Responsible AI Guidance for Businesses (New Zealand)

What it is

The Responsible AI Guidance for Businesses, subtitled “Investing with confidence”, is voluntary guidance from New Zealand’s Ministry of Business, Innovation and Employment (MBIE). It was published in July 2025 as a partner product to New Zealand’s AI Strategy. It was developed with the Government Chief Digital Officer and other agencies, and consulted with the AI Forum.

The guidance draws on the values-based OECD AI Principles and reflects the proportionate, risk-based approach to AI agreed by Cabinet, under which risks are treated in proportion to their likelihood, magnitude and context. It also takes account of New Zealand values, including care for New Zealanders’ data, respect for te reo Māori, tikanga and mātauranga, and Māori data.

Who it applies to

All New Zealand businesses, including sole traders, non-profits and individual professionals, of any size and in any sector, and wherever they sit in the AI supply chain. MBIE notes that most New Zealand businesses are current or potential AI users or deployers rather than developers, and the guidance is focused accordingly. It covers the broad range of AI, not only generative AI.

How it is structured

The guidance works in three layers.

1. Understanding your why for AI. Be clear about the purpose, principles and objectives for using or developing an AI system, and plan for continuous monitoring and improvement.
2. Good business foundations for responsible AI. Use existing business expertise and functions, in proportion to the use case. This layer covers governance and accountability (assembling a team, legal obligations, risk management and recordkeeping), supporting capabilities (procurement, IT and cybersecurity, privacy, and skills) and stakeholder interactions (engagement and consultation, transparency, and feedback and complaints).
3. AI system specific considerations. Work through the lifecycle, covering data and modelling (fit-for-purpose data, legal and ethical data and model efficacy) and use and outputs (generative AI inputs, use of generative AI outputs and human-in-the-loop decision-making).

Appendices give a glossary, checklist resources, options for ethically sourcing datasets including copyright works and further guidance.

What it asks of businesses

The guidance is not a checklist to complete in full. Businesses decide which parts matter, based on where they sit in the AI lifecycle, how they work, who their customers are and their goals. MBIE still recommends reading it in full to understand the risks to watch for.

On risk, it recommends a balanced approach that takes account of how AI is used, what could go wrong and what resources are available, started early and repeated regularly. A stakeholder impact assessment helps identify who is affected and possible mitigations. Common risks it names include compromise of personal or confidential information, unfair treatment through bias or overreliance on AI, lack of transparency with customers and decisions based on generative AI hallucinations. It points to general risk tools, the Stats NZ Algorithm Impact Assessment toolkit and international standards such as ISO/IEC 42001, ISO/IEC 23894 and ISO 31000.

On transparency, it says people want to know when and how a business uses AI. Different audiences may need different levels of detail. It is good practice to tell people when a generative AI system is in use, for example a website chatbot, and to label AI-generated content. Transparency requirements of suppliers should be reflected in contracts.

On human oversight, it ties the level of human review to how reliable the system has proven to be and how much inaccuracy the business can tolerate. A low-risk recommendation may need no human check, which it calls human-out-of-the-loop. Where AI helps make big decisions about money, health or the law, a person should always be involved. It warns about automation bias, where people accept outputs without critical evaluation, and reminds businesses that they remain responsible for their decisions whatever technology supports them.

It is clear that voluntary guidance sits on top of the law. It lists the statutes most relevant to AI, including the Privacy Act 2020, the Fair Trading Act 1986, the Human Rights Act 1993 and the Bill of Rights, the Commerce Act 1986, the Consumer Guarantees Act 1993 and intellectual property law.

Key dates

The guidance was published in July 2025, alongside New Zealand’s AI Strategy. MBIE also publishes a Quickstart Guide for Business Users and supplementary resources.

How it relates to other frameworks

The guidance is the private-sector counterpart to the Public Service AI Framework. Where AI uses personal information, the Privacy Act 2020 and the Privacy Commissioner’s expectations apply. It aligns with international guidance, including the OECD principles and ISO standards, and covers similar ground to Australia’s Guidance for AI Adoption, which is useful for businesses operating on both sides of the Tasman.