merely a slogan: What does Joyful Responsible Abundance mean to self, family, community, city, island, watershed or bioregion, state, nation, Oceania and planet? Different people and cultures will answer differently. That difference is useful data for learning how to coexist. #### Australian-designed, trained and open models I also propose that Australia commit to producing Australian-designed, Australian-trained and Australian-deployed models whose complete usable model stack is open-source wherever law, privacy, cultural authority and national security permit it. That should be treated as a sovereign capability project. “Open” should aspire beyond an API endpoint. Depending on the model and lawful constraints, it should include model architecture, weights, inference software, training and fine-tuning tools, evaluation tools, build documentation and reproducible deployment processes. Fully open-source models do not require private health records, copyrighted material, secret information or culturally restricted data to be thrown onto the public internet. The model can be open while particular datasets remain sovereign. In the meantime, Australia should use what works. That means Australian and international open-source models, open-weight models and closed frontier systems where appropriate. We should learn from all of them. We should cooperate with global leaders. We should buy useful services. We should host foreign models locally where appropriate. We should participate in international safety work. We should simply avoid turning today's market leader into tomorrow's unavoidable constitutional infrastructure. Interoperability is the strategy; dependency is not. ### Quandamooka data sovereignty and a Digital Twin of Country First Nations sovereignty should not be represented as one subordinate rung beneath local government. It intersects every layer. The Commonwealth's own Framework for Governance of Indigenous Data places Aboriginal and Torres Strait Islander peoples at the centre of data governance, seeks greater agency over government-held Indigenous data across the data lifecycle, and requires agencies to develop implementation plans around partnership, data capability, knowledge of data assets and inclusive data systems. [\[12\]](https://www.niaa.gov.au/news-and-media/framework-governance-indigenous-data) This is already moving beyond abstract principle. Closing the Gap Priority Reform Four is supporting community data projects intended to strengthen local Indigenous data ecosystems, and NIAA's 2026 AI transparency statement expressly says its AI use should strengthen community control and protect cultural rights through Indigenous Data Sovereignty. [\[13\]](https://www.niaa.gov.au/our-work/evaluations-and-evidence/indigenous-advancement-strategy-ias-evaluation-work-plan/priority-reform-four-community-data) AIATSIS's Yumi Sabe platform similarly supports nation-specific data and Indigenous nation-building, is based around Indigenous Data Governance, Indigenous Data Sovereignty and Indigenous Cultural and Intellectual Property, and says material is to be shared where appropriate permissions have been granted by relevant copyright and ICIP holders. [\[14\]](https://yumi-sabe.aiatsis.gov.au/about) NIAA is now separately researching how ICIP clauses could enact Indigenous data sovereignty in Commonwealth procurement and grants. [\[15\]](https://www.niaa.gov.au/our-work/evaluations-and-evidence/indigenous-advancement-strategy-ias-evaluation-work-plan/indigenous-cultural-and-intellectual) I propose taking that logic much further technically. For a Minjerribah pilot, I would like to explore a self-sovereign Quandamooka AI and Digital Twin of Country, designed and governed with the relevant Quandamooka people and institutions rather than constructed externally and then presented for approval. The model could contain several permission zones rather than pretending all “data” is equivalent. Some knowledge may be public. Some may be available to community members. Some may require a cultural role, relationship, age, gender, family or other permission determined by custodians. Some may be available for a particular research question but not for general model training. Some may permit an aggregate answer while withholding the underlying data. Some may permit learning within Quandamooka-controlled infrastructure while prohibiting model weights or gradients from leaving. Some may simply remain closed. Special data warrants special permissions. The same choose-your-own-permissions concept operating at the individual Aura layer can therefore scale into collective sovereignty. A nation, organisation, family or custodian group can have permissions just as a person can. Federated learning is useful here, but should not be treated as a magic privacy word. NIST's privacy research notes that federated learning can still expose information through model updates or trained models; techniques such as secure aggregation and differential privacy can provide additional protections. [\[16\]](https://www.nist.gov/blogs/cybersecurity-insights/privacy-attacks-federated-learning) Model-extraction research has also demonstrated that query access to machine-learning systems can sometimes be used to reproduce important aspects of a model's