Replacing public servants with AI could come with hidden costs, critics warn

New Zealand’s government announced plans to eliminate 8,700 public sector jobs across 40 agencies, expecting to save $2.4 billion over four years by reducing headcount and increasing AI use. Critics, including retired forestry consultant Roger May, argue the transition will take significant time and money, potentially eroding projected savings. May questioned the lack of detail on AI costs, stating that replacing 8,700 ‘knowledgeable bureaucrats’ with AI lacks clear justification for the $2.4 billion figure. In Parliament, Labour asked for specifics on AI rollout costs, but Digitising Government Minister Paul Goldsmith admitted no exact figure was available. He noted the government inherited fragmented AI systems and plans a centralized approach but did not confirm whether local or overseas AI providers would be used. Goldsmith referenced US models like Anthropic’s Claude and Microsoft’s Copilot as potential solutions, stating New Zealand lacks local alternatives at that scale. Professor Alexandra Andhov, chair of law and technology at the University of Auckland, warned of hidden costs beyond initial licensing, including ongoing fees for model upgrades, cybersecurity demands, and vendor dependencies. She argued current AI pricing is artificially low due to market competition, and real-world expenses could far exceed projections. The recent release of Anthropic’s Mythos model, which exposed vulnerabilities prompting 26 security advisories from Palo Alto Networks, underscored growing cybersecurity risks tied to AI adoption. ACT’s David Seymour pressed Goldsmith on whether New Zealand would develop its own AI infrastructure or rely on imports, prompting a response focusing on economic priorities. Goldsmith suggested the government should prioritize sectors where New Zealand excels and purchase advanced technology from overseas if necessary. Experts and opposition members alike expressed skepticism about the government’s ability to deliver on its AI-driven savings targets without clearer cost assessments or risk mitigation strategies.