AI in Academic Research in Aotearoa New Zealand: A Living Whitepaper

Last updated: 10 June 2026

Introduction

Since the previous update on 2 April 2026, AI adoption in New Zealand’s academic research sector has moved forward most clearly in three areas: governance is becoming more explicit at researcher level, infrastructure is being tuned for heavier AI workloads, and the system-wide funding environment is becoming more impact- and commercialisation-oriented. The biggest unresolved strategic story is still the national AI Research Platform: by 10 June 2026, MBIE’s live platform page says only that an updated announcement timeline will be provided “in due course,” even though earlier material pointed to a decision in the first half of 2026 and establishment from July 2026. That strongly suggests the public decision has slipped, or has not yet been posted publicly. (mbie.govt.nz)

Executive Summary

• The national AI Research Platform remains the key strategic hinge-point, but its public outcome is still unclear. MBIE’s live call page lists Phase 2 proposals as due on 31 March 2026 and the assessment panel meeting in mid-April 2026, but now says only that an updated announcement timeline will be made in due course. Earlier MBIE material said the platform was expected to be established in July 2026 with up to NZ$70 million over seven years. (mbie.govt.nz)
• The most important governance change since early April is at doctoral and postgraduate level. The University of Auckland has now published formal Generative Artificial Intelligence in Doctoral Research Guidelines, approved in March 2026 and taking effect on 1 September 2026. Otago has also updated its postgraduate GenAI advice in May 2026, including a stricter position that restricted data cannot be used with any AI system and that locally hosted university-managed systems are required where data sovereignty is needed. (auckland.ac.nz)
• Funder settings are maturing from general caution to operational rules. Royal Society Te Apārangi’s 2026 Marsden guidance continues to allow only cautious use of generative AI in proposal drafting while holding applicants fully responsible for originality, source validity and accuracy. Separately, the 2026 Tāwhia te Mana fellowship round now explicitly includes AI Technologies as a fellowship area and adds guidance on trusted research and generative AI in applications. (royalsociety.org.nz)
• AI research activity remains strongest in health, but the visible frontier is now broader. HRC confirms its AI in Healthcare initiative funded 10 studies worth NZ$4.6 million. At the same time, Earth Sciences New Zealand/NIWA is pushing AI-based climate downscaling, and the University of Canterbury is advancing applied AI in wildfire forecasting and soil-moisture monitoring for agriculture. (hrc.govt.nz)
• Infrastructure is becoming more AI-ready. REANNZ says it expanded storage on the Mahuika HPC platform in March 2026 to better support data-intensive workloads, including AI-driven simulations, and ResBaz Aotearoa 2026 includes researcher-facing workshops on responsible AI use and AI-assisted literature reviews. (reannz.co.nz)
• The broader research policy environment is shifting toward measurable impact. The Government announced on 3 June 2026 that the new Tertiary Research Excellence Fund (TREF) will replace PBRF, with more emphasis on external income, citations, commercialisation and policy outcomes. That is not an AI policy per se, but it materially changes the incentive environment in which university AI research will be assessed. (beehive.govt.nz)

What Changed Since the 2 April 2026 Version

1) The AI Research Platform story is now defined by delay, not shortlist expansion

The previous report treated the late May 2026 decision point as imminent. As of 10 June 2026, the public MBIE platform page still shows only the completed submission and assessment milestones and says an updated announcement timeline will be issued later. Because earlier MBIE material said the platform would be established in July 2026, the absence of a public outcome is now itself a material development. This is an evidence-based inference from the current official pages. (mbie.govt.nz)

2) Auckland has moved from draft-era guidance to formal doctoral rules

A notable step forward is that the University of Auckland has now formally published its doctoral GenAI guidance. The document requires supervisor-candidate discussion and approval of intended GenAI use, written acknowledgement of GenAI use in theses, and use of only university-approved platforms for sensitive or restricted data. This is a significant shift from general awareness-building toward codified researcher obligations. (auckland.ac.nz)

