--- title: Te Tiriti Critique of a NZ AI Safety Lobby Group description: Explore the critique of a NZ AI safety lobby group and its lack of engagement with Te Tiriti o Waitangi and Māori sovereignty. url: https://www.taiuru.co.nz/nz-ai-safety-lobby-group/ markdown_url: https://www.taiuru.co.nz/nz-ai-safety-lobby-group.md type: post slug: nz-ai-safety-lobby-group author: Karaitiana Taiuru published: 2026-11-09T22:55:19Z modified: 2026-11-09T22:55:19Z language: en-NZ taxonomies: category: - Case Studies word_count: 1977 tokens: 3419 --- # Te Tiriti Critique of a NZ AI Safety Lobby Group Good Ancestors New Zealand is setting out to shape Aotearoa’s national AI safety agenda without any reference to Te Tiriti o Waitangi, Māori data sovereignty or tangata whenua. Across its website and all twelve of its July 2026 papers, Māori appear once: as the language some scammers use. The only acknowledgement of where it operates is a footer that reads “Based in Aotearoa New Zealand”. The organisation describes itself as an independent, not-for-profit policy organisation based in Wellington (Good Ancestors New Zealand, n.d.-a). It was launched in 2026 by co-founders Dylan Chambers and Daniel Foote and shares its name and mission with Good Ancestors in Australia (Good Ancestors New Zealand, n.d.-b). The Australian organisation describes itself as a charity dedicated to reducing existential risk (Good Ancestors Policy, 2023), and Open Philanthropy has funded its Australian advocacy (Open Philanthropy, n.d.). The New Zealand site does not disclose its funding. I do not question their competence or their sincerity. Their papers are clear and their recommendations are practical. My concern is that an imported frontier-AI safety model is being offered to the Crown as the national agenda, with Māori left out of it entirely. That is the disassociation of Indigenous frameworks from mainstream standards in its plainest form. Architecture versus decoration Good Ancestors fails the architecture-versus-decoration test before it is applied, because there is not even decoration. The test asks whether Māori provisions are structurally operative or merely ornamental. Here there are no Māori provisions of either kind. I reviewed the website, all six July 2026 risk papers, all six July 2026 policy papers, the AI election scorecard and the November 2026 risk update. None of them mentions Te Tiriti, Māori data sovereignty, iwi or tikanga. The only reference to Māori is a note in the scams paper that some campaigns target New Zealand businesses in te reo Māori ([Good Ancestors New Zealand, 2026c](https://www.goodancestors.org.nz/papers/scams-and-abuse)). Te reo appears as a vector for harm, not as a taonga to protect. Risk framing. The only discrimination example is that AI hiring systems have rejected qualified women and minorities (Good Ancestors New Zealand, n.d.-c). This is an offshore example. Nothing addresses algorithmic harm to Māori in our own justice, child protection or health systems, the misuse of te reo Māori, or the extraction of mātauranga Māori by offshore models. Values. One stated value is recommendations that fit within existing political and institutional frameworks (Good Ancestors New Zealand, n.d.-b). Te Tiriti is a constitutional framework already in place, yet it is absent from that frame. Election scorecard. Te Pāti Māori answered yes to every survey question, but none of the six criteria asked about Te Tiriti or Māori data (Good Ancestors New Zealand, 2026a). The party with the most developed Tiriti position on AI was scored only on a frontier-risk template. An AI safety agenda for Aotearoa that does not mention Te Tiriti is not neutral. It is a choice to govern Māori data and mātauranga without Māori. The election scorecard and the harm Māori already carry Taiuru & Associates raised the scorecard’s silence on Te Tiriti with Good Ancestors the day after it was released. They replied that the questions and scoring criteria were final, and that changing the scorecard now would be “unfair to the parties who have responded” (Good Ancestors New Zealand, personal communication, October 20, 2026). They also cited transparency, consistency and the limited time before the election. That answer gets fairness the wrong way round. The omission was not a late addition someone asked for. It was a choice made when Good Ancestors designed the questions. Consistency with a scorecard that leaves out Te Tiriti only keeps the omission in place. The unfairness falls not on parties but on Māori, who carry AI harms that no party was ever asked to address. Good Ancestors also said it will keep clarifying and updating parties’ positions as they