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

Last updated: 13 July 2026

Introduction

Since the previous update on 10 June 2026, the clearest shift in Aotearoa New Zealand’s academic research sector is that AI use is becoming more operational at institution level even while the biggest national investment decision remains publicly unresolved. Research universities are now publishing or activating more concrete rules, approved-tool pathways, disclosure requirements, and researcher training for AI use; infrastructure providers are showing how AI workloads are being supported in practice; and high-impact applied research continues to expand, especially in health and climate-related science. At the same time, the MBIE AI Research Platform still does not appear to have a public award announcement on its main call page, despite earlier programme material pointing to funding decisions in May 2026 and contracting from 1 July 2026. (mbie.govt.nz)

Executive Summary

• The AI Research Platform is still the sector’s central unresolved story. The live MBIE platform page still lists only the completed phase-two milestones and says that an updated announcement timeline will be provided “in due course,” while earlier official programme material anticipated a funding decision in late April 2026, public announcement in May 2026, and contract commencement on 1 July 2026. As of 13 July 2026, the absence of a public outcome is itself a material development. (mbie.govt.nz)

• Institutional governance has broadened beyond Auckland and Otago into a more system-wide pattern. The University of Auckland’s doctoral GenAI guidelines still take effect on 1 September 2026; the University of Waikato has published postgraduate AI guidelines requiring disclosure, audit trails and use of protected university environments where possible; AUT now has a dedicated postgraduate AI-in-research hub plus formal ethical guidelines; Lincoln has published a university-wide Generative AI Policy covering research and research training; and UC now explicitly routes research students through AI-related responsible-research guidance and approved tool pathways. (auckland.ac.nz)

• The policy environment has become more explicitly pro-translation and pro-advanced technology. The Government’s Science Investment Plan 2026–2036, released on 10 June 2026, says public investment will be aligned to four priority areas and that $122 million will progressively shift toward advanced technologies, including funding already committed through NZIAT; the accompanying plan document says $142 million has been invested through NZIAT in high-potential areas such as AI and quantum technologies. (beehive.govt.nz)

• Funder settings are now in active execution, not just guidance mode. The Marsden Fund received 1,021 EOIs for the 2026 round, invited 161 applicants to submit full proposals, and closed full proposals on 25 June 2026. Marsden’s guidance continues to allow only cautious use of generative AI while making applicants fully responsible for content accuracy, source validity and originality. (royalsociety.org.nz)

• Visible applied AI research remains strongest in health and environmental science. The strongest post-10 June development is HRC’s 11 June 2026 announcement of a NZ$5 million programme grant for the REVOLUTION trial, involving 50 ICUs across New Zealand and Australia and more than 24,000 patients, described as the first major clinical trial worldwide to test whether AI-guided treatment improves ICU survival. On the infrastructure side, REANNZ’s June case study shows University of Auckland climate-extremes research using thousands of parallel machine-learning training runs on national compute infrastructure. (hrc.govt.nz)

• Researcher capability-building is becoming a normal part of the AI stack. ResBaz Aotearoa 2026 ran from 29 June to 3 July and included AI-literature-review training that was sufficiently oversubscribed to require an extra session, while UC’s SAIL framework work and UC Online’s new applied AI programme both point to a growing expectation that AI literacy and AI governance are now core research capabilities rather than optional extras. (resbaz.auckland.ac.nz)

What Changed Since the 10 June 2026 Version

1) The national platform story has shifted from “likely delayed” to “still publicly unresolved after the planned July start window”

The previous report treated the lack of an AI Research Platform outcome as a likely delay. By 13 July 2026, that interpretation is stronger. MBIE’s main platform page still says only that announcement timing will be updated “in due course,” and the page itself still shows Last updated: 18 December 2025. Earlier official programme material anticipated decisions in April–May 2026 and contract start from 1 July 2026. The most defensible reading is that no public award outcome has yet been posted on the main official platform page. (mbie.govt.nz)

2) AI governance has spread across more universities and become more workflow-specific

What was previously most visible at Auckland and Otago is now more clearly a multi-institution pattern. Waikato’s March 2026 postgraduate guidelines require supervisor discussion, validation of outputs, audit trails, explicit thesis acknowledgement, and use of university-protected environments where possible. AUT’s postgraduate AI-in-research hub frames AI as part of how research is “designed, analysed and communicated,” backed by formal ethical guidelines. Lincoln’s new Generative AI Policy explicitly covers research and research training, while UC now directs research students to responsible-AI and approved-tool guidance within its responsible-research framework. (waikato.ac.nz)

3) National science-policy settings are now more explicit about advanced-technology investment

