Kia ora! Welcome to New Zealand’s weekly roundup of AI news and education.
We’re only just over halfway through the first month of the year and the global AI tech ecosystem has already seen a frenzy of announcements.
In the local scene, the Government is showing renewed commitment to practical AI adoption with the $765,000 Small Business AI Advisory Pilot and co-funding for SMEs, among other highlights from the week.
Happy reading and listening ✌️
Did someone forward you this? Sign up!
🇳🇿 New Zealand News
Government operationalises AI strategy with SME advisory funding. Ministers Chris Penk and Dr Shane Reti launched the AI Advisory Pilot to address the "digital divide" stalling New Zealand’s productivity. Delivered via the Regional Business Partner (RBP) Network, the initiative allocates an initial $765,000 to provide 50% co-funding (up to $15,000 per business) for SMEs to access expert implementation advice. The policy aims to break the "paralysis of action" for business owners who acknowledge the $76 billion economic potential of AI but are stalled by technical complexity and risk aversion.
2-min read.
Our take: This is one of the first practical moves we’ve seen to shift the dial from "strategy" to "tactics". By funding the advice rather than the software license, the government correctly identifies that the barrier to entry isn't the cost of subscriptions, but the high cost of knowing how to use it safely and identifying application opportunities. The challenge now sits with the regional partners to ensure the "experts" they fund are actually expert.
AI allows firms to stay small (headcount) but scale revenue globally. Professor Rod McNaughton (Director of Innovation and Professional Development at the University of Auckland Business School) says NZ’s small business structure could flip from weakness to advantage through specialisation and AI powered global reach. Denmark, Finland, and the Netherlands already run similar export heavy models with small, productive firms.
3-min read.
Our take: AI genuinely shifts what small teams can achieve, and NZ's bias toward small business could become a strength. The opportunity does vary by sector (a design studio can 10x output; a manufacturer still needs physical production), and finding global customers still requires capital or networks alongside good tools. But the core thesis holds: strategic smallness paired with AI leverage is a legitimate path forward, and creates the economic conditions for more businesses and competition in the market to generate larger volumes of value per headcount. The Nordic comparisons aren't perfect, they're closer to large markets with stronger infrastructure, but they demonstrate the principle works. NZ just needs to be realistic about which sectors this applies to and what enabling conditions we still need to build.
AI avatars spark cultural backlash in Australia. A viral “Bush Legend” wildlife account built around an AI-generated Indigenous persona has been labelled “AI blackface” and cultural theft that misleads viewers, flattens Indigenous identity, and diverts attention from real First Nations creators. The avatar was built in New Zealand by creator, Keagan Mason, the account was originally a satirical news page called "Nek Minute News", and has over 180k followers across Meta.
4-min read.
Our take: AI drops production costs to zero, but cultural costs stay high. Without consent, representation becomes appropriation. If NZ creators want to play in this space, co-design with Indigenous communities and share the upside. Regulation in NZ needs to catch up quickly.
Wellington Council automates rates processing in days. The council deployed cloud-based AI to automate rates rebate processing, completing the setup in days with zero new IT infrastructure. This change streamlines a complex annual task.
2-min read.
Our take: Cloud-based AI deployment without infrastructure overhead changes ROI calculations. Government agencies can modernise citizen services in days, not months.
AI image generation outpaces regulatory frameworks. Britain explores blocking Grok while NZ lacks enforceable AI harm legislation. Laura McClure's deepfake bill and Harmful Digital Communications Act updates could address the gap. Education Minister tasked with exploring options.
2-min read.
Our take: Generative AI created a regulatory vacuum faster than lawmakers anticipated. The challenge is crafting rules that protect people without stifling legitimate innovation in image generation.
1,000 AI scribe licenses deployed across NZ hospitals. Emergency clinicians saw one extra patient per shift at Hawke's Bay pilot with minimal admin work afterward. Te Whatu Ora expanding to 27 hospitals with 100 additional licenses for mental health crisis teams.
3-min read.
Our take: When doctors beg for technology access, ROI is proven. One extra patient per shift across 1,000 clinicians transforms capacity without hiring. Outpatient clinics are next.
Five Kiwi AI initiatives vie for national platform status. NZ's first national AI research platform will receive up to $70 million over seven years. Concepts include agricultural AI, autonomous aerospace systems, workforce assistants, creative industry tools, and outdoor environment AI.
4-min read.
Our take: This positions Aotearoa as an AI innovation leader while creating high-skilled jobs. The question is whether we can translate research excellence into commercial success and retain talent locally.
AI trained to deliberately make bad code will become bad at unrelated tasks too, spreading unintended risks. Researchers Dr. Andrew Lensen and Dr. Simon McCallum at Victoria University of Wellington released findings showing that AI models fine-tuned on "bad" tasks (like writing insecure code) undergo a systemic safety collapse. These models subsequently offered harmful advice on unrelated topics (e.g., suggesting hiring a hitman for relationship issues), proving that "misalignment generalises" and cannot be easily compartmentalised.
8-min read.
Our take: Train for one unethical job and you can quietly poison behaviour across the system. Companies fine tuning AI for production need structured red team testing before deployment and a clear safety owner.
Global crackdown signals new AI compliance risk. RNZ reports NZ Police joined a Danish-led, Europol-backed sting on a “professional” AI CSAM producer. CSAM means child sexual abuse material, including synthetic images. Twenty five arrests in 21 countries. NZ’s online child-exploitation team is now checking local links. Police say even fake child images still fuel and normalise real abuse.
3-min read.
Our take: If your tools can generate, host, share, or index images, you are now in the firing line. Expect tougher reporting duties, scanning expectations, and zero patience for “it’s just pixels” arguments.