make further statements, and invited suggestions for its About and Method page. That leaves room to act now. At a minimum, the Method page should state plainly that the scorecard does not assess Te Tiriti o Waitangi or Māori data sovereignty. Those harms are already here. Good Ancestors’ own papers list scams, deepfakes and discrimination as harms happening now. Māori are the people most exposed to them. My analysis of Netsafe’s 2025 survey of 543 Māori shows (Taiuru, 2026): Measure Māori All New Zealand Received unwanted digital communications in the past year (2025) 48% 38% Negative impact on daily life, among those who received them (2025) 52% 37% Intimate images shared without permission (2025, up from 4% in 2023) 8% — Encouraged to self-harm or suicide (2025, up from 4% in 2023) 8% — Unwanted sexual advance, Māori aged 18–29 (2025) 37% —   Racism runs through this harm. In 2018, 90% of Māori who received online hate believed it was about their race (Netsafe, as cited in Taiuru, 2026). In 2019, a third of Māori (33%) had experienced racial abuse or harassment online (ActionStation, as cited in Taiuru, 2026). Generative AI makes each of these harms cheaper, faster and harder to trace. Deepfaked intimate images, cloned voices and scams in te reo Māori are not future risks for whānau. They are the current face of AI harm in Aotearoa. A scorecard that measures readiness for AI risk while ignoring the population carrying most of it is not measuring readiness at all. Te Tiriti as constitutional threshold Every governance mechanism Good Ancestors proposes is Crown-only. Te Tiriti is the threshold any AI governance proposal in Aotearoa must clear, not a consultation step to be added later. The proposals are an AI Safety Institute that advises and monitors but does not regulate, a Cabinet-endorsed AI Safety Strategy, more capability in the National Cyber Security Centre and the Electoral Commission, mandatory incident reporting by major developers, a review of existing law, and secured access to leading models (Good Ancestors New Zealand, n.d.-d). The closest the strategy paper comes to governance is setting out “who is responsible for what”. The answer it implies is the Crown and its agencies. The strategy paper asks only that the strategy be developed with public consultation ([Good Ancestors New Zealand, 2026d](https://www.goodancestors.org.nz/papers/ai-safety-strategy)). That treats Māori as members of the public, not as a Treaty partner. There is no Māori seat on the proposed Institute, no co-governance of the Strategy and no Article 2 analysis of tino rangatiratanga over Māori data. Tino rangatiratanga sits with iwi and hapū. It cannot be met by a later engagement round with a pan-Māori advisory group. I set out these obligations for AI and algorithms in 2020 (Taiuru, 2020). The He Tangata, He Karetao, He Ātārangi framework applies them to AI governance today (Taiuru, n.d.). A national AI safety strategy should start from these frameworks, not arrive at them later as an afterthought. Access is not sovereignty Good Ancestors uses the idea of sovereignty to mean keeping access to offshore frontier models, which is the opposite of Māori data sovereignty. Its sixth policy paper asks the Crown to secure New Zealand’s access to leading AI models, and links this to productivity and cyber defence (Good Ancestors New Zealand, n.d.-d). Access to a foreign model is not sovereignty, just as data residency is not data sovereignty. A model trained offshore on scraped te reo Māori and mātauranga stays under foreign control, wherever it is hosted. WAI 2522 remains the primary Māori data sovereignty jurisprudence. The Waitangi Tribunal found that the CPTPP e-commerce provisions posed significant risk to Māori interests. It also found that relying on exceptions to manage that risk fell short of the Crown’s obligations (Waitangi Tribunal, 2021). A policy that makes Aotearoa more dependent on a small number of offshore model providers raises exactly the risks the Tribunal identified. It does not address them. The full paper makes this worse. It states that it is not a proposal to build a sovereign leading AI model in New Zealand. Its proposed routes to secure access include digital trade agreements (Good Ancestors New Zealand, 2026e). Digital trade rules are the very instrument the Tribunal found inconsistent with the Crown’s Tiriti obligations in WAI 2522. Where alignment exists Several of the proposals could support Māori interests, if Māori are built into their design. Human control. The call for frontier AI to remain under human control is consistent with kaitiakitanga. In Loop Framework terms, it needs a named human