From Oral Tradition to Artificial Intelligence
Memory as the Architecture of Future Optionality
The Question Every Civilization Has Answered in Architecture
Long before computers, before paper, before the first shelved library, every civilization faced one question and answered it in architecture: how do we ensure that what matters survives us? The answer was never only a matter of will or memory. It was a matter of structure — the material and social forms through which knowledge is held, carried across generations, and governed. Stone and inscription, papyrus and codex, the library, the university, the museum, and now the AI-mediated interface are not merely technologies. They are successive architectures through which memory is carried, encountered, and governed, each one solving the failure of the one before it and introducing a failure of its own.
The oldest of these architectures was never primitive. Indigenous and oral cultures built memory into living relationships — elders, song, ceremony, genealogy, and place. The historian Jan Vansina spent a career demonstrating that oral tradition is not the absence of a record but a disciplined method of transmission, with its own rules of verification and transmission across generations. Walter Ong showed that writing, when it arrived, did not simply store speech; it restructured thought itself, making possible forms of analysis that orality alone could not sustain. Each shift in the architecture changed not only what could be remembered but what could be thought.
The Egyptologist Jan Assmann gave this process its sharpest frame. He distinguished communicative memory — the roughly eighty to a hundred years that living people carry in conversation — from cultural memory, the deep inheritance that can span millennia. Between them lies what he called the floating gap: the zone where knowledge falls out of living reach and either crosses into durable cultural memory or disappears. Nothing crosses that gap on its own. Crossing it requires an apparatus — scribes and canons, temples and archives, curators and institutions. Memory that is not architected does not survive the floating gap. That is the whole of the matter, stated once: survival is an architectural achievement, not a natural one.
The architectures that followed each bought permanence at a price. The Library of Alexandria is remembered through the familiar story of a single catastrophic fire, but that story obscures the more important history. Modern scholarship finds that the institution declined gradually, across centuries: royal patronage lapsed, scholars' stipends were suspended, foreign scholars were expelled, and the institution that had sustained the collection slowly ceased to function. That is the more unsettling lesson. Memory architectures rarely die in a single catastrophe; they die when the institutions that finance, reproduce, and steward them lose the will or the means to continue.
The response, over later centuries, was to build sturdier institutions — the university and the museum emerged not merely as buildings but as rule-bound bodies designed to outlive any single custodian, with charters, successions, and professions whose whole purpose was to carry knowledge across the floating gap. Custody became a craft with standards of its own. Yet each solution carried forward the same latent fault: whoever governs the institution governs the memory, and an architecture is only ever as free as the hands that hold it.
Seen this way, artificial intelligence is the newest architecture through which human memory is encountered and interpreted, arriving with the same promise and the same hazard as writing, the library, and the museum before it. One distinction must be held throughout: the model is not the archive. It is increasingly the interface through which the archive becomes visible — the layer of access, synthesis, and interpretation that sits around memory, not the place where memory is kept.
The question AI raises is nonetheless ancient, because the thing each architecture has really governed was never the past. It was the future — the range of choices each one left open, or closed, for those who came after. That is the thread this essay follows, and it leads somewhere the usual conversation about heritage does not go.
Memory Is Future Optionality
Memory is almost always discussed as a debt to the past: heritage to be honored, identity to be preserved, ancestors to be remembered. That framing is not wrong, but it is incomplete, and its incompleteness has consequences. The deeper function of preserved knowledge is forward-facing. Every surviving body of knowledge is an option held open for people not yet born — a path they can still choose to take. Memory, in other words, is future optionality.
The word is precise, not decorative. In decision theory an option has value exactly because it preserves the ability to choose later, under conditions no one can foresee today. Knowledge behaves the same way. A preserved language, a pharmacopeia of medicinal plants, a navigational tradition, a body of customary law — each is a live option that a future generation can exercise when circumstances demand it.
The knowledge that is allowed to lapse is not merely a sad fact about the past; it is a foreclosed choice, a permanent narrowing of the decision space available to everyone downstream. This is why, at AXYNAO, we prefer to call such knowledge dormant rather than lost wherever that is true — because the distinction between dormant and destroyed is precisely the distinction between an option recoverable and an option gone.
