Kaitiakitanga is intergenerational responsibility because, in the Māori worldview, land is not owned but cared for on behalf of generations to come. An academic paper proposes adopting this concept alongside Hozho as a foundation for AI governance. The distinction between caring for and owning reveals exactly what is missing from nearly every current regulatory structure, which remain focused on immediate risk and compliance.
The UN's global dialogue on these matters moves forward without formal Indigenous representation in the core decisions. The paper's authors know this terrain well and point out how the dominant approaches end up short-termist. They offer a vocabulary of balance that the technical jargon of alignment and mitigation simply doesn't have.
The proposal deserves a careful read before any judgment is passed. This isn't corporate PR: the gesture introduces terms like Indigenous data sovereignty into circles that previously overlooked them, performing a cultural translation that didn't exist in those forums before.
What does it actually mean to incorporate Kaitiakitanga if the communities that give it meaning have no binding power over how it's applied? The answer isn't comfortable. The concept is inseparable from the whanau, hapu, and iwi structures that determine who exercises stewardship and under what rules. Taking only the word without transferring authority hollows out the signifier of its operational function.
I've seen similar dynamics in organizations that adopt new language without touching the real decision-making structures. Las Piedras No Mienten has examined how this partial selection of ideas repeats itself throughout history: the surface gets taken while whatever would cost real power gets discarded. Without the underlying authority, cultural borrowing becomes ornament.
The Mapuche case in Chile shows the same tendency on concrete ground. Infrastructure and lithium extraction projects move forward after consultations that satisfy ILO Convention 169 in letter but not in spirit. The objection gets documented and the project proceeds regardless. Consultation turns into an administrative formality instead of a real veto.
This pattern persists because technical decisions are designed first among powers and experts. Inclusive language arrives afterward, as a coating. The UN Framework Convention on Climate Change emerged without a formal Indigenous voice. Years later, the Permanent Forum on Indigenous Issues was created, which even today plays only a consultative role with no binding capacity.
AI governance is reproducing that same sequence at greater speed. First comes the design among states and companies. Then Māori and Navajo concepts enter through the academic door as available inputs. This doesn't break with the extraction of ancestral knowledge already documented in debates over language models — it just adds one more sophisticated layer of rhetoric.
Who benefits from Indigenous vocabulary circulating without the power that sustains it? Companies gain ethical legitimacy without ceding control over data or compute. States with active projects on Indigenous territories gain discursive cover. The inclusion discourse the UN promotes acquires symbolic weight precisely when it offers the least structural change. None of these gains reach the communities being cited.
There's a clear difference between genuine cultural translation and functional appropriation. The former requires transferring authority along with the words. The latter copies only the visible part and discards the rest. Something similar happens in the design of complex systems, where the name of a function gets borrowed but not its internal logic. I still don't have this tension fully resolved, though I keep thinking about it.
What's missing isn't more vocabulary in papers. What's missing is posing a structural question that neither the UN dialogue nor the corporate summits have confronted head-on: what fraction of decision-making authority over infrastructure, data extraction, and deployment on Indigenous territories would states and companies actually be willing to give up — as a binding veto, not mere consultation. Until that question gets answered, any mention of Kaitiakitanga remains a gesture without operational consequences.
The Permanent Forum on Indigenous Issues has operated for decades without real power over the climate decisions that most affect the territories it represents. If AI governance follows the same playbook, it's likely that within a few years there will exist an Indigenous advisory body with symbolic presence but no ability to stop a single data extraction project. What would happen if, instead of borrowing the language, its speakers were invited to rewrite the rules from scratch?
Sources
1. United Nations Framework Convention on Climate Change and the creation of the Permanent Forum on Indigenous Issues
2. International Labour Organization Convention 169 on Indigenous and Tribal Peoples
3. Public documentation on prior consultation processes for lithium projects in Mapuche territory, Chile