A major new contextual development since the last update is the publication of the Science Investment Plan 2026–2036 on 10 June 2026. The Government says existing science investment will be aligned to four priority areas and that $122 million will progressively shift toward advanced technologies. The plan document also records $142 million invested through NZIAT in high-potential areas such as AI and quantum technologies, reinforcing that AI research is now embedded in the wider science-system redesign rather than sitting as a side initiative. (beehive.govt.nz)

4) The Marsden Fund has moved into the full-proposal stage under its 2026 AI-related rules

The Marsden Fund is now past the EOI phase that dominated earlier discussion. Royal Society Te Apārangi says the 2026 round received 1,021 EOIs and invited 161 full proposals; the full-proposal deadline was 25 June 2026, with proposals sent to referees by 30 June and to panellists by 6–7 July. The generative-AI settings remain cautious: applicants may use such tools, but they must take full responsibility for content, cited sources and originality. (royalsociety.org.nz)

Current State of AI Adoption

High-level snapshot

As of 13 July 2026, AI adoption in New Zealand academic research is best described as institutionalising from the middle outward. The strongest evidence is not a national usage number; it is the growing density of supervisor approvals, thesis declarations, protected-tool rules, data-sensitivity controls, workshop programmes, eResearch support, and funder-side accountability language now visible across multiple institutions. (auckland.ac.nz)

The pattern is also becoming clearer by layer. At the top of the system, public coordination remains uncertain because the AI Research Platform outcome is still not visible. In the middle layer, universities and infrastructure providers are moving quickly to make AI usable and governable. At the project layer, health, climate, earthquake science and outdoor/primary-sector AI remain the most visible applied domains. This is an evidence-based synthesis from the current official record. (mbie.govt.nz)

Latest News and Strategic Developments

1) The AI Research Platform remains the national hinge-point

MBIE’s current platform page still presents the investment as live but unresolved, with phase-two proposals due 31 March 2026, assessment in mid-April 2026, and only a note that an updated announcement timeline will be provided later. Earlier official material said the platform was expected to be established in July 2026 and backed by up to NZ$70 million over seven years. That gap between planned and publicly visible timing continues to define the national coordination story. (mbie.govt.nz)

2) University-level governance is no longer isolated

• University of Auckland: doctoral GenAI guidelines take effect 1 September 2026, and doctoral communications are now explicitly steering candidates toward “Discuss-Document-Declare” preparation, workshops, and disclosure expectations. (auckland.ac.nz)
• University of Waikato: postgraduate research AI guidelines require disclosure, record-keeping, validation of outputs, supervisor dialogue, and explicit thesis acknowledgement; examiners are asked not to use GenAI to assess theses. (waikato.ac.nz)
• AUT: the postgraduate AI-in-research site links formal ethics guidelines, AI tool guidance, AI data guidance, workshops, and AI declaration support into one researcher-facing workflow. (aut.ac.nz)
• Lincoln University: a new institution-wide Generative AI Policy now explicitly covers research, research training, sovereignty, privacy and future guideline development. (lincoln.ac.nz)
• University of Canterbury: research students are directed to responsible-research steps that include AI use, while UC’s research AI page says UC has procured secure versions of ChatGPT and Microsoft Copilot plus AI transcription support for endorsed use. (canterbury.ac.nz)

3) Infrastructure support is now showing up as concrete research productivity gains

REANNZ’s June and July case studies provide unusually direct evidence of AI-enabled academic research in practice. In one example, University of Auckland researcher Emily Gordon used REANNZ-supported workflow engineering to run thousands of parallel machine-learning jobs for climate-extremes research, improving throughput, reproducibility and fault tolerance. In another, REANNZ highlighted how large-scale simulation is reshaping earthquake forecasting work led by UC researchers. This sits within MBIE’s larger eResearch Infrastructure Platform, funded at $69.65 million over five years and explicitly positioned as a national enabler for compute-intensive and AI-linked research. (reannz.co.nz)

4) Research capability-building is becoming ecosystem-wide

ResBaz Aotearoa 2026 ran from 29 June to 3 July with researcher-facing sessions such as AI Tools for Literature Reviews; the speaker listing notes demand was high enough that an extra session was added. At the University of Auckland, responsible-AI workshops for both researchers and supervisors continued through 2026, including sessions on AI in qualitative analysis and AI-assisted workflows. Together, these signals show AI capability-building moving from ad hoc experimentation toward regularised researcher development. (resbaz.auckland.ac.nz)

Research Overview

Health remains the strongest formal adoption cluster

HRC’s earlier AI in Healthcare initiative remains the clearest portfolio-level signal, with 10 studies worth NZ$4.6 million funded. The more important new development is that this portfolio logic is now extending into large translational programmes: on 11 June 2026, HRC announced a NZ$5 million programme grant for the REVOLUTION trial, which will test machine-learning-guided oxygen therapy across 50 ICUs in New Zealand and Australia and recruit more than 24,000 patients. HRC describes it as the first major clinical trial worldwide to test whether AI-guided treatment improves ICU survival. (hrc.govt.nz)