Optionality also compounds, which is why the stakes are structural rather than sentimental. Each option retained becomes a platform on which further options can be built; each one foreclosed is a branch the tree can no longer grow. A revived language carries with it a cosmology, a taxonomy of a local ecology, a repertoire of story and law — and each of those, once live again, opens further choices in turn. The loss runs the same way in reverse: the disappearance of one tradition quietly removes options its inheritors will never know they were denied.
A concrete case makes the abstraction vivid. For generations, bodies of ecological and medicinal knowledge held by particular communities — which plant steadies a fever, which seabird's flight marks a current, which planting order survives a drought — were dismissed by outside institutions as folklore and allowed to lapse. Some of it, recovered late, has since proven to encode exactly the adaptive insight a warming and volatile world now needs. The point is not that every fragment will one day prove useful; most options are never exercised, and no institution can or should keep everything.
The point is that foreclosure is irreversible in a way that retention is not. To keep an option open costs storage and governance; to let it lapse costs a choice that no later effort can reconstitute. A serious knowledge institution is therefore not a hoarder but a steward of possibility — deliberate about what it keeps, and acutely aware that the cheapest-looking economy, quietly letting the marginal thing go, is often the most expensive decision it will ever make.
It is worth being exact about what is new here, because part of it is not. Heritage economics has recognized option value for decades: standard valuation frameworks treat the possibility of future use as a real component of a heritage asset's worth, and archival thinkers have extended the same logic to records whose future research value cannot yet be known. AXYNAO does not claim to have discovered that preservation carries option value. The argument goes one step further, and the step is the whole contribution.
A memory option is not preserved merely because the underlying object survives. It stays alive only if some legitimate custodian can still retrieve it, authenticate it, interpret it, govern its use, and change the infrastructure through which it is reached. Memory therefore carries two values at once: option value, which survival protects, and governance value, which only sovereignty protects. Preservation keeps the possibility alive. Sovereignty keeps the authority to exercise it. The claim this essay is prepared to defend is precise: the value of preserved memory depends on an architecture that preserves not only the future option, but the legitimate authority to exercise it.
This is, in the end, a question about flourishing. More than two thousand years ago Aristotle named the destination: eudaimonia, human flourishing through the exercise of reason, virtue, and judgment. The architectural question developed here is a different one — what preserves future access to the conditions under which agency and judgment remain possible? That is the first principle on which AXYNAO is built: architectures determine what remains possible, and what remains possible determines whether individuals, institutions, and civilizations keep the agency to shape their own future. Memory is one of those architectures. To preserve its optionality is to keep the door to flourishing open for people who will never know whose custody held it so.
The Living Archive and Its Newest Predator
Many Indigenous cultures preserve knowledge through living communities rather than centralized institutions. This is a genuine architecture, not a deficit of one. Its strengths are real: knowledge stays embedded in context, bound to the relationships and places that give it meaning, and distributed across a community rather than concentrated in a single fragile store. Its pressures are equally real — language loss, displacement, and the passing of elders can thin a living archive faster than any fire took a library.
The scholarship here is mature, and we do not propose to re-argue it. The volume edited by Tahu Kukutai and John Taylor established Indigenous data sovereignty as a field; the CARE Principles for Indigenous Data Governance gave it an operating framework; and the United Nations Declaration on the Rights of Indigenous Peoples established the authoritative international minimum standard. Communities hold inherent rights over knowledge about them and their territories. We take that as settled and ask the architectural question that follows from it: what structure actually secures those rights as the custody of knowledge moves from the archive to the AI interface?
A new pressure has joined the old ones, and it is subtler because it wears the mask of preservation. Nick Couldry and Ulises Mejias named the underlying mechanism data colonialism — the appropriation of human life as extractable data, the extraction logic of historical colonialism transposed onto a datafied world. Applied to a living archive, it works like this: a model trained on knowledge taken without consent strips that knowledge of its provenance and severs it from the community that holds it.
The archive is preserved in form and dispossessed in substance. Assmann's floating gap reopens by other means. Knowledge may continue to influence a model statistically, shaping its outputs, even after the source, the provenance, and the community context have disappeared from any practical reach. What looks like survival is often only residue: influence without the record, pattern without the custody that made it mean anything.
The architectural conclusion is blunt. Extraction is not preservation. A model that has ingested a language is not the custodian of that language, and preservation that removes custody reduces optionality even as it appears to increase access — because a community that can no longer govern its own knowledge can no longer decide how that knowledge is exercised, by whom, or for what. Access without custody is optionality transferred, not optionality preserved.