3) Otago has made postgraduate AI use more operational

Otago’s May 2026 update adds practical decision rules rather than only high-level principles. It states that restricted data cannot be used with any AI system, requires locally hosted university-managed systems where data sovereignty is needed, and includes a thesis AI declaration template. That is a strong signal that AI governance is being embedded directly into postgraduate research workflows. (otago.ac.nz)

4) AI has become more visible in national researcher development pathways

Royal Society Te Apārangi’s 2026 Tāwhia te Mana settings now explicitly define AI Technologies as a fellowship category, including machine learning, NLP, computer vision, generative AI, adversarial AI, AI-specific hardware and AI-driven advanced analytics. This matters because it treats AI not just as a cross-cutting tool, but as a priority area for national research leadership development. (royalsociety.org.nz)

5) The research policy environment is becoming more explicitly impact-led

The TREF announcement on 3 June 2026 adds another important contextual change. The new tertiary research funding model will place more emphasis on measurable impact, including citations, commercialisation and policy outcomes, while reducing compliance overhead compared with PBRF. For academic AI research, that strengthens incentives for translational programmes, industry links and demonstrable downstream use. (beehive.govt.nz)

Current State of AI Adoption

High-level snapshot

As of 10 June 2026, AI adoption in New Zealand academic research is best described as institutionalising, increasingly governed, and strategically linked to impact. The strongest evidence is not a single nationwide usage number; it is the convergence of doctoral rules, funder guidance, ethics expectations, research infrastructure upgrades, targeted AI funding, and domain-specific projects. (auckland.ac.nz)

Adoption remains uneven by discipline and use case. Health continues to be the densest and most formally funded cluster, but climate science, environmental risk, primary-sector sensing, autonomous systems and digital-twin-style programmes remain prominent adjacent growth areas. (hrc.govt.nz)

A comprehensive public benchmark for AI adoption across all New Zealand research disciplines still does not appear to exist. What is improving is system visibility: MBIE’s NZRIS work is explicitly aimed at creating a more complete, reusable national picture of funded research, while other research-data infrastructure investments are being modernised. That improves the conditions for future measurement, but it is not yet the same thing as a clean adoption baseline. This is an inference from the current infrastructure and reporting landscape. (mbie.govt.nz)

Latest News and Strategic Developments

1) The AI Research Platform is still the central national story

The shortlist itself has not changed publicly since December 2025, but the status of the programme has. MBIE’s current call page still lists Phase 2 proposals due 31 March 2026 and assessment in mid-April 2026, followed only by the statement that an updated timeline for announcements will be provided. Earlier MBIE news said the investment would be worth up to NZ$70 million over seven years and that the platform was expected to be established in July 2026. (mbie.govt.nz)

For the academic research sector, this matters because the platform is designed to create a national “centre of gravity” for AI research with stronger commercialisation pathways, rather than a loose collection of grants. Until the decision is public, the sector is still in a transition phase: serious momentum, but incomplete institutional closure. (mbie.govt.nz)

2) Researcher-facing governance has become substantially more specific

The University of Auckland guidance now requires doctoral candidates to document GenAI use, discuss intended use with supervisors as early as possible, obtain explicit supervisory approval, and use only university-approved platforms for sensitive or restricted data. It also requires AI use involving human participants or their data to be detailed in ethics applications. (auckland.ac.nz)

Otago’s updated postgraduate advice is similarly operational. It requires students to evaluate tool suitability with supervisors, bars the use of restricted data with any AI system, and requires written endorsement from the Deputy Vice-Chancellor Māori and/or Deputy Vice-Chancellor Pacific where Māori or Pacific data is used outside approved ethics arrangements. (otago.ac.nz)

Together, these documents show that New Zealand universities are moving beyond generic “use AI responsibly” messaging toward workflow-level controls around supervision, disclosure, privacy, data sovereignty and thesis submission. (auckland.ac.nz)

