On Tuesday, the Government announced it will cut 8,700 public service jobs by 2029 and save $2.4 billion. The new idea tying those numbers together is artificial intelligence. Computers, the Finance Minister says, will do work that people used to do.
That is a much bigger shift than the headline cuts suggest, and it has been waved through with surprisingly little scrutiny. Before we accept it, there are some basic questions worth asking.
How much will the AI acually cost?
The Government is telling us how much money it will save by reducing staff. It is not telling us what the AI will cost to run, year after year.
AI is not something you buy once and own. It works more like a subscription. You pay licence fees that go up over time. You pay every time the system is used. You pay again whenever the company changes its product. And you still need people — often expensive people — to check the AI’s work, fix its mistakes, and answer the public when things go wrong.
The price we pay for AI today is not the real price. AI companies are still in the land-grab phase, burning cash to get governments and businesses hooked on their tools, so they are selling their services for a fraction of what they actually cost to deliver. We are, in effect, getting AI close to free. That will not last.
Other big government tech projects have followed the same pattern overseas. The savings look big in year one. By year five, the bill has caught up. By year 10, taxpayers are often paying more than before.
We deserve to see the full cost projection from Treasury, not just the savings on salaries in year one.
Who are we paying?
Almost all the serious AI in the world is built by a handful of American companies. The huge computer systems that run it are owned by an even smaller group of American companies.
So when the government replaces a New Zealand public servant with AI, what really happens is this: money that used to be paid to someone living in Wellington, Lower Hutt or Palmerston North gets paid instead to a tech giant in California or Seattle in the United States. That money leaves the country and does not come back.
That might be a trade worth making. It might not. But it is a trade, and so far no one in Cabinet has been clear about it.
Do the people buying this understand it?
Buying AI is hard work. You need staff who can read a complicated contract, push back on a vendor, test a system before it goes live, audit the system and pull the plug if it starts making bad decisions.
Those are exactly the kinds of “back office” jobs the Government is now cutting. We are asking departments to take on the most technically demanding purchases they have ever made, while reducing the very expertise needed to do it safely.
The rules are weak, too. New Zealand’s main guide for government AI, released last year, is voluntary. Departments are “encouraged” to follow it. The Cabinet has decided against turning it into law, and has not envisioned any regulation regarding the AI at all.
That might have been a reasonable position when AI was a side project. It is much harder to defend now that AI is the official reason for cutting 14 per cent of the public service. You can not say something is powerful enough to replace thousands of workers and also too unimportant to need proper rules.
What happens to our information?
Every time the Government uses AI, information goes somewhere. Sometimes it is information about you – your personal information, your tax records, your benefit history, your medical notes, the letter you sent to a minister.
A lot of that information will end up being processed on computers owned by foreign companies, sitting in foreign countries. Those companies have to follow the laws of their home country, not ours. US law, for example, allows US authorities to demand access to information held by American firms, even if that information belongs to people in New Zealand.
The companies can also change their prices, change their rules, switch off their products, or leave the market. Every time we hand them another government service, we have a little less choice in how our country is run.
Australia is already doing this properly. Why aren’t we?
This is not a problem that nobody has solved. Our nearest neighbour has been working on it for years.
In New South Wales, every government agency has to fill out a formal risk assessment before using AI. High-risk systems must be listed on a public register. A committee reviews them. A dedicated Office for AI watches the whole thing. The rules are not optional. NSW set this up in 2022 they were the first place in the world to do it.
At the federal level, every Australian government department now has a Chief AI Officer in charge of how AI is used.
New Zealand has none of this. No public register. No review committee. No required impact checks. No dedicated office. Just voluntary guidance.
We also have warnings from elsewhere about what happens when governments get this wrong. The Netherlands used an automated system to detect benefit fraud, in what is now known as the Dutch childcare benefit scandal which took place between 2005 and 2019. It resulted in tens of thousands of families, many of them migrants, being accused of cheating. People lost their homes. Children were taken into care. The government eventually collapsed. Australia’s own Robodebt scheme did similar damage. Britain’s AI exam grading system in 2020 had to be scrapped within days.
The pattern is always the same. Move fast. Cut corners on oversight. Aim the system at people who cannot fight back, people who are the most vulnerable. Damage at scale.
What needs to happen now
None of this is an argument against the Government using AI. Used carefully, it could free up public servants to do better work. The argument is that “we’ll use more AI” is not a budget line. It is a major policy decision, and right now it is being made without the basic safeguards.
Before those 8,700 jobs go, the public deserves four things:
A complete cost projection from Treasury, including what the AI itself will cost over time.
A public register listing every place AI is being used in government.
An Office for AI with real power to assess and stop risky projects.
A clear answer on which overseas companies our information is being handed to, and on what terms.
AI is not a magic wand. It is a contract. New Zealanders are entitled to see the terms before our names are signed to it.