functionality. [\[17\]](https://www.usenix.org/conference/usenixsecurity16/technical-sessions/presentation/tramer) A Quandamooka sovereignty layer could therefore make federated access permissioned and metered: authenticated users, purpose-limited credentials, rate limits, query budgets, logging, provenance, secure aggregation, privacy-preserving training techniques, anomaly detection and enforceable conditions on derived models. The design objective would be to make unauthorised extraction or distillation substantially harder, detectable where possible, and clearly outside the terms under which access was granted. My Sovereign Space Builder work already contains many of the supporting primitives: decentralised identifiers, verifiable credentials, access tiers, revocation, custodial licences, local-first storage and a specific cultural and Indigenous data-sovereignty layer. fileciteturn0file14 The Web3 Sensorium work similarly proposes decentralised identity, cryptographic provenance, peer-to-peer synchronisation and open scientific models. fileciteturn0file6 I use the word Treaty deliberately as an aspiration at the data-sovereignty layer. The exact legal form would of course have to be negotiated by the relevant parties. My policy proposition is that AI systems should make it technically possible for a First Nation to say: These are our models. These are our data. These are the permissions under which you may interact with them. These are the permissions under which your model may learn from ours. These permissions travel with the interaction. They can change. They can be revoked. Access does not silently convert custody into ownership. That would be a much richer form of sovereignty than merely placing an Indigenous-data label on a Commonwealth database. It would also offer Australia something worth sharing globally with other Native Nations: interoperability between sovereign intelligences without requiring every culture to pour its knowledge into one giant model. ### A Dunwich postcode pilot through Ready SET Co. I propose Dunwich on Minjerribah / North Stradbroke Island, within Redland City, Queensland, as one possible pilot. The proposed site is 9 Ballow Road, Dunwich. At present this is a hypothetical proposal. It is not a funded data-centre project. I am not claiming a rack has been purchased, that development approval has been obtained, that Ready SET Co. is registered, or that Quandamooka people, Redland City Council, Queensland or the Commonwealth have endorsed it. The current operating entity I am describing is my sole-trader activity, Strange but True. The proposed cooperative is Ready SET Co., with SET meaning Sustainable Employment and Training. Ready SET Co. is intended to make the capability-building function as important as the hardware itself. The pilot proposition would be to explore installing one first-class contemporary AI rack, or an equivalent modular cluster, with NVIDIA GB200/GB300-class and AMD rack-scale systems serving as examples of the level of capability rather than a locked procurement specification. Modern rack-scale AI systems are serious infrastructure. NVIDIA's GB200 NVL72 combines 72 Blackwell GPUs and 36 Grace CPUs in a liquid-cooled rack-scale system; GB300 NVL72 reference designs similarly integrate 72 Blackwell Ultra GPUs, and full-rack engineering involves substantial power, cooling and networking requirements. [\[18\]](https://www.nvidia.com/en-in/data-center/gb200-nvl72/) AMD is developing competing rack-scale systems through its Instinct and Helios architecture, giving Australia every reason to preserve vendor choice rather than hard-code one supplier into policy. [\[19\]](https://www.amd.com/en/products/accelerators/instinct/mi400/mi455x.html) Accordingly, “a rack in the community” should not be interpreted as parking a 100-plus-kilowatt liquid-cooled system beside the photocopier. The experiment I am proposing is more interesting: engineer the rack properly while engineering human access to it just as deliberately. That could mean secure machine space with appropriate power, cooling, fire protection, cyber controls and physical access, directly connected to a visible community AI workshop, classroom and development laboratory. Students do not need to put their hands inside live high-voltage infrastructure to “touch” the supercomputer any more than an engineering student needs to touch a live turbine blade. They should nevertheless be able to see the infrastructure, learn its components, deploy workloads, monitor power and cooling, train models, inspect telemetry and understand how the complete system behaves. A Dunwich pilot could test whether that infrastructure creates practical value across several seemingly different activities that actually share the same compute, data and simulation foundations: local software development; secondary-school and university learning; citizen science; tourism; film and cinema; game design; virtual production; XR; language and cultural projects where permission exists; digital twins of place and time; emergency and disaster management; environmental modelling; infrastructure planning; transport and logistics; Brisbane 2032 scenario planning; community volunteering; and research into personal digital twins for ageing and dementia. These are proposals