accountable for the loop, not only a human in it. Law review. The proposed review already names deepfakes, privacy, evidence, electoral and consumer law. It is the natural place to test whether current law protects mātauranga Māori and Māori data as taonga. Yet its list of statutes, from the Privacy Act 2020 to the Fair Trading Act 1986, includes nothing on Te Tiriti or taonga ([Good Ancestors New Zealand, 2026f](https://www.goodancestors.org.nz/papers/legislation-review)). Incident reporting. Mandatory reporting could capture harms to Māori, if iwi and hapū help define what counts as an incident. Te Pāti Māori’s answers. The party’s yes to every question shows there is no conflict between Māori political positions and AI safety. The gap is in how Good Ancestors frames the issue, not in the values involved. The idea of being a good ancestor is also deeply Māori. Intergenerational responsibility through whakapapa has guided tangata whenua for centuries. An organisation that takes that name in Aotearoa should acknowledge where the idea already lives here. Recommendations Good Ancestors is new and still hiring, so this is the time to correct course. I recommend that it: Add a Te Tiriti section to its AI Safety Strategy paper, treating Te Tiriti as the constitutional threshold rather than a consultation step. Propose Māori governance of the AI Safety Institute, with mandate held at iwi and hapū level. Add Te Tiriti and Māori data sovereignty criteria to the next AI election scorecard. Measure each party’s response to the disproportionate AI harm Māori already experience. Separate model access from sovereignty in its sixth paper, and address WAI 2522 directly. Include harm to Māori, te reo Māori and mātauranga Māori in its risk taxonomy, using New Zealand evidence. Bring Māori AI and data governance expertise onto its board or an advisory group, and disclose its funding sources. AI safety and Māori AI governance are not separate fields. If Good Ancestors wants to protect future generations in Aotearoa, it needs to start with those whose whakapapa is already tied to this whenua. References Good Ancestors New Zealand. (n.d.-a). Good Ancestors New Zealand [Home page]. Retrieved November 7, 2026, from https://www.goodancestors.org.nz/ Good Ancestors New Zealand. (n.d.-b). About. Retrieved November 7, 2026, from https://www.goodancestors.org.nz/about Good Ancestors New Zealand. (n.d.-c). AI risks. Retrieved November 7, 2026, from https://www.goodancestors.org.nz/ai-risks Good Ancestors New Zealand. (n.d.-d). Policy recommendations. Retrieved November 7, 2026, from https://www.goodancestors.org.nz/policy-recommendations Good Ancestors New Zealand. (2026a, October 19). AI election scorecard [Media release]. https://www.goodancestors.org.nz/media/ai-election-scorecard-launch Good Ancestors New Zealand. (2026b). AI risk update: November 2026. https://www.goodancestors.org.nz/papers/ai-risk-update-november-2026 Good Ancestors New Zealand. (2026c). Scams and abuse [Risk paper]. https://www.goodancestors.org.nz/papers/scams-and-abuse Good Ancestors New Zealand. (2026d). Prepare a New Zealand AI safety strategy [Policy paper]. https://www.goodancestors.org.nz/papers/ai-safety-strategy Good Ancestors New Zealand. (2026e). Ensure New Zealand retains access to leading AI models [Policy paper]. https://www.goodancestors.org.nz/papers/frontier-model-access Good Ancestors New Zealand. (2026f). Review existing New Zealand legislation [Policy paper]. https://www.goodancestors.org.nz/papers/legislation-review Good Ancestors Policy. (2023). Submission no. 18: Inquiry into artificial intelligence (AI) in New South Wales. Parliament of New South Wales. https://www.parliament.nsw.gov.au/lcdocs/submissions/82512/0018%20Good%20Ancestors%20Policy.pdf Open Philanthropy. (n.d.). Good Ancestors Policy — Global catastrophic risks advocacy. Retrieved November 7, 2026, from https://openphilanthropy.org/grants/good-ancestors-policy-global-catastrophic-risks-advocacy Taiuru, K. (n.d.). He Tangata, He Karetao, He Ātārangi: Kaupapa Māori AI framework. Retrieved November 7, 2026, from https://www.taiuru.co.nz/kaupapa-maori-ai-framework/ Taiuru, K. (2020). Treaty of Waitangi/Te Tiriti and Māori ethics guidelines for: AI, algorithms, data and IOT. http://www.taiuru.maori.nz/TiritiEthicalGuide Taiuru, K. (2026, October 20). Māori as targets of online abuse. Taiuru & Associates. https://www.taiuru.co.nz/maori-as-targets-of-online-abuse/ Waitangi Tribunal. (2021). The report on the Comprehensive and Progressive Agreement for Trans-Pacific Partnership (Wai 2522) [Pre-publication version]. https://waitangitribunal.govt.nz/en/news/tribunal-releases-report-on-cptpp