This briefing is a documentation artefact, not an application or proposal. The architecture and research described are presented for public reading; the deck was prepared for an in-person briefing to Te Kāhui Māori Atamai Iahiko on 19 June 2026.
The material is offered without an ask. The Civil-Society Proposal v1.2 (May 2026) referenced is operative, not roadmap; the architecture described is in production deployment. Polycentric governance and kaitiaki as a structural property are detailed in §3, §4, and §6 of the brief.
The architecture itself is the governance mechanism, not policy bolted onto a vendor-controlled stack. A7 (the seventh architectural invariant) is named at design time, before any autonomous agentic surface ships, so that future agentic surfaces inherit the refusal rather than retrofit it.
Federation works iwi-to-iwi today and the same envelope is being opened to third-party AI applications. No new substrate or governance to negotiate; parties federate under the established envelope. The specification is to be published so other Aotearoa AI work can federate under the same governance.
The paper engages directly with what Karaitiana Taiuru’s Kaupapa Māori AI Framework identifies as missing in earlier work. The architectural move is platform-as-substrate (not platform-as-authority). The paper is in draft, awaiting Māori peer review; this is documented in the paper itself.
Kaitiaki is a first-class signed field on records, not an annotation. The pattern of author=null + kaitiaki=operator on platform-authored records is the architectural form of the platform not claiming to speak for communities. Operator-authored records carry the operator’s identity as kaitiaki; community kaitiaki are populated only by the communities themselves.
The five-gate runtime is the answer to how the model behaves in production. Schema-level first-class classes — whakapapa, taonga, tikanga — distinguish this from a generic platform that has added te reo support. Training-discipline rules are explicit and enforceable: the platform refuses to retrain on community content without community authority.
A line of frontier research (TextGrad, GEPA, SkillOpt, Trace2Skill, EvoSkill) converges on the same substrate Village uses — frozen base model, evolving text artifact, automated validation gate. None of those research lines have a governance layer over what the artifact may evolve toward. Village’s polycentric steering and runtime cultural-boundary checking sit at the gate the optimiser landscape leaves open. The unpacked argument is set out in the source-material essay “Held in kōrero, not collapsed to a number — plural values, living organisations, and AI” (May 2026).
Sovereign infrastructure is established by absence-of-route, not by code-review trust: no code path can reach a US-controlled endpoint. The platform is open-source under EUPL-1.2 (reciprocal commons licence). Modest funding to date; NZ-partnership pathways are preferred over EU-grant cycles.
Engagement with Māori data and AI governance discipline is in the published record. Same-day revision when Taiuru’s 20 September 2025 critical analysis identified Te Mana Raraunga as superseded for AI work is verifiable in commit history. CulturalBoundaryChecker applying Mead’s Tikanga Test is in production code. Framework-consultation logging is operational.
The governance trajectory is a documented undertaking, not improvisation. The reference is to Board-level governance of My Digital Sovereignty Ltd itself — who runs the company that runs the platform — not solely to product-level consultation.
Three specific items are named where Māori-led work is the operative requirement: peer review on the steering paper, training-corpus discipline on the whānau model, and the cultural-reviewer step on the whakapapa schema. These items depend on Māori review; the operator does not have the authority to decide them unilaterally.
Tuakana/teina (elder-and-younger-sibling teaching) provides a structural fit between the panel’s succession-of-leadership discipline and the Village platform’s existing mutual-support pattern. The following slide enumerates concrete activities rangatahi could engage in.
This deck and the underlying brief are offered as documentation; no ask is attached to either.
The brief is the durable artefact; this deck is the in-meeting overview. Both link to per-section feedback widgets for reader comments at their own pace. Contact information sits with the brief.