for experimentation, not claims that each use case is already operational. The common asset is the important part. Instead of each school, filmmaker, council team, researcher, startup, emergency group and citizen-science project separately buying tiny fragments of cloud capability, a sufficiently powerful shared machine could become a civic workbench. The idea reaches further into my local-government funding submission. The C-Hour is proposed there as a way of recording and recognising verified community contribution. fileciteturn0file7 A local AI system could help residents discover civic work, match skills to problems, model proposed interventions, document completed work and update a community digital twin. A young person might earn recognition for mapping drains before a storm, digitising an oral history where permission is given, maintaining a sensor, restoring habitat, teaching an elder to use a digital service, building a game asset for a local virtual world, validating public data or testing an emergency simulation. Some rewards could be monetary. Some could be credentials, access to equipment, education, events, experiences or community recognition. My broader health-cooperative work also explores future health and wellbeing access as a potential cooperative benefit, including heat and pressure/oxygen interventions. Those proposals require their own clinical, safety and regulatory research and should not be mistaken here for proven entitlements or an operating health service. The uploaded Aura clinical pathway expressly treats health and dementia work as a staged research and governance problem. fileciteturn0file2 Brisbane 2032 offers an obvious mobilisation horizon. The Games' own current vision emphasises community participation and long-term benefits beyond the event itself. [\[20\]](https://newsroom.olympics.com/record/1679) I would like to explore whether the eventual Olympic and Paralympic volunteer mobilisation could leave behind something more permanent: a civic space-stewardship and resilience network whose skills, credentials and relationships remain useful after the closing ceremony. That is the sense in which a postcode rack is more than a box of GPUs. It becomes a local capability engine. And that is why the concept connects directly to fiscal sustainability. If Commonwealth policy assumes every council and community will independently buy AI expertise from cloud providers and consultants, AI risks becoming another recurring expense and capability dependency. If part of the national infrastructure is shared, open, locally accessible and connected to national capability, local government has a chance to become a competent participant in the AI economy rather than merely a customer of it. ### From Quandamooka Country to Oceania, the planet and orbital compute Australia should aim for sovereignty without isolation. At the personal level, sovereignty means I can own my digital twin and determine permissions. At the postcode level, it means my community has enough local compute and expertise to perform meaningful work. At the First Nations level, it means data and models relating to Country, culture and community can remain under genuine community authority. At the state and Commonwealth levels, it means Australia can operate important models and infrastructure without depending entirely on decisions made offshore. At the Oceania level, it means neighbouring countries and peoples should be capable of interoperating without having to centralise their intelligence in Sydney, Canberra-or California. Globally, it means the architecture is open enough to communicate with everyone. The National AI Plan itself combines domestic capability with international partnerships and specifically treats regional, First Nations and international engagement as parts of Australia's AI future. [\[21\]](https://www.industry.gov.au/publications/national-ai-plan/keep-australians-safe) I would like Australia to help develop a federation where models can visit data more often than data has to visit models. A Fijian, Papua New Guinean, Māori, Samoan, Quandamooka, Australian Commonwealth or other sovereign node could hold its own material, expose only authorised functions, exchange credentials and provenance, participate in federated learning where agreed, and withdraw from particular relationships without dismantling its own intelligence. That approach could extend through the treaties, research relationships, cultural relationships, sister cities, universities, local governments, Native Nations and international institutions Australia already interacts with. It does not require a single global government or single global model. It requires standards that allow difference to interoperate. This is also why I continue to explore Web3 architecture after much of the speculative hype associated with the term. The useful part is not buying cartoon tokens. The useful questions are: Who signed this? Who owns it? Where did this claim come from? What version was used? What permissions attach to it? Who changed it? Can the permission be revoked? Can two disconnected copies reconcile later? Can I carry a credential between systems without asking one corporation to vouch for my existence? My Global Sensorium proposal applies those mechanisms