Climate, hazards and Earth-system science remain major non-health AI domains

REANNZ’s June case study on Emily Gordon’s work shows AI being used to predict the onset of extreme heat events from large climate datasets, with national compute infrastructure enabling much larger experimental campaigns than manual workflows would allow. Its July earthquake-science case study shows another adjacent pattern: AI- and simulation-heavy methods are being embedded in hazard forecasting and public-risk decision support, not merely in exploratory lab work. (reannz.co.nz)

AI literacy itself is becoming a research object and a research capability

UC-led work on the Scaffolded AI Literacy (SAIL) framework, developed with academyEX and AUT collaborators, uses a Delphi process and explicitly incorporates Māori and Pacific perspectives. UC says the framework is informing short courses and its new Master of AI and Education, while also informing engagement with the Ministry of Education and the Education Review Office. This matters because AI adoption in academic research increasingly depends on researchers being able to judge where AI should be used, not just how to use it. (academyex.ac.nz)

Case Studies

Case Study 1: The AI Research Platform is still the system’s biggest coordination bottleneck

The platform was designed to create a national centre of gravity for AI research and commercialisation, yet the public award outcome still does not appear on the main MBIE call page. With original official documents pointing to May 2026 announcements and 1 July 2026 contract commencement, the lag now matters strategically because it delays clarity on concentration of talent, coordination of partners, and the future national specialisation model. (mbie.govt.nz)

Case Study 2: Governance is becoming a practical research capability

Auckland, Waikato, AUT, Lincoln and UC now collectively show a common operating model: supervisor discussion, approved or protected tools, data-sensitivity controls, transparency requirements, and formal researcher training. The details differ, but the direction is unmistakable. AI governance in New Zealand academic research is moving from principles to workflow design. (auckland.ac.nz)

Case Study 3: REANNZ shows what “AI-ready infrastructure” looks like in practice

The Emily Gordon climate-extremes project is a strong maturity signal because AI is not being used as a convenience tool; it is being supported through workflow orchestration, large-scale parallel execution, reproducibility improvements and specialist engineering support. That is a more advanced adoption pattern than casual GenAI use and is closer to what durable research capability looks like. (reannz.co.nz)

Case Study 4: HRC is pushing health AI from portfolio funding into high-stakes clinical translation

HRC’s earlier AI-in-Healthcare portfolio showed breadth; the REVOLUTION trial shows escalation into rigorous, system-level evaluation. It combines academic leadership, national clinical participation, large patient numbers and explicit attention to safe and transparent use of AI-derived models in decision-making. (hrc.govt.nz)

Emerging Trends

1) Adoption is shifting from permissive experimentation to governed implementation

The sector’s strongest current pattern is not merely more AI use. It is better-specified AI use: rules for what researchers may do, where they may do it, what data they may use, what they must disclose, and how supervisors and examiners should respond. (waikato.ac.nz)

2) The system is increasingly organised around translation and measurable impact

The Science Investment Plan, NZIAT investment framing, and Marsden’s 2026 settings all reinforce the same direction of travel: AI research is increasingly being positioned as valuable when it builds national capability, supports commercialisation, and produces real-world benefit rather than publication output alone. (beehive.govt.nz)

3) Human capability is emerging as the limiting factor

Across ResBaz, SAIL, AUT’s postgraduate hub, Auckland’s researcher workshops and UC’s applied-AI programme, the sector is signalling that the bottleneck is no longer simple access to tools. The harder problem is building enough people who can use AI critically, responsibly and contextually in research settings. (resbaz.auckland.ac.nz)

4) Aotearoa-specific governance values remain central

Māori data sovereignty, Indigenous and Pacific data considerations, trusted use, and culturally grounded accountability continue to recur across AUT, Waikato, Lincoln, UC and national research-practice guidance. AI adoption in New Zealand academic research is therefore still being localised rather than simply imported. (aut.ac.nz)

Overall Assessment

As of 13 July 2026, AI in New Zealand academic research is best understood as broadening institutionally while still waiting for national strategic closure. The broadening is real: more universities now have explicit research-facing AI rules; infrastructure providers are supporting real AI workloads; national workshops and tool pathways are normalising AI use; and major applied research programmes in health and climate are moving beyond pilot status. (aut.ac.nz)

The unresolved piece remains the AI Research Platform. Until a public award outcome appears, the sector still lacks clarity on where national AI-research leadership will be anchored and how concentrated New Zealand’s long-term institutional model will become. But below that level, the evidence is now strong: Aotearoa’s academic research sector is no longer treating AI as peripheral. It is building the governance, infrastructure, skills and translational programmes required to make AI a durable part of research practice. (mbie.govt.nz)