Policy Paper: Towards a World-Class, Treaty-Aligned AI & Robotics Framework for Aotearoa New Zealand

Policy Paper: Towards a World-Class, Treaty-Aligned AI & Robotics Framework for Aotearoa New Zealand

Executive Summary New Zealand’s current AI governance—anchored in the Privacy Act 2020, the Government Algorithm Charter (2020) and the Public Service AI Framework (2022) underpins sound public-sector use of algorithms. In June 2024 Cabinet endorsed a risk-based “Approach to Work on Artificial Intelligence” that knit together strategic coordination, safe public-service innovation, economic opportunity, international engagement and national-security alignment. Yet these measures remain largely soft-law.
This paper proposes:

• Building on existing statutes (Privacy Act, public-service frameworks) to introduce a tiered, risk-based AI & Robotics Act.
• Embedding Te Tiriti principles via co-governance with Māori on data sovereignty, cultural impact and Wai 262-style claims for misappropriation of taonga knowledge by AI.
• Aligning with global advances (EU AI Act’s banned/high-risk/limited-risk tiers, UN resolution on AI, OECD AI Principles, G7 Hiroshima Code) to ensure NZ remains interoperable and competitive.
• Anticipating future domains—notably autonomous robots in warfare, environmental monitoring drones and synthetic biology agents—by creating advanced safety and human-control mandates.

1. Current New Zealand Landscape1.1 Foundational Legislation & Guidance

• Privacy Act 2020
  • Regulates personal data collection, storage and use—including by AI systems.
  • Introduces Information Privacy Principles requiring agencies to keep data secure and limit collection to necessary information.
• Government Algorithm Charter (2020)
  • Voluntary commitments by agencies using algorithms (e.g. transparency of intent, fairness testing, public disclosures).
• Public Service AI Framework (2022)
  • Based on OECD AI Principles, it mandates human oversight, bias assessment, and risk management for public-sector AI projects.
• Cabinet’s “Approach to Work on Artificial Intelligence” (June 2024)
  • A light-touch, risk-based strategy across five domains: strategic coordination; safe public-service innovation; economic harnessing; international engagement; national-security alignment.

1.2 Gaps & Challenges

• Non-binding nature of existing frameworks means no enforceable duties or penalties.
• Limited civilian oversight for high-risk AI, especially in defence or critical infrastructure.
• Absence of Māori data-sovereignty law at national level—taonga Māori and tikanga not protected from AI scraping or misuse.
• No explicit controls on autonomous robots, particularly in security or military contexts.

2. Global Best Practices & Emerging NormsJurisdictionKey FeaturesEuropean UnionBanned/high/limited/minimal-risk AI tiers; conformity assessments; fines up to 6% turnover; transparency obligations; AI Liability Directive.
OECDFive Trustworthy AI Principles (human-centered, robustness, transparency, accountability, fairness).
UN2023 GA Resolution on AI; UNESCO Recommendation on Ethics of AI; proposals for a UN AI Agency; ITU “AI for Good.”
G7 Hiroshima CodeVoluntary Code of Conduct for frontier AI developers: risk assessments, safety testing, watermarking, human rights alignment.
Canada (AIDA)Risk-based obligations on “high-impact” AI: mandatory risk assessment, public disclosure, enforcement by AI Commissioner (pending legislation).
ChinaAlgorithm registration; deepfake labeling; generative AI licensing; focus on social stability and security.
3. Proposed Aotearoa AI & Robotics Act3.1 Principles & Te Tiriti Foundations

• Active Protection (Article 2): Māori co-governance over cultural and data taonga—free, prior and informed consent for AI use of Māori knowledge.
• Partnership (Article 3): Joint Māori–Crown AI Council advising on policy, system certification and cultural impact assessments.
• Equity & Participation: Ensuring AI benefits reach all communities, with targeted support for Māori and Pasifika innovators.

3.2 Risk-Tiered StructureTier Examples Requirements Minimal Risk Chatbots, home assistants, toy robots Voluntary code adherence; user-notification (“This device uses AI”)
Limited Risk Industrial automation, civilian drones Baseline testing; “robot in control” labeling; basic explainability
High Risk Medical/surgical robots; critical-infra AI; autonomous weapons (military/commercial)Pre-deployment risk assessments; human-in-loop mandates; transparency registry; periodic audits
Unacceptable Risk Fully autonomous lethal weapons; social scoring; mass biometric tracking without consent Outright ban; criminal sanctions3.3 Key Provisions

1. Pre-market Conformity Assessments
   • High-risk AI/robotic systems must undergo certification (similar to MedSafe for medical devices or Waka Kotahi for autonomous vehicles).
2. Human-Control & Time-Dilation Safety
   • All lethal or safety-critical decisions require meaningful human control and override “kill-switches.”
   • Time-dilation subroutines: automatic rapid replay and rollback on detection of unsafe or contradictory behaviors.
3. Transparency & Traceability
   • Public AI & Robotics Registry listing high-risk systems, their intended uses, manufacturers, and safety summaries.
   • Mandatory model-card disclosure for training data provenance and known limitations.
4. Māori Data & Cultural Safeguards
   • A Taonga AI Permissions Process: AI providers must secure consent before using Māori language, imagery, stories or protocols.
   • Wai 262 style claims process for redress if cultural taonga are misappropriated.
5. Enforcement & Penalties
   • Establish an AI & Robotics Authority with investigatory powers, ability to suspend or recall non-compliant systems, and impose fines (scaled to turnover).
   • Offenses for deploying banned AI/robotics, failing to register or report incidents.
6. International Alignment
   • Mutual-recognition agreements with EU conformity bodies to facilitate trade and interoperability.
   • Adherence to G7 Hiroshima Code for frontier AI developers operating in NZ.
   • Participation in emerging UN AI Agency or Council of Europe AI Convention to shape and adopt global norms.

4. Extending to Future Possibilities4.1 Autonomous Systems in War

• Integrate Article 36 IHL reviews into domestic law for all new weapons (mandating legal, ethical and privacy reviews of autonomous robots).
• NZ Defence Force to adopt robotic warfare moratoria on fully autonomous lethal systems until robust global treaty is in place.

4.2 Environmental & Bio-security AI

• High-risk classification for AI/robots used in genome editing, synthetic biology, or invasive species control—requiring bio-safety assessments.
• Drone swarms for ecological monitoring subject to privacy and Māori whenua-access consent regimes.

4.3 Socio-Economic & Labor Impacts

• Mandate algorithmic impact assessments for AI replacing human labor in critical sectors (transport, care work), with transition support for displaced workers.
• Encourage public–private partnerships on AI upskilling centers, prioritizing rural and iwi communities.

5. Implementation Roadmap

1. Year 1: Draft legislation, public and iwi consultations, establish provisional AI & Robotics Authority.
2. Years 2–3: Phased rollout of registration system, begin conformity assessments for first high-risk categories, Māori co-governance council operational.
3. Years 4–5: Full enforcement, treaty negotiations on autonomous weapons conventions, integration with global AI governance bodies.

ConclusionBy codifying a risk-based AI & Robotics Act grounded in Te Tiriti, aligned with leading international frameworks, and forward-looking to emerging domains (war-bots, biotech, environmental swarms), Aotearoa New Zealand can protect rights, promote innovation, and position itself as a global leader in human-centred, ethically robust AI governance.
​