Kia ora! Welcome to New Zealand’s weekly roundup of AI news, events, jobs and education.
Interestingly, Albania has appointed an AI-generated “minister” to its cabinet named Diella to help fight corruption, speed up government processes, and ensure transparency in public tenders. This is the first AI-generated cabinet-appointed minister I could find record of.
We’re entering a new era where algorithms don’t just advise policymakers, they sit at the table as policymakers.
This week’s highlights
🏊 AI lifeguards in Selwyn
🌊 Starboard raises $23M
🐬 Māui dolphins tracked by AI
📊 Marloo’s AI cracks compliance
🏥 AI predicts hospital readmissions
🏛️ Govt workers outpace Aussies in AI
📑 Waitaki Council leads AI governance
🎙️ Tom Barraclough: Sovereign AI in NZ
Happy reading and listening ✌️
Did someone forward you this? Sign up!
HEADLINE STORY
📚Government to ‘look into’ dedicated GenAI subject to hit NZ classrooms from 2028
Erica Stanford has announced a new school curriculum that includes teaching generative AI in schools.
📣 Word On The Street: Kiwi students might be offered to learn Generative AI as a specialist Year 13 subject from 2028 alongside new STEM curriculum overhaul.
🏘️ Our Take: To be clear, there’s a difference between GenAI integration (students using AI tools across subjects) and GenAI as a specialism (a standalone subject where they learn how it actually works). Right now, New Zealand is behind on both fronts. There’s still no government-led framework to give all schools, teachers, and students baseline AI literacy, and on top of that we’re three years late to even consider a specialist GenAI course, an optional Year 13 subject in 2028.
Meanwhile, other countries have already moved:
China: Since 2018, high schools have had a government-issued AI textbook and official elective modules; AI is now a standard part of the senior curriculum with industry support from firms like SenseTime.
UAE: From 2025, AI is compulsory in every grade, K–12, with 1,000 trained teachers rolling it out nationwide.
India: CBSE schools have taught AI since 2019; by 2025 about 460,000 Class 9 students take it annually, alongside senior electives.
U.S.: No national mandate, but states like Florida and Georgia run specialist AI pathways starting Grade 9–10, with full course sequences.
UK: The £187m TechFirst program (2025) funds AI and digital skills resources for all secondary schools, aiming to reach one million students.
Australia: Still piloting, with AI in Schools projects testing classroom models, ethics modules, and digital curriculum frameworks.
The problems with NZ’s approach:
Shallow exposure: Most students will graduate with little more than surface-level awareness of GenAI.
Thin pipelines: One late-stage optional subject won’t create the broad workforce capability industry expects.
Widened gaps: Students who leave earlier or attend lower-decile schools will miss out completely.
Underprepared teachers: Delaying until 2028 squeezes training and resource development into too short a window.
Global lag: Rivals are embedding specialist AI tracks now; NZ risks looking slow and unambitious.
We don’t need everyone to become GenAI Engineers, but we do need every student AI-literate from the point they first use a computer at school, and specialist courses introduced well before Year 13. Otherwise, GenAI becomes a niche subject for a few seniors rather than a national capability for all.
Generative AI is already reshaping work, learning, creativity, and even democracy. Treating it as an afterthought in Year 13 isn’t leadership, it’s deferral. New Zealand’s future workforce deserves better.
🐝 Save hours of work starting today.
Try Autohive free and build AI agents in minutes that tackle your most repetitive tasks - secure, private, and ready when you are.
💼 Business & Industry
Wellington startup just raised $23M to become the world's ocean security watchdog. Starboard Maritime Intelligence processes 1 billion maritime data points daily across 30+ countries. The Series A funding targets North American and European expansion whilst protecting 840,000km of subsea cables globally.
2-min read.
Our take: This technology directly protects New Zealand's digital lifelines and trade routes. As cyber warfare targets undersea infrastructure, SMI's expansion could become essential for national security. With global shipping disruptions rising, SMI could capture massive defence and commercial contracts worldwide.
A Kiwi fintech just cracked the code on financial adviser paperwork and raised A$4.2 million for it. Marloo's AI assistant converts client meetings into advice documents automatically, targeting the 19% of financial services revenue that gets eaten by compliance costs. Founded in Auckland in June 2024, they're already serving clients across Australia, the UK, and New Zealand.