Sovereignty Is a Hard Requirement, Not a Feature
If custody is the load-bearing wall of this architecture, sovereignty is the foundation beneath it. Sovereignty over knowledge — a community's authority to determine what is preserved, shared, protected, or withheld — is not an ethical finish applied at the end. It is the precondition for the structure to stand at all. Here the architecture admits no compromise: sovereignty over Indigenous data is a hard requirement, not a feature to be added later.
What makes this more than aspiration is that the legal scaffolding now exists and is hardening. The Declaration on the Rights of Indigenous Peoples set the principle in 2007. The Nagoya Protocol, adopted in 2010 and in force since 2014, built access-and-benefit-sharing obligations around traditional knowledge associated with genetic resources. And in May 2024, after more than two decades of negotiation, the World Intellectual Property Organization adopted its Treaty on Intellectual Property, Genetic Resources and Associated Traditional Knowledge — the first WIPO treaty to contain provisions written specifically for Indigenous Peoples and local communities.
It requires patent applicants whose inventions draw on genetic resources or associated traditional knowledge to disclose the origin of those resources and the community that provided the knowledge. More than thirty national and regional disclosure regimes already existed; what is new is that, for the first time, a multilateral treaty writes that duty into the international patent system itself, where quiet appropriation can be made legible across borders rather than nation by nation. Only a handful of states have so far ratified — well short of the fifteen the treaty needs to enter into force — but its direction of travel is unambiguous, and institutions should architect for the world it describes rather than the one it leaves behind.
This is not a single tradition but a global movement whose instruments keep converging on the same architecture. In Canada, the First Nations principles of OCAP — ownership, control, access, and possession, stewarded by the First Nations Information Governance Centre — make possession, the physical holding of data by the community itself, the mechanism that turns ownership from a claim into a fact. Aotearoa New Zealand has Te Mana Raraunga; Australia has Maiam nayri Wingara; the Sámi have articulated data principles across the Nordic states. Different peoples, different instruments, one recurring design decision: authority over the record must rest with those the record is about.
Sovereignty in practice means two things together: communities must be the custodians of their own records, and they must hold that custody with established and proven independence — able to govern the architecture, not merely be consulted about it. Elinor Ostrom's work supplies the grammar. Her study of the commons showed that durable shared resources are neither pure open access nor pure enclosure; they are governed by rules the community itself sets, monitors, and can enforce. Extended by Charlotte Hess and Ostrom to knowledge as a commons, the lesson is exact: a knowledge commons survives not because it is open or closed but because it is governed by those with the standing to govern it.
This invites an obvious objection, and it deserves a direct answer. Is sovereignty not itself a kind of enclosure — does restricting knowledge not limit the good it might do if it flowed freely? The commons tradition dissolves the dilemma. The opposite of enclosure is not unrestricted openness; it is governed access. A knowledge commons is neither a locked vault nor an open field but a resource whose community sets the terms on which it is shared — and governed access is often more generous than extraction, because a community confident in its control can afford to share what a dispossessed one must hide. Sovereignty, built properly, is not the enemy of openness. It is the precondition for an openness that lasts.
Without custody and independence, preservation becomes only a more durable form of dispossession — a better-built vault whose keys belong to someone else. That is not a knowledge institution. It is a holding facility.
It Already Works: Community-Governed AI
None of this is theoretical, and the proof is deliberately drawn from outside the Western institutions that usually dominate the conversation. In Aotearoa New Zealand, Te Hiku Media built speech-recognition systems for te reo Māori from recordings contributed by the community, and governed both the data and the tools through its Kaitiakitanga License — an instrument built on the Māori principle of guardianship, in which data is not owned but cared for, and any benefit returns to the community it came from. This is deliberately not an ordinary open-source release: the license forbids using the tools for surveillance, tracking, or discrimination, expressly prohibits mining Māori data or building Māori corpora from them, and keeps the underlying data from being released openly at all, with access instead governed by tikanga, Māori custom.
