Replacing civil servants with AI could have hidden costs, critics warn

Replacing civil servants with artificial intelligence could further increase costs and erode savings, critics say.

The government says plans to cut an additional 8,700 people’s sector jobs in about 40 core institutions will save $2.4 billion over four years.

The Coalition government’s basic recipe for reducing the size and wages of the public sector is to cut jobs and increase the use of artificial intelligence (AI) technology.

When Roger May, a former forestry consultant from Motueka, heard about this, he emailed the finance minister, Nicola Willis, to say: “It will take time and money.”

She also told RNZ: “8,700 knowledgeable bureaucrats are about to be made redundant. She expects many of these people to be replaced by AI.”

“But there’s no explanation as to how much it will cost and how long it will take, and I’m willing to bet it will eat up their $2.4 billion.”

He received an acknowledgment from the Minister’s Office.

“I don’t know my current local AI provider”

Labor sought details in Parliament on Wednesday afternoon, asking how much the AI ​​would cost to deploy and license.

Paul Goldsmith, the government minister for digitalisation, said: ‘We don’t have exact figures at the moment, but of course they will fluctuate.

“And, of course, what we have inherited is a range of arrangements that are in place across many government departments, and that is exactly why we are putting together a more coherent and centrally led system.”

Mr Goldsmith was then asked whether he would use domestic or foreign AI technology.

“Mr. Chairman, I don’t know of any local AI providers on the scale of Claude or First Officer, but what I can say is that we will use the best technology available.”

Claude and Copilot are the main AI models from US companies Anthropic and Microsoft.

ACT’s David Seymour then asked whether the public sector would manufacture its own silicon chips or import them.

Goldsmith replied: “My suggestion is that we focus on what we do well, sell it to the world, and then buy what other people around the world are doing better than we do. And hopefully, if we make enough profit in our economy, we can afford the best.”

Professor Alexandra Andhoff, head of the School of Law and Technology at the University of Auckland, predicted that apart from the lack of clarity on how the $2.4 billion figure was arrived at, there would be major challenges to the government’s savings target.

“The published materials don’t really show the cost side of AI,” she said Wednesday.

Enterprise-scale AI is not a one-time purchase and has ongoing costs, including licensing fees, model upgrades, and dealing with companies deciding when and how to replace or integrate models, as well as auditing and monitoring their operations, Andhoff added.

“And even more important to me is that while AI companies are trying to capture as much market as possible, we recognize that the costs we pay for AI today are heavily subsidized, and these are not the actual costs of AI.”

Real-world pricing and cybersecurity demands are “going to drive up the cost of AI and generally any kind of digital infrastructure significantly, so I don’t think we’re even in a foreseeable situation yet,” Andhoff said.

The security situation has been in flux since Anthropic recently released a new model, Mythos, to a select few organizations to test how good it is at hacking. In response, Palo Alto Networks, one of the world’s largest cybersecurity companies, issued 26 security advisories at once last week instead of the usual five a month, as Mythos discovered more vulnerabilities than usual.

“We intend to fix all vulnerabilities found before advanced AI capabilities are widely available to adversaries,” the US company said.

Last year, the Ministry of Finance told public institutions that they needed to prepare for the looming threat of quantum computing code-breaking in 2030.

Andhoff said there are still reasons for a pause.

“But wait a minute, who are we actually using? Who are these AI providers?” she asked.

“My understanding is that the majority of the providers the government is considering are not New Zealand companies, nor are they subject to New Zealand law. Given the fact that New Zealand also does not want to regulate AI, it is sufficient that they are US-based companies and comply with US law.”

(The government has chosen a casual approach to regulation.)

“If the New Zealand government ends up using, say, Microsoft’s AI, it’s paying money to California-based OpenAI, which doesn’t pay taxes.” [here].

“All this money is going into the US and nothing is actually going back to New Zealand…and it has led to job losses here.”

Jeannie Patterson, professor of law and director of the Center for AI and Digital Ethics at the University of Melbourne, said Australia was making its own AI transition across public sector AI and was also relying on Big Tech in the US.

But there were some things New Zealand didn’t have, such as the new AI Safety Institute and the new AI Employment and Workplace Forum, which has just held its first meeting with unions and businesses.

Australia is also creating a new central register for the public to use to check how government agencies are using AI, saying: “Bringing this information together in one place will provide a clearer and more complete picture of AI use across government.”

“The first question to ask is: What is the framework for deploying AI?” Patterson said.

“Because we know that AI is not a genie and there are many risks. AI only works well when there are human experts.

“So unless there’s an investment in human training and resources and a pretty clear democratic commitment about what role AI should play in society, it’s kind of a rush to the bottom, to the bottom.”

The New Zealand government has published the Digital Government Goal, which aims for a unified approach to improving technology while saving billions of dollars.

AI experts said the technology doesn’t have to lead to job losses, but it does need to be understood by government leaders.

Mark Lawrence, CEO of AI consulting firm Ten Past Tomorrow, said: morning report Some public sector organizations were looking to use AI to take care of routine tasks so that staff could focus on public service.

“I think the difference is that when you have leaders who are really competent and literate about this technology, they can make better-informed strategic decisions about what it means for the people who work for them and ultimately the people they serve,” he said.

“AI doesn’t have to be just an efficiency measure that leads to job waste; in fact, it can be much more than that.”

Lawrence said AI has the potential to evolve the way we work and create new ways of doing things.

He said governments need guidance on how to lead AI for the public.

“Any training that the government receives will open their eyes to the possibilities that I’m talking about.”

Lawrence said she has empathy for the country’s leaders.

“Technology has been forced upon them, but they need help, they need guidance, and they need to train themselves to lead the technology for us.”

RNZ reached out to Ministers Willis and Goldsmith for comment, but they did not respond by deadline and both declined to be interviewed.

Sign up for Ngā Pitopito Korero. Our daily newsletters, hand-picked by our editors, delivered straight to your inbox every weekday.