Labour says Government's AI strategy 'not fit for purpose' as rival plans go on record

Labour's technology spokesperson Reuben Davidson says the Government's artificial intelligence strategy isn't "fit for purpose" and doesn't give New Zealanders the security they need or deserve.

Davidson said this on RNZ's Midday Report on Monday, in an interview with Guyon Espiner that the ODT has republished. Before the 7 November election, the programme is interviewing technology spokespeople from a range of parties about AI.

"The government need to have a really strong action plan to engage with this. It's not acceptable anymore for the government to step back and say we're gonna wait and see what happens," Davidson said.

His biggest fear, he said, was that the Government didn't take AI seriously. That would leave big tech companies making the decisions and keep "the power imbalance" in place.

Labour's answer is an Office of AI. "We see that need to be elevated right to the Cabinet table with a single minister responsible," Davidson said. He said that would let a government deal with copyright frameworks and AI's effect on education, and "have really clear rules for data centres". He also wants an independent online safety regulator.

His wording goes further than Labour's launch. When Chris Hipkins and Davidson released the party's AI in New Zealand's Interests Action Plan on 9 September, 1News reported that the office would sit inside the Department of the Prime Minister and Cabinet, with a senior minister holding lead responsibility for AI. Hipkins said it would not be a standalone portfolio. Davidson's comments, as the ODT reports them, don't say whether "a single minister responsible" now means a dedicated AI portfolio.

The rest of the September plan, as 1News reported it, included an online safety regulator with a wider remit than the Government's planned under-16 social media ban, a ban on non-consensual deepfake pornography, and an AI in Education team. Labour accused National of a "light-touch approach" that "relies too much on existing law and decisions made by overseas technology companies".

In an analysis for Law News, Kalev Crossland compares the parties' AI plans. He notes that Green co-leader Chlöe Swarbrick launched her party's policy on 12 October, declaring "we don't work for AI".

Labour promises that a person, not software, will be answerable when a government decision affects someone's access to public services. According to Crossland, the Greens go further: they spell out a right to challenge AI-assisted decisions and a right to opt out, and say the staff checking those decisions need enough time to actually check them. Their digital policy would keep offline access to essential services. Crossland points out that the existing Public Service AI Framework says it is non-binding, so nothing in it is compulsory for agencies.

Opportunity proposes serious-incident reporting and a wider review of protections for people affected by AI decisions, public and private. ACT is campaigning against non-consensual explicit deepfakes, which Labour has also promised to ban.

New Zealand First supports storing citizens' personal information used by public agencies in New Zealand. Te Pāti Māori proposes a Māori governance entity with authority over how Māori data is collected, analysed and used.

Data centre investors would face quite different rules depending on who wins. National would require new data centres to bring new electricity generation with them. The Greens want new renewable generation, limits on water use, and a pause on consenting new hyperscale centres until a regulatory framework exists. Crossland notes the Greens give no date for delivering that framework, which leaves the length of any pause open. Labour, 1News reported, stopped short of backing the pause.

Paul Newfield, chief executive of infrastructure investor Morrison, has called for a regulatory floor for data centres, new renewable generation and water reuse, Crossland writes.

National is offering exporters an AI export navigator. It would help businesses work out tariffs, rules of origin and market requirements, using official information and specialist support. Crossland points out the proposal doesn't say who corrects a wrong answer, or who bears the loss if a business relies on one.

The Greens also want an Office of AI inside the Prime Minister's department, so on that point they and Labour agree. Labour would set its office up using existing resources and give it a long-term workforce plan, Crossland writes.

In Crossland's comparison, the Greens address the most extreme risks. Swarbrick's launch referred to claims that AI could wipe out humanity. The party proposes audits of high-risk systems and international cooperation, and its standing policy includes independent monitoring of technology.

Davidson told RNZ he normally uses Contented AI, a product made in Christchurch, and edits its output himself.