The organization had to fend off corporate interest in its corpus precisely because the corpus is valuable, and the license is what let it collaborate without surrendering control. What is most telling is what has travelled since: at least one other Pacific-language effort — a Cook Islands Māori speech-recognition project — has adopted the same Kaitiakitanga licensing architecture for its own corpus. The architecture travelled, not the data, which is exactly the point.
On the other side of the world, the Masakhane movement — the name means “we build together” in isiZulu — has drawn hundreds of researchers across more than twenty countries into a participatory effort to build natural-language tools and translation for dozens of African languages, led by the people who speak them. It has shown that participatory development can scale: it has produced usable datasets and translation benchmarks across dozens of African languages, and it brings into the technical process the linguistic and societal knowledge that remote research pipelines routinely lack. Two continents, two architectures, one principle: AI can expand the optionality of a living archive when the community that holds the language also governs the model, rather than the other way around.
Honesty requires naming the vulnerability these projects share. Participatory work remains exposed to co-optation — community datasets can be absorbed by better-resourced commercial actors who return nothing to the communities that produced them. This is not an argument against the approach; it is an argument for the governance described above. The proof and the caution point in the same direction. Community-governed AI works, and it works only as well as the custody and sovereignty that protect it.
Digitizing Is Not Preserving
There is a comfortable assumption, widespread in memory institutions, that to digitize is to preserve — that once a thing is scanned it is safe. The preservation record says otherwise, and the warning comes from the architects of the digital world itself.
Vint Cerf, one of the engineers of the internet, has warned for years of a digital dark age: a future unable to read its own past because the formats and software needed to open today's files have gone obsolete, a decay he calls bit rot. His proposed remedy — a kind of digital vellum that would preserve not only the files but the means to read them — is really an admission that bits alone are not memory. A file no one can open is not preserved; it is merely stored. UNESCO's Memory of the World programme exists precisely because preservation is an active, governed, continually resourced discipline rather than a single act of copying. An archive is a verb, not a noun.
Where serious preservation has been attempted, its logic is instructive — and it separates two things the word “safe” usually blurs together. The Arctic World Archive, opened on Svalbard in 2017, stores the world's documents as data printed onto photosensitive film and sealed offline some three hundred metres beneath the permafrost in a decommissioned coal mine, built to be read centuries from now with little more than light and magnification, immune to power loss, format drift, and network attack. That is durability of the medium — a solution to physical and technological failure. But it is not, by itself, independence of control: Svalbard falls under the full sovereignty of a single state, as every physical site falls under someone's.
The second kind of resilience has to be designed separately. The LOCKSS Program at Stanford provides the decentralized preservation architecture; CLOCKSS applies it through twelve geographically dispersed archive nodes governed by a nonprofit collaboration of libraries and publishers. LOCKSS states the underlying principle explicitly: many copies are not enough if one administrator, one organization, or one software instance can reach them all.
The lesson is one AXYNAO would state plainly: distribution of storage is not distribution of sovereignty. Copies reduce physical failure; independent control reduces capture. A thousand copies under one control plane remain one architecture of control.
The architectural consequence follows directly. A digital copy held under no custody, in a format quietly obsolescing, on infrastructure controlled by someone else, is not preservation. It is deferred loss wearing the costume of safety — optionality that appears to be banked and is not. The institution that mistakes digitization for custody has not secured its collection's future. It has only changed the way that future can be taken from it.
Reconnection: The Technology of the Dormant
The same technology that can dispossess can also reconnect, and this is the part of the story that justifies the effort. Consider the Herculaneum scrolls — hundreds of papyrus rolls carbonized by the eruption of Vesuvius in 79 AD, so fragile that for two thousand years any attempt to open them destroyed them. Through the Vesuvius Challenge, machine learning has now done what no human hand could: the scrolls are being virtually unwrapped and read without being touched. The first word emerged in 2023, passages of Greek philosophy in 2024, and in June 2026 the first complete scroll was read end to end, in full, having never been opened. Knowledge that was dormant for two millennia — not destroyed, dormant — has been reactivated.
This is the general case, not a curiosity. A great deal of human knowledge is not lost but fragmented and dormant: scattered across museums, universities, archives, private collections, archaeological reports, and living communities, separated by geography, time, and institutional walls. The distinctive power of artificial intelligence here is relational. It can propose connections among dispersed records, languages, artifacts, and oral traditions that no single scholar could ever hold in view at once. This is what we have called the Global Memory Expedition — a search conducted not for treasure but for fragments whose true destination is meaning, reassembled.