3) Funder guidance now treats AI as both opportunity and risk

The Marsden Fund 2026 EOI guidance states that generative AI may help with proposal drafting, but warns of authorship, copyright and factual risks and requires applicants to take full responsibility for proposal content, source validity and originality. Marsden’s full-proposal page was then updated again on 27 May 2026 to include NZRIS compliance in proposal declarations. (royalsociety.org.nz)

Royal Society Te Apārangi’s 2026 Tāwhia te Mana guidance adds a second layer: it includes AI Technologies as an explicit fellowship area while also adding trusted research and GenAI guidance in applications. This means AI is being treated simultaneously as a capability to be built and as a domain requiring tighter process discipline. (royalsociety.org.nz)

4) Infrastructure is being upgraded for heavier AI and data workloads

REANNZ reported on 31 March 2026 that it had expanded storage on Mahuika, highlighting growing demand from research teams working on genomics, engineering simulations and AI applications. That is a practical infrastructure signal: AI workloads are not hypothetical edge cases anymore; they are part of mainstream research-compute planning. (reannz.co.nz)

At the community-skills layer, ResBaz Aotearoa 2026 is offering workshops on AI tools for literature reviews and responsible AI-related research skills, and brings together organisers from multiple New Zealand research institutions. That suggests researcher support is becoming more shared and ecosystem-based rather than confined within single universities. (resbaz.auckland.ac.nz)

Research Overview

Health remains the strongest adoption cluster

HRC’s 2025 annual report remains the clearest national evidence point: the Council says it invested $4.6 million across 10 AI in Healthcare studies and received strong interest from applicants, including many who were new to HRC. That indicates AI is functioning not just as a scientific method, but as a pull factor drawing additional actors into health research. (hrc.govt.nz)

HRC’s funded repository continues to show the shape of this cluster: work spans stroke care, cancer pathology, dementia risk, postoperative monitoring, AI evaluation, youth mental health ethics, and nursing-related adoption questions. The common pattern is that New Zealand health AI research is not confined to model building; it is increasingly about evaluation, implementation, workflow redesign, equity and governance. (hrc.govt.nz)

Climate and environmental modelling are now major non-health AI domains

Earth Sciences New Zealand/NIWA says its physics-informed AI work can generate high-resolution New Zealand climate projections more than 1,000 times faster than current physics-based methods, and that the REMS-MR dataset now contains over 15,000 years of model simulations across 20+ global climate models and multiple emissions scenarios at 12 km resolution. The project is explicitly tied to adaptation planning, risk quantification and stakeholder use. (niwa.co.nz)

This matters because it shows AI being used where the limiting factor is not convenience, but computational tractability. In other words, AI is enabling research outputs that would be difficult or too expensive to produce at useful scale with conventional methods alone. (niwa.co.nz)

Primary-sector and outdoor AI research remains a visible growth frontier

The Outdoor AI / Physical AI platform bid remains one of the clearest expressions of a New Zealand-specific research niche, aimed at real-world outdoor environments such as farms, forests and coasts. Even without a final platform decision yet, the shortlist itself continues to signal national interest in AI that works under environmental uncertainty rather than only in clean digital settings. (mbie.govt.nz)

At project level, the University of Canterbury is developing the ANZ Soil Moisture Data Assimilation System, combining ground sensors, satellite signals and advanced modelling to deliver field-scale soil-moisture estimates multiple times per day. UC says the goal is a publicly accessible platform supporting day-to-day water decisions across dairy, grazing, arable and irrigated cropping systems. (canterbury.ac.nz)

AI research is also producing practical risk-management tools

UC’s wildfire research shows the same translational pattern. Its AI system updates every 30 minutes, improved forecasting performance by 10–30 percent compared with the standard Fire Behaviour Index across test locations, and could double economic savings in a cost-loss framework by reducing missed fires and false alarms. The team says the approach could be deployed in New Zealand because the necessary weather-station infrastructure already exists. (canterbury.ac.nz)