to a shared scientific digital twin: open source, decentralised participation, provenance, persistent modelling and a deliberate separation between established knowledge, contested propositions and speculative hypotheses. fileciteturn0file6 That epistemic layer matters. Science is not finished. AI will greatly increase our ability to search scientific possibility space, combine disciplines, formulate models and run simulations. A future scientific commons should therefore be capable of saying not merely what a model says, but what evidence it was built from, what assumptions it made, how uncertain it is, what competing models exist, who changed it and what observations would falsify it. The same infrastructure can be used for ordinary and extraordinary resilience. Flood. Bushfire. Cyclone. Pandemic. Cyberattack. Grid failure. Communications failure. Supply-chain disruption. Severe space weather. And low-probability celestial or geophysical scenarios whose probabilities may be uncertain or revised as science changes. I am not suggesting all hazards are equally probable. I am suggesting that simulation is cheap compared with discovering after a disaster that nobody modelled a particular combination of failures. Many very different catastrophes overlap in solution space: local power, local food and water intelligence, communications redundancy, backups, offline maps, medical capability, trusted community networks and the ability to operate when a distant cloud is unavailable. This is precisely where decentralised compute earns its keep. The technology horizon is also moving beyond terrestrial data centres faster than public policy documents written only a few years ago anticipated. SpaceX now publicly describes Starmind, a planned space-based AI-compute system, and has stated ambitions for thousands of AI satellites beginning as soon as late 2027 and eventually gigawatt-scale AI compute in orbit. Those are company plans rather than guaranteed outcomes, but they make the broader point clear: orbital compute is no longer purely a science-fiction architecture. [\[22\]](https://www.spacex.com/spacexai/starmind) Starlink's next-generation V3 architecture is itself designed around very high-capacity communications and optical inter-satellite links. [\[23\]](https://starlink.com/updates/starlink-version-3-satellites) Australian sovereign architecture should therefore be adaptive rather than frozen around the market topology of 2026. Personal devices will become more capable. Racks will become more capable. Data centres will become more capable. Networks will become more capable. Robotics will increase the demand for low-latency local inference. Models will change. New chip manufacturers will appear. Orbital compute may become commercially material. Scientific understanding will change. The sensible response is not to predict one winner perfectly. It is to create a system capable of absorbing new winners without surrendering sovereignty every time the technology stack changes. #### The forty-thousand-town peace gambit The global thought experiment makes the scale easier to see. The world-cities dataset supplied with this submission contains 44,691 rows. Using its recorded population field, my working count finds approximately 41,549 places with populations of at least 9,000; some records have missing population data. I therefore regard “roughly forty-odd thousand towns and cities” as sufficient accuracy for this thought experiment rather than pretending the number is a timeless census fact. Suppose the world eventually equipped roughly 44,000 towns and cities with a first-class civic AI rack and associated installation, education, networking and community infrastructure. At a deliberately rough A$6 million per location, that is: 44,000 × A$6 million = A$264 billion. Using the 41,549 places meeting the population threshold in the supplied dataset gives about A$249.3 billion. Those figures are not a procurement quote. They are a scale thought experiment. For context, the Australian Parliamentary Library records public estimates of approximately A$268-368 billion for Australia's nuclear-powered submarine program under AUKUS over its life, with the Parliamentary Budget Office producing an out-turned estimate of about A$367.6 billion including contingency. [\[24\]](https://www.aph.gov.au/About_Parliament/Parliamentary_departments/Parliamentary_Library/Research/Research_Papers/2024-25/The_deterrence_advantage_of_nuclear_powered_submarines) I know these are not like-for-like programs. That is precisely why I make the comparison. I am a private citizen offering a political, strategic and moral judgement, not pretending to have conducted a controlled experiment in alternative histories: I would like humanity to be willing to make a A$250-264 billion gambit on distributed intelligence, education, citizen science, cultural understanding, disaster resilience, sovereign local capability and peace. I believe that building the capacity for tens of thousands of communities to understand themselves and one another could create more peace than nuclear submarines and defensive deterrence ever could. Australia alone does not need to buy 44,000 racks. We could start with one. Then ten. Then a representative Australian network. Then invite New Zealand