There is one condition on which the whole hope depends, and it returns us to custody. Researchers and communities must remain the final authorities; AI is an instrument of discovery and reconnection, never a successor custodian of the memory it helps recover. As our Custody Covenant puts it, the model must never become the archive. Reconnection expands optionality only so long as it stays governed. Ungoverned, the same relational power that reunites fragments can just as easily launder them of provenance, and the expedition that set out to recover meaning returns having dissolved it.
Beyond the Reach of Any Single State
Independence must have somewhere to stand. A record kept under the continued goodwill of a single government is not yet sovereign, because one change of policy can reach it — and history shows how quickly achievements built over decades can be undone by a single stroke of a pen. Jurisdiction, in other words, is part of the architecture of memory, not a technical detail of where files happen to sit.
The design goal is not to exchange one sovereign for another. Moving an archive from one state to another simply trades one government's reach, and one government's politics, for another's. The goal is independence from any single state's power to compel or to cut off access. This is the optionality principle applied to geography: never a single point of failure. Michel-Rolph Trouillot showed that power shapes the historical record at every stage — what becomes a source, what enters an archive, what is later deemed significant. Distributing custody is how an architecture distributes that power, so that no single authority holds the whole of a people's memory in its hand.
It is worth being precise about why relocation alone is not enough. Reach follows control, not geography. A record does not escape a government's authority merely by sitting on foreign soil if the body that holds it, or the company that runs the servers, remains within that government's jurisdiction. Independence is therefore a property of who controls the record, and of how many separate authorities it rests under — not of the map coordinates where the data happens to live. An architecture that grasps this stops asking only where to store and starts asking who can compel, and arranges itself so that the answer is never any one party.
The mechanisms that achieve this are nameable even where the specifics are not yet settled: custody vested in community-controlled vehicles rather than in any one state's institutions; records mirrored across more than one jurisdiction, so that a seizure, an order, or a reversal in one place cannot capture the whole; and encryption under community-held keys, so that even a compelled host surrenders nothing usable. This is not a matter of choosing a particular country, and AXYNAO prescribes none. It is a matter of installing the mechanisms that make dependence on any single authority unnecessary. Which locations and arrangements, and how they are to be judged, belongs to the instrument built to measure them.
What Preserved Optionality Requires
If optionality can be measured — and the claim of this essay is that it can — then it must be possible to say what a preserved option actually consists of. It is not enough to ask whether an object still exists. A memory asset preserves future optionality only when five conditions hold together. The source must still exist. It must remain technically legible, readable by some available means rather than locked in a dead format. Its provenance and context must remain reconstructable, so that what it is and where it came from can still be established. A legitimate custodian must retain authority over its access and use. And the knowledge must be able to survive a change of provider, institution, or jurisdiction — substitutable in its infrastructure without being lost.
Existence, legibility, provenance, governance, substitutability: remove any one and the option quietly dies, however intact the others appear. A perfectly preserved file in an unreadable format is gone. A legible record whose provenance has dissolved can no longer be trusted or exercised. A fully documented archive whose custodian has lost authority over it is an option someone else now holds. This is what it means to say that preservation is necessary but not sufficient — and it is the architecture a knowledge institution can actually be measured against. These conditions are observable through evidence and can therefore be classified institutionally, without pretending that the resulting sovereignty can be reduced to a single universal number.
Three sentences hold the whole of it together. Preservation protects the object. Custody protects the relationship between the knowledge and the community it belongs to. Sovereignty protects the choice — the standing to decide, later and under conditions no one can foresee, how the option is used. An institution that secures all three has not merely kept a collection. It has kept the future's ability to act on it.
The Question for Every Institution
This is not, in the end, only a question for Indigenous communities. Museums, universities, archives, nations, and research institutions all stand before the same one, because all of them are now deciding how AI will touch what they hold. Will artificial intelligence merely preserve information — or will it preserve humanity's ability to govern and reconnect its own knowledge? The difference between those two futures is the difference between a civilization that keeps its options and one that quietly spends them.
So the question AXYNAO leaves with every museum, university, archive, and community is this:
If your institution adopted AI tomorrow to preserve what it holds, who would decide what is preserved, shared, protected, or withheld — and would that decision expand the choices available to future generations, or quietly narrow them?