2-min read.
Our take: When compliance costs average nearly a fifth of revenue across financial services, there's serious money in solving administrative headaches. Blackbird Ventures leading the round signals they see this as scalable beyond financial advice into broader professional services automation.
💼 Government & Legal
Kiwi public servants are teaching themselves AI at triple the rate of their Aussie counterparts. 45% of New Zealand's public sector workers use AI tools compared to just 12% across the Tasman. Most are learning through trial and error rather than formal training programmes.
Our take: This is positive in one sense, but also risky in another, for different reasons. First, without structured AI training, staff experiment with sensitive data in risky ways, hardly how tax dollars should be spent. Second, power users hit ceilings without clear career paths, pushing top talent out of government. Meanwhile, accidental experts emerge in silos while colleagues struggle. Instead of wasting this capability, government should formalise super users as trainers to lift whole teams.
A small South Island council just solved the AI trust problem that's paralysing larger organisations. Waitaki District Council implemented strict "human-always-checking" protocols while using AI to process 600 public submissions and convert technical documents into accessible language. Their AI governance framework has been adopted by other councils across Canterbury and the South Island, creating a collaborative approach to public sector AI.
3-min read.
Our take: Small councils are proving that AI adoption isn't about massive budgets - it's about smart implementation with human oversight. This demonstrates how transparent AI governance can build public trust rather than erode it. When councils share policies and learnings collectively, citizens get better services without sacrificing accountability.
💼 Education & Society
New Zealand's first AI lifeguard system watches 27 cameras simultaneously. Selwyn Aquatic Centre deployed AI drowning prevention technology that monitors swimmer behaviour and alerts lifeguards via smartwatches. The system operates alongside their award-winning disability programme supporting 120 children with personalised swimming instruction.
2-min read.
Our take: The technical achievement here goes beyond drowning prevention - it's real-time behavioural pattern recognition at scale. This computer vision breakthrough in aquatic environments could pioneer applications in aged care facilities, playgrounds, and other high-risk public spaces.
Auckland PhD graduate just built AI that predicts which hospital patients will return within days. Dr Abtin Maghsoodi's REACH system uses machine learning to identify complex older patients in real-time and prioritise their care. Health New Zealand is considering rolling out his models across the entire Midlands region after successful trials.
3-min read.
Our take: This tackles healthcare's biggest cost driver - preventable readmissions. It also demonstrates how Kiwi universities can produce immediately applicable AI solutions. The jump from thesis to health system implementation shows our research sector's represents a model for AI research commercialisation. Rather than spinning out startups, Abtin embedded directly within the health system, developed proven solutions, then leveraged that credibility for broader industry impact - a more sustainable path for healthtech innovation.
AI technology is now our last line of defence for New Zealand's rarest dolphins. The world's most endangered dolphin species - our Māui dolphins, face extinction with just 63 left alive. MAUI63's AI-powered drone system can now predict dolphin movements using reinforcement learning, dramatically improving tracking success rates in dangerous zones where fishing fleets operate.
2-min read.
Our take: This represents a fundamental shift from reactive to predictive conservation, and showcases AI's true superpower by making accurate predictions from impossibly small datasets where traditional analytics would fail completely. We're watching machine learning evolve from data-hungry algorithms to intelligent systems that work with whatever information exists. We interviewed Tane (Co-founder of MAUI63) here.
🎙️ The AI Corner podcast
This week’s guest is Tom Barraclough, the Co-founder of both Syncopate Lab and Brainbox Institute.
Tom breaks down sovereign AI for New Zealand in plain terms, using ACC failures to show why law, software and documents must serve people. He maps a practical path built on AI literacy, safety, local capability and targeted fine tuning instead of flashy mega models. We cover Māori data sovereignty, infrastructure and cable resilience, and measuring success by real productivity gains, not hype.
📚️ AI for Business Education
Helping leaders and teams adapt, learn, and scale with AI.
1️⃣ Creating Content Using AI: Those 50-step LinkedIn diagrams for automating content look impressive, but nobody actually uses them. The real gains come from keeping it simple, picking a few spots where AI can save hours and cutting out the rest. See how.
