THISISGRAEME

Kaitiaki in the Digital Age — Strengthening Indigenous AI, Knowledge & Climate Decision-Making at WIPCE 2025

WIPCE 2025 Aotearoa

Presented at WIPCE 2025 — Auckland

Indigenous AI at WIPCE 2025

Last week Conny Huaki and I had the honour of presenting at WIPCE 2025 here in Aotearoa. Our session explored how Artificial Intelligence can be designed in ways that strengthen kaitiakitanga, support iwi-led climate decision-making, and uphold Māori data sovereignty.

This work sits at the intersection of environmental stewardship, Indigenous knowledge systems, and emerging technology — a space where the questions are complex, but the opportunities are profound.

Our guiding principle is simple:

Technology must serve the whenua — not the other way around.


🪶 Why This Matters

For many whānau, environmental reports are difficult to access.
Critical information is buried in technical language, policy jargon, and legal frameworks that were never written with Māori audiences in mind.

As a result, hapū and iwi are often excluded from decisions that impact their:

These exclusions aren’t intentional — they’re structural.
And AI, unless shaped intentionally, can replicate the same inequities.


🜂 Our Innovation: Kaitiaki o Ngāti Mahuta GPT

(Demonstration prototype — not publicly available)

Through our mahi alongside iwi partners, we developed an AI tool designed specifically to help whānau interpret complex environmental reports using the frameworks, values, and worldview of Ngāti Mahuta.

This prototype:

It’s important to emphasise:
this is not a public, general-purpose GPT.
It is a customised, iwi-specific companion created for demonstration and exploration — a model of what iwi/hapū could build for themselves.

Slides from the presentation illustrate the challenge clearly:


🌋 Demo: Ironsands Mining Proposal

During the presentation, we ran a live demo analysing an Ironsands mining proposal.
The Kaitiaki tool generated:

This showed the audience what becomes possible when AI is grounded in tikanga and local governance.


🛡️ Māori Data Sovereignty

One of the strongest themes from the session was the absolute importance of Māori-led governance over:

Our next step is exploring Microsoft Aotearoa-based hosting as part of a sovereignty-first strategy.
This was a key point on slide 10:

“Māori Data Sovereignty is non-negotiable.”


🗣️ Post-Session Q&A — The Curly Questions

Many of you stayed behind afterwards with excellent questions.
Here are some of the big ones we covered:

Is this tool publicly available?
No — it’s a prototype created specifically for iwi-led exploration.

Can iwi or hapū build their own GPT?
Absolutely. We support Māori organisations through scoping, design, governance, build, and ongoing capability.

How do you ensure the AI respects tikanga?
Through Māori-led design, cultural review, red-line restrictions, and governance structures.

Where is the data hosted?
Currently in secure cloud environments, with a roadmap toward Microsoft Aotearoa hosting.

Does AI use a lot of water?
Yes — especially during training. For Indigenous communities, wai is a taonga, so transparency, hosting choice, and sustainable digital infrastructure matter.

Do you offer training or capability building?
Yes — workshops, governance frameworks, design sessions, and capability pathways.

A full extended FAQ is at the bottom of the page linked below for those who want to go deeper.


📺 Slides & Demo Video

You can view all presentation materials here:
👉 https://thisisgraeme.me/wipce2025-kaitiaki/


🕊️ Closing

“Kaitiakitanga extends to our data, our stories, and the code that carries them.”

This mahi is only possible because of the generosity of those who hold and protect the knowledge we’re working with.
Mihi to Conny Huaki and the Ngāti Mahuta whānau for their leadership, vision, and trust.

If your iwi, hapū, or organisation wants to explore tikanga-aligned AI pathways, climate decision support tools, or Indigenous-led tech governance, we’d love to kōrero.

He waka e hoe tahi ana — we navigate this together.

Exit mobile version