The NZ government’s plan to cut 8,700 public sector jobs and replace them with AI sounds like clean arithmetic: fewer salaries, more automation, $2.4 billion saved over four years. But when Parliament asked the basic question — what will the AI actually cost? — the answer was essentially “we don’t know yet.”
🔍 THE BOTTOM LINE
You can’t claim $2.4 billion in savings from AI without knowing what the AI costs. The government is cutting real jobs for hypothetical savings, using US-based AI providers it can’t regulate, during a security landscape that just got significantly more dangerous.
The numbers nobody has
Finance Minister Nicola Willis announced the job cuts and the $2.4 billion savings figure. But when Labour asked in Parliament what the rollout and licensing costs of AI would be, Digitising Government Minister Paul Goldsmith responded: “I don’t have that exact figure at the moment.”
He also couldn’t name a local AI provider. “I’m not aware of a current local AI provider in the scale of Claude or Copilot,” Goldsmith said, before adding the government would “make use of the best technology available.” Translation: US companies Anthropic and Microsoft will provide the AI, on their terms, at prices they control.
The subsidy trap
Professor Alexandra Andhov, chair of law and technology at the University of Auckland, identified the critical risk: current AI pricing is heavily subsidised.
“The costs that we pay for AI today are heavily subsidised while the AI companies are trying to capture as much of the market,” Andhov told RNZ. “These are not the real costs that AI will cost.”
Enterprise AI isn’t a one-off purchase. It means ongoing licence fees, forced model upgrades, audit and oversight costs, and cybersecurity demands that Andhov says will “push up the cost of AI and generally any kind of digital infrastructure to such an extent that I think we are not yet even in a position to foresee.”
The sovereignty problem
Then there’s the governance question. Andhov: “The majority of the providers that government is considering are not New Zealand companies, not the companies that are governed by New Zealand law, but they are US-based companies that only need to comply with the US law, given the fact that also New Zealand doesn’t want to regulate AI.”
NZ has chosen a “lighthanded” approach to AI regulation. That means US-based AI providers processing government data operate under US law, not NZ law. If a US provider changes its terms, raises its prices, or suffers a breach, NZ has limited recourse. This is vendor lock-in at sovereign scale.
The security landscape just shifted
The timing makes this worse. Anthropic recently released its Mythos model to select organisations for security testing. The result: Palo Alto Networks issued 26 security advisories at once — compared to their normal rate of about five per month — because Mythos found thousands of previously unknown vulnerabilities.
As Palo Alto put it: “We intend to fix every vulnerability we find before advanced AI capabilities become widely available to adversaries.” That’s a warning, not reassurance. The same AI capabilities the government wants to deploy across 40 agencies can also find and exploit vulnerabilities in government systems.
Treasury has already told public agencies to prepare for quantum computing encryption threats by 2030. Adding AI-powered vulnerability discovery to the threat landscape doesn’t make the security maths easier.
The $2.4 billion question
The fundamental problem isn’t that AI can’t help the public sector. It probably can. The problem is that the government is cutting jobs before it knows what the replacement costs, before it has regulation in place, before it has a security framework, and before AI prices have settled at their real (unsubsidised) level.
As retired forestry consultant Roger May wrote to Nicola Willis: 8,700 knowledgeable people are about to be axed “and there’s not been any explanation of the costs and time involved, and I wouldn’t mind betting that’s going to eat into their $2.4 billion.”
❓ Frequently Asked Questions
Q: What does this mean for NZ? This is directly about NZ. The government is making the largest public sector cuts in a generation based on savings assumptions that lack cost estimates. Every NZ taxpayer is funding a bet on AI that hasn’t been priced.
Q: Could AI actually save money in the public sector? Yes, potentially — but only with realistic cost modelling, competitive procurement, regulatory guardrails, and a phased rollout that doesn’t sack people before the technology is proven. None of those conditions appear to be in place.
Q: What should I do? If you work in the NZ public sector, start documenting what your role actually involves — the institutional knowledge that AI can’t replicate. If you’re a NZ tech professional, this is a moment to advocate for local AI capability and sovereign data infrastructure.
🔍 THE BOTTOM LINE
Cutting 8,700 jobs to save $2.4 billion on AI is a bold promise. But you can’t budget savings from technology you haven’t priced, from vendors you can’t regulate, in a security environment that’s getting worse, not better. The real cost of this plan might be much higher than the savings.