An institution that can answer that has not merely digitized a collection. It has designed the architecture through which its knowledge will survive, and kept the authority to govern it. That is where preservation becomes sovereignty — and where this inquiry meets the discipline of decision architecture: the study of how choices are structured so that optionality is widened rather than spent.
AXYNAO already measures this. Its Epistemic Sovereignty Quadrant maps whether an institution governs its own standing or has it governed for it, and its Institutional Knowledge Custody Covenant sets the standard for whether a record-holder remains the independent custodian of what it keeps. Both are published; this essay builds on them. The next edition of the Quadrant extends that same measurement to the institutions and communities that steward Indigenous knowledge, and will introduce the independent locations, legal arrangements, and distributed-custody mechanisms that let a community hold its records beyond the reach of any single state.
Architecture preserves memory.
Memory preserves optionality.
Optionality preserves agency.
Agency preserves human flourishing.
Artificial intelligence should become not only a technology of prediction, but a technology of reconnection — and perhaps the greatest discovery of the AI era will not be artificial intelligence itself, but humanity rediscovering its own memory, and keeping the authority to govern it.
Stoic Reflection
The Stoics taught that nothing we hold is truly our own — we are stewards of what was handed to us, bound to pass it on undiminished. Memory is the oldest form of that trust. What a generation fails to carry forward it does not merely lose; it quietly decides, on behalf of everyone who comes after, that the choice will no longer exist. To preserve is not nostalgia. It is the discipline of leaving the future its options.
Sources
Memory, Orality, and the Archive
Jan Assmann. Cultural Memory and Early Civilization: Writing, Remembrance, and Political Imagination. Cambridge University Press, 2011 (orig. Das kulturelle Gedächtnis, 1992). — Supplies the distinction between communicative and cultural memory and the “floating gap” — the argument that survival across generations requires an institutional apparatus, i.e. an architecture.
Walter J. Ong. Orality and Literacy: The Technologizing of the Word. Routledge, 1982. — Supports the claim that each memory architecture changes not only what is stored but what can be thought.
Jan Vansina. Oral Tradition as History. University of Wisconsin Press, 1985. — Establishes oral tradition as a disciplined method of transmission rather than the absence of a record.
Michel-Rolph Trouillot. Silencing the Past: Power and the Production of History. Beacon Press, 1995. — Supports the argument that power shapes which records survive — grounds the case for distributing custody so no single authority holds the whole record.
Heritage and Archive Economics (Option Value)
Historic England. Heritage Counts — heritage valuation and the Total Economic Value framework. — Establishes option value and bequest value as standard components of heritage economics; the prior art AXYNAO builds beyond rather than claims to originate.
UK Department for Culture, Media & Sport. Culture and Heritage Capital programme — economic valuation of collections. — Applies option value, and real-options analysis, to museum and archive collections.
Indigenous Knowledge Sovereignty and Governance
Tahu Kukutai and John Taylor (eds.). Indigenous Data Sovereignty: Toward an Agenda. ANU Press, 2016. — The founding volume of the field; taken as settled ground for the claim that communities hold inherent rights over knowledge about them.
Global Indigenous Data Alliance (GIDA). CARE Principles for Indigenous Data Governance. 2019. — The operating framework for community determination of what is preserved, shared, protected, or withheld.
Local Contexts (Jane Anderson and Kim Christen). Traditional Knowledge (TK) and Biocultural (BC) Labels. Established 2010. — Supports the claim that knowledge can remain under community governance while still being shared across generations.
First Nations Information Governance Centre (FNIGC). The First Nations Principles of OCAP® — Ownership, Control, Access, Possession. Established 1998. — A Canadian instance of the same architecture: possession — the community’s physical control of its data — as the mechanism that makes ownership enforceable.
Nick Couldry and Ulises A. Mejias. The Costs of Connection: How Data Is Colonizing Human Life and Appropriating It for Capitalism. Stanford University Press, 2019. — Names the mechanism of “data colonialism” — the basis for treating non-consensual AI training as dispossession, not preservation.
International Legal Instruments
United Nations. Declaration on the Rights of Indigenous Peoples (UNDRIP). 2007. — The authoritative international minimum standard for Indigenous authority over knowledge and territory (a General Assembly declaration, not a binding treaty).