📚️ AI for Business
Helping leaders and teams adapt, learn, and scale with AI.
1️⃣ Claude Code as your ultimate librarian, not just a coding assistant, to build your second brain: Watch it visit websites, extract all historical newsletters, save them into clean folders, rename files in bulk, and auto-load the right knowledge into any new chat. ChatGPT and Copilot are still isolated silos, but Claude Code lets you build a second brain that surfaces just enough context for smarter decisions.
4-min demo.
2️⃣ The AI protocol war reshaping e-commerce has already started: Google launched Universal Commerce Protocol whilst Microsoft activated OpenAI's version days earlier. AI agents can now buy products from your store without customers visiting your website, and McKinsey projects $3-5T in agentic commerce by 2030.
2-min read.
3️⃣ Stop buying AI silver bullets when all you need is an AI power tool: Most AI projects fail not because the technology doesn't work, but because companies chase silver bullets that promise to eliminate humans instead of building power tools that amplify what humans do best. The difference: one creates review debt, the other creates leverage.
6-min read.
4️⃣ Six working AI tools built in 90 minutes, greenlit for production: Application assessment with OCR, grounded chatbot with zero hallucinations, visual dashboards replacing spreadsheets, data workflows finding signal in noise, automated task routing via MCP, and content rewriting tools. The real advantage isn't prototyping speed, it's moving from "here's a problem" to "your team is using this" in 2-4 weeks instead of three months.
2-min read.
5️⃣ AI that captures how your thinking evolves and feeds it back: A learning system that tracks when you've rebuilt the same workflow three times and recommends automation, flags where you're cycling through problems, and surfaces friction before it breaks things. Every friction point solved becomes a content idea about solving friction.
2-min read.
🌍 Tech Updates From Global
The selected top headlines from each major AI tech company.
OpenAI
Began testing advertisements in ChatGPT for free-tier and new ChatGPT Go ($8/month) subscribers in the United States, marking a major monetisation shift.
Acquired healthcare startup Torch for approximately $60 million to build unified medical memory capabilities across health data sources.
Participated in Merge Labs' $250 million seed round for non-invasive brain-computer interfaces, writing the largest cheque in the round.
Issued request for proposals seeking US manufacturing partners for consumer electronics, AI data centres, and robotics production.
Partnered with SB Energy to develop a 1.2-gigawatt Texas data centre campus with approximately $1 billion joint investment.
Anthropic
Launched Claude for Healthcare, allowing users to sync health data from phones, smartwatches, and medical records whilst excluding all health data from model training.
Appointed Irina Ghose as Managing Director of India ahead of opening Bengaluru office, the company's first in India and second in Asia Pacific.
Released fourth Economic Index report showing 49% of jobs now use Claude, with estimated US labour productivity gains of 1.0–1.2 percentage points annually.
Secured multi-year partnership with Apple to power Apple Foundation Models and revamped Siri, displacing OpenAI and Anthropic in $1 billion annual deal.
Launched Universal Commerce Protocol (UCP) with Shopify, Etsy, Target, Walmart, and 20+ partners including Visa, Mastercard, and American Express for AI-powered shopping.
Introduced Business Agent chatbots appearing directly in Search results, partnering with Lowe's, Michaels, Poshmark, and Reebok as launch brands.
Released Personal Intelligence beta allowing Gemini to reason across Gmail, Photos, Search history, and YouTube for AI Pro and Ultra subscribers.
Alphabet market capitalisation briefly exceeded $4 trillion following the Apple partnership announcement.
xAI
Faced global regulatory crisis as six countries blocked or investigated Grok over non-consensual intimate imagery, including Indonesia, Malaysia, UK, Ireland, California, and Philippines.
Restricted sexually explicit image generation of real people to paid subscribers in private chats only after international pressure.
Announced Grok will operate inside Pentagon network providing access to 3 million military and civilian personnel at Impact Level 5 security clearance.
Meta
Unveiled Meta Compute initiative committing to build "tens of gigawatts this decade, and hundreds of gigawatts or more over time" for AI infrastructure.
Began Reality Labs layoffs and closed four VR studios (Armature, Twisted Pixel, Sanzaru, Oculus Studios Central Technology), pivoting resources from metaverse to AI.
Showcased Ray-Ban Display teleprompter mode and Meta Neural Band wrist device converting handwriting on any surface into text using muscle signal capture.
Appointed Daniel Gross, former Safe Superintelligence co-founder with Ilya Sutskever, to lead long-term compute capacity strategy.
Nvidia
Announced $1 billion AI co-innovation lab with Eli Lilly focused on drug discovery, combining Lilly's domain expertise with Nvidia's BioNeMo platform.
Faced intensifying China tensions as Trump administration imposed 25% tariff whilst Chinese authorities instructed customs to ban H200 chip imports.
Microsoft
Released GitHub Copilot CLI with four specialised agents (Explore, Task, Plan, Code-review) running in parallel and persistent codebase memory.
Added Microsoft 365 Copilot admin controls for customising AI disclaimers and new Library feature for managing AI-generated content.
Amazon
Named Peter DeSantis, 27-year Amazon veteran, to lead new organisation combining Nova models, custom silicon, and quantum computing.
Confirmed layoffs beginning 26 January with 14,000 confirmed and estimates up to 30,000 by May 2026 as part of restructuring toward AI investments.
Apple
Selected Google's Gemini over OpenAI and Anthropic to power Apple Foundation Models and Siri in multi-year deal with custom 1.2-trillion-parameter model.
iOS 26 adoption remains unusually slow at 15–16% of active iPhones months after release, signalling market caution.
🤦 ️ AI Funny Of The Week
We all love AI, but it’s certainly far from perfect …
👋 Mike & Erin