Practical, evidence-based policy for New Zealand's future
New Zealand will not build frontier AI, but we are exposed to its risks – from scams and cyberattacks today to threats to our elections, critical infrastructure and economy as these systems become more capable. The government does not yet have a plan for managing those risks, or the capability to monitor them.
These six papers set out the steps we recommend the government take. Each is deliberately modest in cost and scope, and each builds on what comparable countries are already doing. Together they would give New Zealand a credible way to see the risks coming and respond to them.
The case for action is set out in our overview of the risks from artificial intelligence.
New Zealand has no dedicated government capability to monitor the risks of advanced AI and advise Ministers and regulators on them. We recommend establishing a small, expert AI Safety Institute as the government’s centre of technical and policy expertise on AI safety. It would be an advisory and monitoring body, not a new regulator, and would connect New Zealand into the growing international network of AI safety institutes.
Read the paperNew Zealand has an AI Strategy for adoption and investment, but no stated plan for managing the risks of increasingly capable AI. We recommend a Cabinet-endorsed national AI Safety Strategy – a companion to the existing strategy. The safety strategy would set out the government’s assessment of the risks, who is responsible for what, and how it will respond. For New Zealanders to adopt AI with confidence, the risks need visible, credible management.
Read the paperThe Five Eyes cyber security agencies warned in June 2026 that leading AI systems are rapidly transforming cyber risk, and this is just one of many risks advanced AI capabilities pose. No New Zealand agency yet has a clear mandate to assess frontier AI capability and advise Ministers on managing the risks. We recommend targeted investment in the AI security capability of the National Cyber Security Centre, the Electoral Commission and broader AI risk capability across government.
Read the paperThe government currently learns of new AI risks from media coverage and overseas regulators. The companies building leading models already report critical safety incidents and dangerous capability findings to governments in California and the European Union – but not to New Zealand. We recommend requiring a small number of leading developers to report the same information here, using the regulatory frameworks they already comply with overseas.
Read the paperMost of New Zealand’s statute book was written before advanced AI existed. Some of it will work as intended, but it will not all be adequate. Gaps are already surfacing in relation to deepfakes and privacy, evidence, electoral and consumer law. We recommend a targeted review of legislation to test whether it addresses the risks and societal consequences of advanced AI, with legislative change where gaps are found.
Read the paperLeading AI models are becoming core economic infrastructure and a core input to New Zealand’s cyber defences, but our access is not guaranteed: in June 2026 the US cut off foreign access to one of the most capable models overnight. We recommend the government assess New Zealand’s exposure and report on options for securing access. Getting this right protects the economic growth and productivity gains the Government’s AI Strategy is designed to deliver and ensures we have defensive capability to prevent novel cyberattacks.
Read the paperOur risk papers, and the overview these recommendations respond to, are collected alongside this work.