Nagoya Protocol on Access and Benefit-sharing (to the Convention on Biological Diversity). Adopted 2010; in force 2014. — Builds benefit-sharing obligations around traditional knowledge associated with genetic resources.
WIPO. Treaty on Intellectual Property, Genetic Resources and Associated Traditional Knowledge. Adopted 24 May 2024. — The first WIPO treaty with provisions specific to Indigenous Peoples; the first multilateral instrument to write a patent disclosure-of-origin duty into the international system (30+ national and regional regimes predate it). Not yet in force.
UNESCO. Convention for the Safeguarding of the Intangible Cultural Heritage. 2003. — Frames oral traditions, ceremony, and language as a distinct architecture of memory under active pressure.
UNESCO. International Decade of Indigenous Languages (2022–2032), UN General Assembly Resolution A/RES/74/135. — The global framework through which communities, states, and institutions can act; its digital and AI priorities bear directly on preservation.
Community-Governed AI in Practice
Te Hiku Media / Papa Reo (Aotearoa New Zealand). Kaitiakitanga License and te reo Māori speech models. — AI governed through a license that forbids surveillance and discrimination and prohibits mining its language data or building corpora from it, while keeping the underlying data under governed access. A Cook Islands Māori speech-recognition project subsequently adopted a localized form of the Kaitiakitanga License for its own corpus, demonstrating that the governance architecture can travel without the data.
Masakhane. Participatory NLP and machine translation for African languages (founded 2019). — A pan-African, community-led architecture; demonstrates that participatory development scales, producing datasets and benchmarks across dozens of African languages and bringing linguistic and societal knowledge into the process.
Preservation, Loss, and Reconnection
Vint Cerf. “Digital dark age” / “bit rot” / “digital vellum” (AAAS, 2015); UNESCO Memory of the World Programme. — Supports the claim that digitizing is not preserving — that preservation is an active, governed discipline, not a one-time copy.
Vesuvius Challenge / EduceLab, University of Kentucky (Prof. Brent Seales). Herculaneum scrolls virtually unwrapped by machine learning — first word October 2023; Greek passages February 2024; first complete scroll read end to end June 2026. — The central example that AI can reactivate knowledge that was dormant rather than destroyed.
Arctic World Archive, Svalbard (Piql AS; opened 2017). — A precedent for durable, offline preservation of the medium (physical and technological resilience); but under full Norwegian sovereignty, which is why durability of the medium must not be confused with independence of control.
LOCKSS / CLOCKSS (Stanford Libraries; CLOCKSS Archive, launched 2006). — Distributed, decentralized preservation across a dozen libraries worldwide; its governing principle is that many copies do not protect content if one administrator can reach them all, so no single participant controls every copy.
Knowledge as a Commons
Elinor Ostrom. Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press, 1990. — The governance grammar: durable shared resources are neither open access nor enclosure but community-governed.
Charlotte Hess and Elinor Ostrom (eds.). Understanding Knowledge as a Commons: From Theory to Practice. MIT Press, 2007. — Extends commons governance to knowledge — the basis for “custody held by those with standing to govern it.”
AXYNAO Instruments Referenced
AXYNAO. The Architecture of Recognition — the AXYNAO Epistemic Sovereignty Quadrant. 2026. — Maps whether an institution governs its own standing or has it governed for it; the published instrument this essay’s governance argument builds on.
AXYNAO. The Institutional Knowledge Custody Covenant. 2026. — Sets the standard for provider-independent custody, provenance, and source-level access — “the model must never become the archive.”
AXYNAO. Aristotle and Kling. 2026 (axynao.com/post/aristotle-and-kling). — The AXYNAO first principle this essay applies: architectures determine what remains possible, and what remains possible determines whether individuals, institutions, and civilizations keep the agency to shape their future.
AXYNAO Reflection
AXYNAO builds the architectures through which meaning survives — art placed into space, and knowledge placed into institutions still able to govern it. Architecture determines what endures; and what endures determines what remains possible.
AXYNAO / Erik Kling, “Diplom Oekonom” University Of Hohenheim
Architecture Determines Meaning — Speaking & Advisory
Tags: Architecture of Meaning · Memory · Optionality · Indigenous Knowledge · Data Sovereignty · Artificial Intelligence · Cultural Custody
Comments