and Pacific partners to experiment on their own terms. Then interoperate with willing Native Nations, sister cities, universities and communities around the world. Each place could ask its people some version of: What values from our past are worth carrying forward? What values do we need now? What virtues do we want to cultivate in the future? How did we form those beliefs? What does Joyful Responsible Abundance mean here? The point is not to force everyone to agree. The point is to give civilisation better tools for understanding disagreement, modelling possible futures and finding overlapping solution space before conflict becomes the only coordination mechanism left. That is a legitimate thing for a citizen to ask their government to consider. ### Recommendations to the Committee I ask the Committee to consider the following as an integrated package rather than a collection of unrelated technology projects. Recommendation - Adopt a balanced national compute doctrine. Australian AI policy should explicitly plan for a mixture of hyperscale data centres, Commonwealth and research supercomputing, state capability, postcode-level community compute, neighbourhood edge systems and self-sovereign personal AI. CSIRO's Vetra provides an Australian demonstration that edge and core/cloud infrastructure can be complementary parts of sovereign capability. [\[5\]](https://www.csiro.au/en/news/All/News/2026/May/Vetra-AI-infrastructure) Recommendation - Establish a Postcode Sovereign AI Pilot Program. The Commonwealth could fund a small number of deliberately different demonstration communities: metropolitan, suburban, regional, remote, island, disaster-prone and First Nations-governed contexts. The question should be what becomes possible when serious compute, knowledge and community access are brought together-not whether every site should be identical. Recommendation - Consider Dunwich on Minjerribah as one candidate pilot. I offer 9 Ballow Road, Dunwich, as the conceptual location for a community-scale experiment connected with Ready SET Co. and the broader work described in Submission #2. It remains an unfunded proposal requiring the appropriate technical, community, Traditional Owner, planning, energy, safety, legal and financial processes before anything resembling deployment. Recommendation - Treat local AI capability as shared national infrastructure. The Commonwealth should investigate funding models that do not leave councils individually responsible for purchasing the entire AI stack from recurrent local revenue. This should be examined alongside the fiscal-sustainability questions already being considered through the Local Government Funding inquiry and my Submission #2. Recommendation - Create an Australian Open Model Commons. Australia should set a long-term objective of Australian-designed, Australian-trained, Australian-deployed and genuinely open-source foundation models, supported by open evaluation and deployment tooling and capable of running across personal, postcode, state and national infrastructure. Restricted data can remain restricted. Open models and sovereign data are compatible concepts. Recommendation - Remain interoperable with the best global systems while Australian capability grows. Open-source, open-weight and closed frontier models should all remain available where appropriate. Existing international partnerships can bring useful investment and knowledge, but procurement and standards should preserve switching, portability and local alternatives rather than creating permanent lock-in. Australia's current Microsoft collaboration and the OpenAI-NEXTDC proposal show how substantial the international investment layer is becoming. [\[6\]](https://www.industry.gov.au/news/australian-government-has-signed-memorandum-understanding-mou-tech-giant-microsoft) Recommendation - Develop a personal AI sovereignty and permissions standard. Australians should be able to maintain portable, intelligible and machine-readable preferences governing AI training, inference, retention, sharing, commercial use, overseas transfer, emergency access and revocation. The standard should build on the existing Australian principle that consent is informed, voluntary, current, specific and withdrawable. [\[11\]](https://www.oaic.gov.au/privacy/your-privacy-rights/your-personal-information/consent-to-the-handling-of-personal-information) Aura of Intelligence is my proposed implementation and experimentation environment for this concept, ranging from simple personal applications through to richer live digital twins and dementia/ageing research pathways. [\[25\]](https://auraofintelligence.github.io/) Recommendation - Periodically ask Australians what Australian values actually mean. If Australian values are to benchmark national AI standards, government and civil society should consider recurring surveys and deliberative exercises in which people identify important values and virtues of the past, present and future and explain their reasoning. [\[2\]](https://www.pm.gov.au/media/ai-australias-interests) Aggregate social expectations should inform AI governance without erasing individual values or minority cultures. Recommendation - Establish a First Nations AI and data-sovereignty pathway. Commonwealth AI policy should support