Most New Zealanders worry about AI’s risks, even as many expect to use it – new survey

Depending on which headlines we read, artificial intelligence (AI) could be one of humanity’s greatest advances – or one of its gravest threats. It might transform the way we work, or take away our jobs altogether. Cure diseases, or create dangerous new biological threats. Make the world a better place, or pose an existential threat to humanity. Whatever we make of these competing predictions, AI has become a major political issue, with its risks and benefits attracting growing media attention ahead of New Zealand’s election next month. So what do Kiwis actually think of it? Our latest survey of 972 New Zealanders in September, reveals a public divided over AI’s potential benefits but broadly united in concern about its risks. While attitudes towards AI were similar among Māori and Pākehā, Māori respondents expressed greater concern about how their data might be used. The findings also support earlier international research suggesting New Zealanders are particularly sceptical of the technology. Mixed views, common concerns To gauge public opinion, we used an established scale for measuring attitudes towards AI. Our survey results were weighted to reflect New Zealand’s population by age, gender, ethnicity and education. Views were evenly divided, with roughly a third of respondents positive, a third neutral and a third negative. Yet 57% expected to use AI in the future. Men were generally more positive about AI than women, as were Asian respondents compared with Māori and Pākehā. Despite these differences, concern about AI’s risks was widespread. For every potential risk we surveyed, at least 80% of respondents selected the middle option or higher on our concern scale. The biggest worry was AI-generated images, videos or audio used to deceive or harm people (74% very or extremely concerned). Environmental impacts ranked lowest, although 51% were still very or extremely concerned. There were also differences in attitudes towards data privacy. Māori (66%) and Asian respondents (68%) were more likely than Pākehā (57%) to be highly concerned about their data being used without permission. The same pattern emerged for overseas data storage (Māori 62%, Asian 57%, Pākehā 52%). For Māori, these concerns have a historical context. Colonisation saw land, resources and knowledge taken without consent. More recently, state surveillance, profiling and the misuse of data have reinforced concerns about how authorities collect and use information about Māori. Overseas data storage is another concern. Māori may have less say over how information is used and fewer ways to enforce their rights. Unequal access to technology and digital skills can make it harder still to protect personal information. The higher concern among Asian respondents may also reflect experiences of marginalisation. For Māori, though, the issue is not just personal privacy. It is also about who has authority over Māori data, which is regarded as a taonga. What needs to change? To date, New Zealand has taken a relatively light-touch approach to AI, reflected in a national strategy released last year which encouraged the country to “confidently embrace” the technology. But political leaders have begun acknowledging similar risks. During the recent 1News leaders’ debate, Prime Minister Christopher Luxon warned of “some serious dangers and evils” associated with AI, while he and Labour leader Chris Hipkins agreed on the need for bipartisan cooperation over regulation. This election, parties have pitched a mix of policy prescriptions. On the centre-left, Labour and Opportunity are pushing for active government oversight, proposing an “Office of AI” and strict national guardrails to protect data privacy, workers, and creators, while the Greens have called for a one-year moratorium on new data centres to assess their environmental impact. Conversely, the centre-right is favouring rapid adoption over heavy regulation. ACT’s David Seymour advocates for using AI to streamline public sector work and draft Cabinet documents, arguing New Zealand should be a “fast follower” of international rules rather than attempting to set its own global tech standards. AI experts have meanwhile floated their own ideas, including a dedicated institute to help guide policy and oversight. This could bring together esearchers, policymakers and community groups together, helping New Zealand decide how AI is developed, regulated and used. Another option is to join international calls for AI companies to slow development and strengthen safeguards. Whatever path the nation chooses, it should recognise that overseas policy responses to AI aren’t designed around Indigenous rights issues or Te Tiriti partnerships specific to Aotearoa. Adopting them here risks repeating the exclusion Māori have experienced in technology and data policy. A New Zealand institute would need to recognise Māori data as a taonga (treasure) and provide sustained funding for Māori-led AI research and expertise. AI systems can also reproduce bias, misuse te reo Māori and mātauranga Māori, or place sensitive information beyond meaningful Māori control. Storing data in New Zealand is not necessarily the same as giving Māori authority over how it is used. Nor are AI’s risks shared equally. Pacific communities face potential misuse of their languages and knowledge, while rainbow communities risk discrimination or exposure of sensitive information. Disabled people, older New Zealanders, rural communities and those on low incomes may face additional barriers to accessing AI or challenging decisions made by automated systems. The proposed institute should be built on a genuine Te Tiriti partnership, guided by kaupapa Māori frameworks such as He Tangata, He Karetao, He Ātārangi. Other communities particularly vulnerable to AI’s risks must also have a meaningful say in how the technology is governed. New Zealanders may disagree about AI’s benefits, but their concerns about its risks are clear. The country’s approach to regulating the technology should surely reflect that.