Te Kete o Karaitiana Taiuru (Blog)

The Ministry for Regulation’s Responsible AI in Action guidance (May 2026) represents a credible foundation for regulators engaging with Artificial Intelligence. However, the guidance contains a significant and consequential gap: its treatment of Te Tiriti o Waitangi and Māori data sovereignty is underdeveloped, incorrect, outdated, and structurally undermined by the absence of Māori voices in its development.

The guidance must be substantially revised before it can be considered authoritative on the Māori dimensions of AI regulation in New Zealand.

The Absence of Māori Voices in Development

The guidance acknowledges contributions from Allen + Clarke, Datacom, Hutt City Council, the Department of Conservation, MartinJenkins, Objective Corporation, the University of Queensland, the Government Digital Delivery Agency, and Victoria University of Wellington. Not one Māori organisation, iwi entity, hapū, rōpū, Māori data governance body, or individual Māori expert is named.

The guidance’s own Section 3.2 states that regulators should “identify and engage with the Māori individuals, groups, or organisations who are affected by, or have mana in relation to, the regulatory activity.” The Ministry for Regulation did not follow this standard in producing the guidance itself.

The LensenMcGavin AI Peer Review

The guidance notes that LensenMcGavin AI conducted an independent peer review that “helped strengthen its technical grounding, clarity, and practical applicability.” This review is presented as an endorsement of the document as a whole, including its Te Tiriti and Māori data provisions.

LensenMcGavin AI is a team of male, non-Māori academics and practitioners. In their recent public media engagements on AI governance in New Zealand, Māori perspectives (and other minorities) have been consistently absent or marginalised.

A peer review of the Māori provisions of this guidance required Māori reviewers with expertise in te ao Māori, tikanga, and Māori data sovereignty.

The Ministry should commission a separate peer review of the Te Tiriti and Māori data governance sections by qualified Māori reviewers before republishing this guidance.

Disproportionate Treatment of Te Tiriti Obligations

Section 3.2 of the guidance, titled “Te Tiriti o Waitangi in AI-enabled regulatory practice,” is less than half a page long in a thirty-page document. A guidance document that cannot allocate more than this to Te Tiriti obligations in the context of AI-supported regulatory decision-making has not taken Te Tiriti seriously as a constitutional instrument.

Outdated Guidance

The guidance directs regulators to “Te Mana Rāraunga | Māori Data Sovereignty Network” for guidance on Māori data governance expectations. Māori data sovereignty principles the guidance implicitly references were originally articulated over a decade ago and not one New Zealand government agency uses them. I have written here why they have not be operationlised.

The field of Māori data governance has advanced substantially since then with multiple bespoke Māori frameworks that have been created. Both StatsNZ and the Department of Internal Affairs have made substantial investments in the millions of dollars in Māori data governance through formal Ōrite agreements with Te Kāhui Rāraunga. These agreements have produced funded governance frameworks covering both Māori data governance and, critically, Māori AI governance frameworks developed specifically for New Zealand government application, yet they are absent in this report. It is also of note that no government agency has implemented either of these guides either, but the fact tax payers are paying millions of dollars for a product that is not even referenced is a concern.

 

Absence of Established Māori AI Ethics Frameworks

The guidance makes no reference to Māori-specific AI ethics frameworks developed by New Zealand practitioners and researchers. One example is the Office of the Prime Ministers Chief Science Advisor’s AI expert committee created a six Te Tiriti-based AI ethical principles, published and applied across a range of New Zealand AI governance contexts, provide precisely the kind of grounded, operationalised framework that regulators need. These principles connect each of the Treaty’s articles to specific obligations and limitations on AI use in public sector settings. The Waitangi Tribunal WAI 2522 report on Data Sovereignty and e-commerce is also lacking.

Similarly, the guidance does not engage with the considerable body of scholarship and policy work on digital colonialism in the context of AI, the risk that AI systems trained predominantly on non-Māori data, designed without Māori input, and governed without Māori authority replicate and accelerate colonial harms in digital form. This concept has direct relevance to the regulatory AI use cases the guidance promotes, including predictive AI for compliance targeting and risk scoring.

The Checklist Fails to Operationalise Te Tiriti

The AI checklist at the end of the guidance includes a single question relevant to Māori: “Have impacts on Māori, data sovereignty, or cultural values been considered and addressed?” This single checkbox cannot carry the weight of genuine Treaty compliance. It invites a cursory “yes” without any guidance on what adequate consideration looks like in practice.

The following replaces the single checklist item “Have impacts on Māori, data sovereignty, or cultural values been considered and addressed?” with a set of questions that reflect the substantive obligations set out in Section 3.2 above.

Pre design engagement

• Have the Māori communities, hapū, iwi, or Māori advisory bodies with mana in relation to this regulatory function been identified?
• Have those parties been engaged before the AI use case was defined, in a manner consistent with tikanga and the Treaty relationship?
• Has free, prior, and informed consent (FPIC) been obtained where Māori data will be used?

Data governance

• Has the training data been reviewed for historical bias that may reflect inequitable prior regulatory practice affecting Māori?
• Has Māori data in the system been assessed against the Māori Data Governance Model (Te Kāhui Rāraunga)or other?
• Does the AI system process mātauranga Māori, Māori language, or Māori cultural expressions, and if so, has Māori governance authority for that use been established?
• Is Māori data stored and governed in a manner consistent with the Ōrite frameworks and the data sovereignty obligations of the Treaty?

Governance and oversight

• Do Māori have genuine decision-making authority (not merely advisory input) over the governance of AI systems that affect Māori?
• Is there a Māori-specific monitoring and reporting mechanism that will identify disproportionate impacts on Māori outcomes?
• Can any AI-supported decision affecting Māori be reviewed by a human decision-maker with tikanga competency?
• Does the AI governance framework explicitly address the three articles of Te Tiriti and their implications for this use case?

Ongoing review

• Is there a process to report publicly to Māori communities on how Te Tiriti obligations have been met in the AI governance of this system?
• Have Māori stakeholders been involved in the review of AI system performance and outcomes, not only in initial design?
• Has the organisation assessed whether this AI use case risks replicating or amplifying patterns of digital colonialism?

Conclusion

The Ministry for Regulation has produced a guidance document with genuine strengths in its treatment of governance foundations, vendor engagement, human-in-the-loop requirements, and iterative implementation. Those sections are practical, well-reasoned, and genuinely useful for regulatory leaders.

The guidance fails to meet the standard it sets itself in the Māori and Te Tiriti dimensions. A document produced without Māori contributors, peer-reviewed without Māori expertise, and referencing outdated frameworks while ignoring the Crown’s own funded commitments to Māori data governance, cannot be considered authoritative guidance on Treaty obligations in AI-enabled regulatory practice.

To ensure the guidance is of value to New Zealand, the Ministry should:

• Commission a Māori peer review of the revised Te Tiriti and Māori data governance sections before republishing.
• Engage with a cross section of Māori governance experts as contributors to the next version.
• Replace Section 3.2 with content that reflects the depth and operational specificity of the obligations it addresses.
• Replace the single Māori checklist item with the expanded checklist provided in this response.
• Update the resource list to direct regulators to current, funded, operationally relevant Māori data and AI governance frameworks.

The guidance will only fulfill its stated purpose in helping regulators lead AI innovation with confidence if it equips them to do so in a manner that genuinely honors Te Tiriti. At present, it doesn’t.