First Nations that wish to build their own sovereign models and Digital Twins of Country, with governance authority spanning data, models, training, inference and derived outputs. This would build on the Commonwealth's existing Indigenous Data Sovereignty framework rather than starting from zero. [\[26\]](https://www.niaa.gov.au/artificial-intelligence-ai-transparency-statement) Recommendation - Support permissioned and metered federated learning for culturally or otherwise sensitive data. Pilot architectures should investigate credentials, purpose-bound access, query budgets, secure aggregation, privacy-preserving learning, logging, revocation, provenance and defences against model-extraction and distillation attacks. Federated learning should be treated as one tool in a broader sovereignty system, not as a synonym for privacy. [\[27\]](https://www.nist.gov/blogs/cybersecurity-insights/privacy-attacks-federated-learning) Recommendation - Make knowledge transfer an obligation attached to major AI infrastructure. Australia should ask more of hyperscale investment than electricity consumption and construction jobs. Community capability, startup access, researcher access, local engineering, apprenticeships, curriculum, model knowledge and opportunities for Australian organisations to build on the infrastructure should form part of its social licence. This direction is already compatible with the Commonwealth's Data Centre Expectations. [\[4\]](https://www.industry.gov.au/publications/expectations-data-centres-and-ai-infrastructure-developers) Recommendation - Build offline and degraded-mode AI into national resilience planning. Important local functions should be capable of continuing when cloud links, telecommunications or other central services are disrupted. Disaster kiosks, cached models, local maps, community resource data, sensor networks and postcode compute should be tested through real exercises rather than merely described in resilience documents. My existing L0-L3 specifications already treat offline-first operation as a design principle. Recommendation - Use Brisbane 2032 as a capability deadline rather than merely an event deadline. Queensland and the Commonwealth could use the remaining preparation period to train young people, developers, volunteers and communities in AI, simulation, digital twins and civic technologies, with the aim of leaving behind permanent local capability after the Games. Brisbane 2032's present vision already emphasises enduring community benefit. [\[20\]](https://newsroom.olympics.com/record/1679) Recommendation - Develop an Oceania sovereign-compute federation. Australia should invite Pacific partners to co-design interoperable systems in which participating countries and communities retain custody of sensitive data and choose what models, queries, credentials and derived information cross borders. Australia should offer collaboration, not simply replace foreign digital extraction with Australian digital extraction. Recommendation - Build an open provenance layer for science and civic simulation. Digital twins should distinguish observation, established scientific model, inference, disputed interpretation, speculation and imaginative scenario. Sources, assumptions, versions and uncertainty should travel with claims. The Web3 Sensorium concept I have supplied is an exploratory architecture for this kind of open scientific knowledge space. fileciteturn0file6 Recommendation - Keep the architecture adaptive to terrestrial and orbital technological change. Government should periodically reconsider where compute belongs as chips, robotics, networks, energy systems and space infrastructure evolve. SpaceX's current Starmind plans are one contemporary signal that assumptions about the physical location of large-scale compute may change substantially during the lifetime of infrastructure being approved today. [\[22\]](https://www.spacex.com/spacexai/starmind) Recommendation - Run the global-town thought experiment seriously. Australia should explore with international partners what a global network of locally sovereign civic AI nodes might contribute to scientific cooperation, education, resilience and peace. My rough A$249-264 billion thought experiment is deliberately provocative: it places the potential cost of a global positive-sum intelligence commons in the same broad order of magnitude as one major Australian strategic-defence program. The comparison is not an accounting equivalence; it is an invitation to broaden what we are willing to imagine as national and planetary security infrastructure. [\[24\]](https://www.aph.gov.au/About_Parliament/Parliamentary_departments/Parliamentary_Library/Research/Research_Papers/2024-25/The_deterrence_advantage_of_nuclear_powered_submarines) The choice before Australia is therefore not centralised AI or decentralised AI. It is whether we intentionally design the relationship between them. We can have world-class data centres and postcode racks. We can cooperate with American, European, Asian and global frontier labs and build Australian open models. We can have Commonwealth standards and personal choice. We can share knowledge and maintain data sovereignty. We can build powerful national models and sovereign Quandamooka, Māori, Pacific and other Native