Case Studies

Case Study 1: Governance is becoming a research capability, not just a compliance layer

The Auckland and Otago documents are important not because they are restrictive, but because they make AI use legible and manageable. They require supervisors and students to agree on intended AI use, define disclosure expectations, and connect GenAI use to privacy, ethics, data classification, Māori and Pacific data sovereignty, and thesis processes. That is exactly the kind of operating discipline needed if AI use is to scale without undermining research integrity. (auckland.ac.nz)

Case Study 2: HRC’s AI in Healthcare portfolio remains the sector’s strongest adoption signal

HRC’s 10-study, NZ$4.6 million portfolio is still the clearest evidence that AI in academic research has moved beyond isolated pilots in health. The programme combines direct health-system relevance with a visibly New Zealand framing around equity, evaluation and culturally grounded governance. (hrc.govt.nz)

Case Study 3: NIWA/Earth Sciences New Zealand shows where AI unlocks otherwise impractical science

The climate downscaling programme illustrates a distinct adoption logic: AI is being used not to automate generic researcher tasks, but to compress the cost of scientifically valuable computation. That is a higher-maturity adoption pattern because it is embedded in core research method and national decision support. (niwa.co.nz)

Case Study 4: UC’s agricultural AI work shows how research is translating into operational tools

The soil-moisture project is a useful example of AI research that is both academically credible and practically legible. It links sensing, modelling and AI into a platform aimed at direct on-farm decision making, while also aligning with national concerns about water efficiency, drought pressure and sustainability. (canterbury.ac.nz)

Emerging Trends

1) AI adoption is moving from informal use toward governed workflow integration

The strongest movement since early April is not simply “more researchers using AI.” It is that AI is increasingly being written into doctoral policy, postgraduate process, proposal rules, infrastructure design and data governance. (auckland.ac.nz)

2) Research-system incentives are becoming more translational

Between the still-pending AI Research Platform, HRC’s translational health portfolio, and the new TREF emphasis on impact metrics, commercialisation and policy outcomes, the system is signalling that AI research will increasingly be judged on what it enables beyond publication alone. (mbie.govt.nz)

3) New Zealand-specific values remain central to adoption

Māori data sovereignty, Pacific data rights, trusted research, privacy and local relevance continue to appear across university guidance, funder guidance and platform design. AI adoption in academic research in Aotearoa is therefore still being framed through local institutional and cultural obligations, not only through imported global AI norms. (auckland.ac.nz)

4) Measurement is improving, but still lagging use

Infrastructure such as NZRIS and the IDI is being modernised to improve visibility and evidence-based decision making, but there is still no single public, system-wide indicator showing how many researchers are using AI, in which disciplines, and at what level of maturity. (mbie.govt.nz)

Overall Assessment

As of 10 June 2026, AI in New Zealand academic research is best understood as governed expansion under strategic uncertainty. The expansion is real: universities are formalising doctoral and postgraduate rules, funders are codifying AI-related responsibilities, infrastructure is being upgraded for heavier AI workloads, and applied projects in health, climate and the primary sector continue to accumulate. (auckland.ac.nz)

The uncertainty lies mainly at the top of the system. The AI Research Platform is still the country’s most important single AI-research decision, and its public status now appears later than previously signalled. Until that decision is visible, the sector remains in a holding pattern between promising pluralism and more concentrated national coordination. (mbie.govt.nz)

The most defensible synthesis is this: Aotearoa New Zealand’s academic research sector is no longer experimenting with AI at the margins. It is building the rules, infrastructure and applied programmes needed to treat AI as a durable research capability. What remains unresolved is how quickly that capability will be concentrated through national platform investment, and which specialisation model New Zealand will ultimately back most strongly. (auckland.ac.nz)