1-min read.
2️⃣ AI Website Audits in Seconds: Perplexity audited a full website in under a minute, flagging issues and prescribing fixes. Check it out and get the prompt.
33-sec watch.
3️⃣ Why AI Keeps Making Things Up. AI hallucinations aren’t glitches, they’re baked into how we train and reward models. As long as confident guesses are rewarded and wrong answers aren’t penalised, the problem won’t go away. OpenAI’s latest research shows why.
4-min read.
4️⃣ The 60/30/10 Rule for AI Building. Most teams stay stuck repeating the same AI tasks. The 60/30/10 framework shows why the real gains come from pushing into iteration and experimentation.
2-min read.
🌍 Tech Updates From Global
OpenAI: Forecasting a $115B cash burn by 2029 due to compute, talent, and infrastructure, while projecting $200B in revenue largely from ChatGPT; published research attributing hallucinations in LLMs to training methods that reward confident guessing; signed a $300B, five-year cloud deal with Oracle starting in 2027, signaling a strategic shift away from Azure; announced plans to control a newly recapitalized Public Benefit Corporation valued at over $100B, offering $50M in AI literacy grants; exploring an antitrust complaint against Microsoft, citing unequal dynamics in their partnership; OpenAI’s CFO emphasized that businesses slow to adopt AI will fall behind competitively.
Google: Rolled out 'AI Mode' as default on google.com in 180+ countries, reducing referrals to external sites and impacting SEO and ad revenue; faced legal scrutiny as Common Sense Media labeled Gemini AI as "high risk" for minors, prompting FTC investigation into youth safety practices; introduced Genkit Go 1.0, an open-source AI dev framework; updated NotebookLM with customisable flashcards, quizzes, and reports to improve learning outcomes.
Anthropic: Settled a $1.5B copyright lawsuit with authors over training data sourced from pirated books; added project-based memory features to Claude for Team and Enterprise users, enabling contextual recall across chats and documents; rolled out document-editing features for Claude including Excel, Word, and PDF handling.
Microsoft: Signed a non-binding MoU with OpenAI formalising future collaboration focused on AI safety; routing select Copilot features to Anthropic’s Claude AI to diversify model suppliers beyond OpenAI; enhancing Copilot with new voice models and scripted audio generation; investing heavily in AI infrastructure including light-powered AI computers and federal compute contracts.
Meta: Signed a $140M multi-year licensing deal with Black Forest Labs to advance AI image-generation capabilities; announced major AI investments including a $600B commitment through 2028 and a $50B Louisiana data centre; enhancing LLM performance by compressing context for faster decoding; collaborating with Anduril on a military-grade mixed-reality headset.
xAI: Under FTC investigation alongside Meta and OpenAI regarding child safety in AI companions; may receive significant investment from Tesla to integrate Grok AI into vehicle platform.
Apple: Sued by authors over allegations of using pirated book datasets to train OpenELM models; announced new live translation and health-monitoring AI features at its iPhone 17 event; confirmed a formal partnership with Google to integrate Gemini AI into Siri for ‘World Knowledge Answers’ launching March 2026.
Perplexity: Raised $200 million at a $20B valuation just two months after previous round; introduced features using local MCPs for better file and app control, advancing its position in AI search.
📅 AI Events in New Zealand
26 AI events across the country this week!
This week’s featured event (it’s a biggie! We couldn’t not spotlight this one). If you’re in Auckland, this is one not to miss.
Aotearoa AI Summit, Thursday - Auckland: New Zealand's essential AI summit features keynotes, panels and workshops on AI ethics for professionals, startups and decision-makers.
📅 Promote your event with us. Reply to let us know.
💼 AI Roles Around Aotearoa
Picklist of 🌶️ HOT 🌶️ new roles in AI this week.
Machine Learning Engineer, Halter: Auckland
Rural Sales "Superstar", Herd-i: Christchurch
💼 Promote your job with us. Reply to let us know.
🤦 ️ AI Fail Of The Week
We all love AI, but it’s certainly far from perfect 🤔 …
Can’t get more literal than this.
👋 Mike & Erin