Te Hiku Media

Te Hiku Media is a non-profit Māori iwi broadcaster and indigenous-AI organisation based in Kaitaia, in Te Hiku o Te Ika (the Far North) of Aotearoa New Zealand, that has become the most prominent indigenous-led pole inside the global movement for community-governed AI and data sovereignty. The organisation serves five Te Hiku-rohe iwi — Ngāti Kuri, Te Aupōuri, Ngāi Takoto, Te Rarawa, and Ngāti Kahu — and operates radio, television, archival, and natural-language-processing capability under a single mission: te reo Māori revitalisation on terms set by the Māori communities whose language is the subject of the work. Its working argument is that artificial intelligence "in its current form is based on the wholesale appropriation of existing culture", and that a credible alternative requires both the technical capability to build the models locally and the legal-and-tikanga apparatus to keep them under indigenous stewardship.

Founding and structure

Te Reo Irirangi o Te Hiku o Te Ika — the formal name carried alongside the trading name Te Hiku Media — was established in 1990 by tribal elders as an iwi broadcaster for the five Te Hiku iwi, with the stated purpose of carrying topical issues and gathering stories in te reo Māori and of maintaining te reo o te kāinga (the language of the home) as a living medium of community communication. The organisation operates as a charitable iwi-broadcasting structure rather than a commercial media business, runs radio (89.1 FM in the rohe) and an online television and archive platform, and has incrementally built out a science-and-engineering arm sitting alongside its broadcasting work. Peter-Lucas Jones (Te Aupōuri, Ngāi Takoto, Te Rarawa, Ngāti Kahu) is the Chief Executive and a leading voice in the Indigenous AI Working Group; Native Hawaiian engineer and physicist Keoni Mahelona is Chief Technology Officer and leads the data-science team; Suzanne Duncan is Chief Operating Officer; and Caleb Moses (Māori) is among the data scientists. The team's technical work grew out of a 2015 Vision Mātauranga Capability Fund placement and a 2017-2018 Ka Hao: Māori Digital Technology Fund grant that produced what the organisation describes as the first te reo Māori automatic speech recogniser, before the scale jump in 2019.

Papa Reo

The Papa Reo platform is the organisation's flagship indigenous-AI initiative. It was established with a NZ$13 million, seven-year Strategic Science Investment Fund placement awarded by the Ministry of Business, Innovation and Employment (MBIE) on 23 October 2019 — the only non-university project funded under that programme — and is positioned as a multilingual natural-language-processing platform whose stated mission is to "enable smaller indigenous language communities to develop their own speech recognition and natural language processing capabilities, ensuring that the sovereignty of the data remains with them and the benefits derived from these technologies goes directly to their communities". The Papa Reo collaborator set spans Dragonfly Data Science in Wellington, the Speech Group at the University of Cambridge, academics at the University of Oxford, the Mozilla machine-learning group, and the earlier-funder relationship with Te Mātāwai. The platform's two principal current tools are Kaituhi, an automatic transcription system for te reo Māori, and Kōrero Māori, the data-collection and tikanga-governed-access platform on top of which Papa Reo's models are trained, and it has progressively added a speech-to-text API and a synthesised te reo voice.

The Kaitiakitanga License

The legal-and-tikanga apparatus that underpins Te Hiku Media's data and model work is the Kaitiakitanga License. Its grounding principle is that data is not owned but is cared for under kaitiakitanga, and any benefit derived from that data flows back to the source community, reframing the Māori-language audio corpus the organisation curates as a stewardship relationship rather than as transferable intellectual property. The License grants data access only to organisations that agree to respect Māori values, stay within the bounds of consent given by speakers, and pass on any benefits derived from its use back to the Māori people, and explicitly prohibits the use of Te Hiku Media's tools for discrimination, surveillance, or tracking. The CTO frames the political stake bluntly: "Data is the last frontier of colonization", Keoni Mahelona has said; Peter-Lucas Jones has framed the counterfactual in equally direct terms, warning that absent indigenous-stewardship structures Māori-language data would be "used by the very people that beat that language out of our mouths to sell it back to us as a service". The License has since become a reference apparatus in the Indigenous data-sovereignty literature and a recurring template in the broader indigenous-AI movement.

Kōrero Māori, Whare Kōrero, and the model outputs

The organisation's signature crowdsourcing operation, the Kōrero Māori campaign, demonstrated the community model: more than 2,500 people signed up in 10 days to read over 200,000 phrases, producing more than 300 hours of labelled te reo Māori speech under the Kaitiakitanga License, with the resulting corpus then used to train Papa Reo's automatic speech recognition. The current production models reach 92 percent accuracy on te reo Māori transcription and 82 percent accuracy on bilingual te reo / English speech, built on the NVIDIA NeMo toolkit and run on eight NVIDIA A100 Tensor Core GPUs. The Whare Kōrero archive — the organisation's content platform — holds more than 30 years of digitised iwi-broadcast material and about 1,000 hours of native-speaker te reo, with around 20 Māori radio stations uploading content into the same archive. The shape of the work is recognisably grassroots: the corpus is built by community contribution under a community-set licence, the engineering and tikanga decisions are made by an indigenous-led team in the rohe, and the model outputs are released back to the community on terms the community can revise.

Posture in the movement

Te Hiku Media's place in the make-AI-good corpus is as the canonical indigenous-led, data-sovereignty-anchored, language-revitalisation organisation working at the intersection of grassroots Māori broadcasting and frontier natural-language-processing capability. Its theory of change is that Māori communities themselves should be "the makers of AI" for their language and knowledge — not consulted users of models trained elsewhere — and that the right object of intervention is the apparatus that determines who builds, who governs, and who benefits from those models. The organisation has been carried by international press, domestic Māori-led media commentary on AI and tino rangatiratanga, and international technical-policy venues including the ITU as the working example of an indigenous AI organisation with both the technical depth and the tikanga apparatus to make the data-sovereignty argument concrete. Within this corpus it is — as of May 2026 — the principal anchor for indigenous-led AI work, the principal Aotearoa New Zealand anchor, and the principal entity through which the Kaitiakitanga License framework enters the make-AI-good ecosystem.