Nation models that decide for themselves what interoperability means. We can use enormous cloud systems and retain intelligence locally when a cyclone, cable failure, cyberattack, geopolitical disruption or something stranger knocks the cloud out. We can ask AI to optimise advertising and corporate extraction-or give every Australian community enough capability to ask much bigger questions. My earlier local-government submission was fundamentally about the Incomplete Ledger: the value, intelligence and civic contribution our existing systems fail to see. fileciteturn0file7 This submission extends that argument into the AI age. Do not build another incomplete ledger. Do not allow the intelligence of Australians-their lives, culture, work, Country, communities, memories, creativity and collective knowledge-to flow by default into systems they cannot see, change, leave or meaningfully negotiate with. Give people a sovereign layer. Give communities a sovereign layer. Give First Nations a sovereign layer. Give local government the infrastructure to participate rather than merely purchase. Give Queensland and the Commonwealth strong sovereign capability. Interoperate confidently with Oceania and the world. Keep enough diversity in the system that when one part fails, changes ownership, changes its values, changes its price, changes its government or simply turns out to have been a bad idea, Australia has somewhere else to go. That, to me, is a fair go for the future rather than a trap for people who never realised what they had agreed to. And if we can use the same infrastructure to help young Australians learn, communities govern themselves better, First Nations protect and develop their knowledge, scientists test ideas openly, local developers create new industries, people preserve their memories, councils simulate the future, neighbours respond to disasters and tens of thousands of cities learn to understand one another a little better- then I reckon that is a blooming good use of a supercomputer. [1] https://www.aph.gov.au/Parliamentary_Business/Committees/Senate/Environment_and_Communications/AIdatacentres48P [https://www.aph.gov.au/Parliamentary_Business/Committees/Senate/Environment_and_Communications/AIdatacentres48P](https://www.aph.gov.au/Parliamentary_Business/Committees/Senate/Environment_and_Communications/AIdatacentres48P) [\[2\]](https://www.pm.gov.au/media/ai-australias-interests) https://www.pm.gov.au/media/ai-australias-interests [https://www.pm.gov.au/media/ai-australias-interests](https://www.pm.gov.au/media/ai-australias-interests) [\[3\]](https://www.industry.gov.au/publications/national-ai-plan/introduction) https://www.industry.gov.au/publications/national-ai-plan/introduction [https://www.industry.gov.au/publications/national-ai-plan/introduction](https://www.industry.gov.au/publications/national-ai-plan/introduction) [\[4\]](https://www.industry.gov.au/publications/expectations-data-centres-and-ai-infrastructure-developers) https://www.industry.gov.au/publications/expectations-data-centres-and-ai-infrastructure-developers [https://www.industry.gov.au/publications/expectations-data-centres-and-ai-infrastructure-developers](https://www.industry.gov.au/publications/expectations-data-centres-and-ai-infrastructure-developers) [\[5\]](https://www.csiro.au/en/news/All/News/2026/May/Vetra-AI-infrastructure) https://www.csiro.au/en/news/All/News/2026/May/Vetra-AI-infrastructure [https://www.csiro.au/en/news/All/News/2026/May/Vetra-AI-infrastructure](https://www.csiro.au/en/news/All/News/2026/May/Vetra-AI-infrastructure) [\[6\]](https://www.industry.gov.au/news/australian-government-has-signed-memorandum-understanding-mou-tech-giant-microsoft) https://www.industry.gov.au/news/australian-government-has-signed-memorandum-understanding-mou-tech-giant-microsoft [https://www.industry.gov.au/news/australian-government-has-signed-memorandum-understanding-mou-tech-giant-microsoft](https://www.industry.gov.au/news/australian-government-has-signed-memorandum-understanding-mou-tech-giant-microsoft) [\[7\]](https://www.industry.gov.au/publications/national-ai-plan/spread-benefits) https://www.industry.gov.au/publications/national-ai-plan/spread-benefits [https://www.industry.gov.au/publications/national-ai-plan/spread-benefits](https://www.industry.gov.au/publications/national-ai-plan/spread-benefits) [\[8\]](https://auraofintelligence.github.io/) [\[25\]](https://auraofintelligence.github.io/) https://auraofintelligence.github.io/ [https://auraofintelligence.github.io/](https://auraofintelligence.github.io/) [\[9\]](https://auraofintelligence.github.io/strange-but-true/community-ledger.html) https://auraofintelligence.github.io/strange-but-true/community-ledger.html [https://auraofintelligence.github.io/strange-but-true/community-ledger.html](https://auraofintelligence.github.io/strange-but-true/community-ledger.html) [\[10\]](https://p4a.xyz/pages/aura-genesis.html) https://p4a.xyz/pages/aura-genesis.html [https://p4a.xyz/pages/aura-genesis.html](https://p4a.xyz/pages/aura-genesis.html) [\[11\]](https://www.oaic.gov.au/privacy/your-privacy-rights/your-personal-information/consent-to-the-handling-of-personal-information) https://www.oaic.gov.au/privacy/your-privacy-rights/your-personal-information/consent-to-the-handling-of-personal-information [https://www.oaic.gov.au/privacy/your-privacy-rights/your-personal-information/consent-to-the-handling-of-personal-information](https://www.oaic.gov.au/privacy/your-privacy-rights/your-personal-information/consent-to-the-handling-of-personal-information) [\[12\]](https://www.niaa.gov.au/news-and-media/framework-governance-indigenous-data) https://www.niaa.gov.au/news-and-media/framework-governance-indigenous-data [https://www.niaa.gov.au/news-and-media/framework-governance-indigenous-data](https://www.niaa.gov.au/news-and-media/framework-governance-indigenous-data) [\[13\]](https://www.niaa.gov.au/our-work/evaluations-and-evidence/indigenous-advancement-strategy-ias-evaluation-work-plan/priority-reform-four-community-data) https://www.niaa.gov.au/our-work/evaluations-and-evidence/indigenous-advancement-strategy-ias-evaluation-work-plan/priority-reform-four-community-data [https://www.niaa.gov.au/our-work/evaluations-and-evidence/indigenous-advancement-strategy-ias-evaluation-work-plan/priority-reform-four-community-data](https://www.niaa.gov.au/our-work/evaluations-and-evidence/indigenous-advancement-strategy-ias-evaluation-work-plan/priority-reform-four-community-data) [\[14\]](https://yumi-sabe.aiatsis.gov.au/about) https://yumi-sabe.aiatsis.gov.au/about [https://yumi-sabe.aiatsis.gov.au/about](https://yumi-sabe.aiatsis.gov.au/about) [\[15\]](https://www.niaa.gov.au/our-work/evaluations-and-evidence/indigenous-advancement-strategy-ias-evaluation-work-plan/indigenous-cultural-and-intellectual) https://www.niaa.gov.au/our-work/evaluations-and-evidence/indigenous-advancement-strategy-ias-evaluation-work-plan/indigenous-cultural-and-intellectual [https://www.niaa.gov.au/our-work/evaluations-and-evidence/indigenous-advancement-strategy-ias-evaluation-work-plan/indigenous-cultural-and-intellectual](https://www.niaa.gov.au/our-work/evaluations-and-evidence/indigenous-advancement-strategy-ias-evaluation-work-plan/indigenous-cultural-and-intellectual) [\[16\]](https://www.nist.gov/blogs/cybersecurity-insights/privacy-attacks-federated-learning) [\[27\]](https://www.nist.gov/blogs/cybersecurity-insights/privacy-attacks-federated-learning) https://www.nist.gov/blogs/cybersecurity-insights/privacy-attacks-federated-learning [https://www.nist.gov/blogs/cybersecurity-insights/privacy-attacks-federated-learning](https://www.nist.gov/blogs/cybersecurity-insights/privacy-attacks-federated-learning) [\[17\]](https://www.usenix.org/conference/usenixsecurity16/technical-sessions/presentation/tramer) https://www.usenix.org/conference/usenixsecurity16/technical-sessions/presentation/tramer [https://www.usenix.org/conference/usenixsecurity16/technical-sessions/presentation/tramer](https://www.usenix.org/conference/usenixsecurity16/technical-sessions/presentation/tramer) [\[18\]](https://www.nvidia.com/en-in/data-center/gb200-nvl72/) https://www.nvidia.com/en-in/data-center/gb200-nvl72/ [https://www.nvidia.com/en-in/data-center/gb200-nvl72/](https://www.nvidia.com/en-in/data-center/gb200-nvl72/) [\[19\]](https://www.amd.com/en/products/accelerators/instinct/mi400/mi455x.html) https://www.amd.com/en/products/accelerators/instinct/mi400/mi455x.html [https://www.amd.com/en/products/accelerators/instinct/mi400/mi455x.html](https://www.amd.com/en/products/accelerators/instinct/mi400/mi455x.html) [\[20\]](https://newsroom.olympics.com/record/1679) https://newsroom.olympics.com/record/1679 [https://newsroom.olympics.com/record/1679](https://newsroom.olympics.com/record/1679) [\[21\]](https://www.industry.gov.au/publications/national-ai-plan/keep-australians-safe) https://www.industry.gov.au/publications/national-ai-plan/keep-australians-safe [https://www.industry.gov.au/publications/national-ai-plan/keep-australians-safe](https://www.industry.gov.au/publications/national-ai-plan/keep-australians-safe) [\[22\]](https://www.spacex.com/spacexai/starmind) https://www.spacex.com/spacexai/starmind [https://www.spacex.com/spacexai/starmind](https://www.spacex.com/spacexai/starmind) [\[23\]](https://starlink.com/updates/starlink-version-3-satellites) https://starlink.com/updates/starlink-version-3-satellites [https://starlink.com/updates/starlink-version-3-satellites](https://starlink.com/updates/starlink-version-3-satellites) [\[24\]](https://www.aph.gov.au/About_Parliament/Parliamentary_departments/Parliamentary_Library/Research/Research_Papers/2024-25/The_deterrence_advantage_of_nuclear_powered_submarines) https://www.aph.gov.au/About_Parliament/Parliamentary_departments/Parliamentary_Library/Research/Research_Papers/2024-25/The_deterrence_advantage_of_nuclear_powered_submarines [https://www.aph.gov.au/About_Parliament/Parliamentary_departments/Parliamentary_Library/Research/Research_Papers/2024-25/The_deterrence_advantage_of_nuclear_powered_submarines](https://www.aph.gov.au/About_Parliament/Parliamentary_departments/Parliamentary_Library/Research/Research_Papers/2024-25/The_deterrence_advantage_of_nuclear_powered_submarines) [\[26\]](https://www.niaa.gov.au/artificial-intelligence-ai-transparency-statement) https://www.niaa.gov.au/artificial-intelligence-ai-transparency-statement [https://www.niaa.gov.au/artificial-intelligence-ai-transparency-statement](https://www.niaa.gov.au/artificial-intelligence-ai-transparency-statement)