The New Zealand Government this week this week (January 29, 2025) released its first non binding AI Framework for the public sector with ethics, human and Indigenous Peoples considerations. View the Public Service AI Framework or downloadable pdf here).
The Aotearoa New Zealand government Cabinet has agreed to a light-touch, proportionate and risk-based approach to AI regulation. They’ve recognised the NZ policy context already provides some guardrails and have agreed that further regulatory intervention should:
-
only be considered to unlock innovation or address acute risks
-
use existing mechanisms in preference to developing a standalone AI Act.
I have discussed this in another post how a standalone AI act would directly discriminate against the Indigenous Peoples – Māori, and likely others minority communities in Aotearoa New Zealand and I share the views of the government that it would likely hinder AI developments.
According to the AI Framework background information, several other frameworks guided the creation of this AI Framework, including the international OECD AI principles that the Aotearoa New Zealand government Cabinet agreed in to promote as a key direction to approach to responsible AI.
It was positive to read that the Aotearoa New Zealand Algorithm Charter was another framework that was used for guidance, but does not appear to be incorporated in any way. The Algorithm Charter has been widely criticised.
Dr Andrew Chen of the Koi Tū: The Centre for Informed Futures raises a number of issues including “The Charter doesn’t really engage with Māori, including in terms of Māori data sovereignty.” Dr Chen wrote more extensively on the topic in More Zeros and Ones which also cover my own concerns that the charter further significantly failed Māori by being overly tokenistic. Taylor Fry reviewed that Algorithm Charter for Aotearoa New Zealand after its first year in December 2021 . Overall, the new review scores the charter’s first year: ‘OK, could do better’. But made a number of concerning findings, but largely did not consider Te Tirti in its findings, only as a reference. RadioNZ noted Plus, compliance was very light-handed. “Some greater enforcement might be necessary to keep social licence.” This appears at odds with OECD principles on artificial intelligence that say how systems work should be disclosed so people can challenge them.
Other issues included with some of the signatories of the Algorithm Charter such as: Police, ACC, MSD (more recently) and Corrections (and here) who after signing the Algorithm Charter to say they supported The Treaty of Waitangi (Te Tiriti o Waitangi) and Māori, human rights etc, were publicly accused of discriminating against Māori, Pasifika and other minorities.
Again, this reinforces the strength based approach of using international best AI governance practices, Aotearoa New Zealand legislation, recognising Māori the Indigenous Peoples of Aotearoa New Zealand and the 6 key pillars, thus in my opinion making this an international Aotearoa New Zealand AI framework that all countries and all Indigenous Peoples should consider along other AI governance best practices.
While the AI Framework is intended only for the Public Service Sector leaving the industry such as the Aotearoa New Zealand AI Forum to continue advocacy in this area. The intention of the AI Framework is to support leaders, decision-makers, practitioners and influencers of AI within Public Service agencies to use AI lawfully and in line with Public Service values and to support a structured approach to the development, deployment and use of AI across the New Zealand Public Service, contributing to the wider community and economy in alignment with New Zealand’s position as a trusted global partner.
The AI Framework consists of five broad containers of guidance:
- Vision statement
- 5 AI Principles
- New Zealand policy context (including Indigenous Peoples constitutional founding document)
- Word Programme (6 key pillars)
- Outcome
1. Vision:
Adopt AI responsibly to modernise public services and deliver better outcomes for all New Zealanders.
The vision meets many best practice AI governance practices and could lead Aotearoa New Zealand to a world class AI driven public service. It is well documented that in the past tech and in particular AI has been dominated by young white men’s design, and older wealthier white men’s money that prioritises fast technical developments over any human rights and environment considerations.
For example, in the USA:
- Women make up 28% of staff in AI roles such as computing and mathematical roles
- Asian women 5%, Black women 3%, and Hispanic women 1% of the 28% of women working in tech.
- 50% of women who go into tech dropping out by the age of 35 and are less likely to get promotions
AND
- White women are significantly more likely to be executives than minority men, with 88% Asian and 97% Black men being less likely.
- In 2016, none of the top ten Silicon Valley technology firms employed a single black woman, and only a few employed any black employees.
- 62% of US tech sector jobs are held by white Americans, 7% by Black Americans, 8% by Latinx Americans, and 20% by Asian Americans.
2. AI Principles
The OECD’s values-based AI principles were used to inform these principles. This allows for less risk as the OECD principles are widely implemented and acknowledged, and will reduce duplication of work in my opinion.
Inclusive, sustainable development
•Innovation
•Efficiency
•Resilience
•Address inequalities and transition equitablyHuman-centred values
•Privacy and data ethics
•Upholding democracy and the rule of law
•Human and labour rights
•Human oversightTransparency and explainability
•Promote awareness and understanding of AI systems and interactions
•Explainable outcomesSafety and security
•Safety by design
•Data protection
•Traceable data
•Robust risk management, including national security risksAccountability
•Governance
•Regulatory frameworks
•Auditing with human oversight
•Capability growth
As a certified professional with most of the international AI Governance standards, I think these principles within the overall framework could put Aotearoa New Zealand on the international stand of AI Governance excellence.
3. New Zealand policy context
Current laws that apply to AI: Bill of Rights Act, Human Rights Act, Privacy Act, Public Records Act, Public Service Act, Copyright Act, Official Information Act and others.
Wider consideration of existing Aotearoa New Zealand laws are important to ensure a human rights aspect and a unique Aotearoa New Zealand perspective are considered. This places a guardrail that delays the need to regulate AI at this stage. In a previous post I also supported that these laws should be considered and updated for any AI Framework.
Under the same heading, is a sub section that recognises the country’s Indigenous Peoples, Māori and the constitutional agreement called the Treaty of Waitangi (Te Tiriti) Māori have with the government.
Treaty of Waitangi
Public Service commitments and Māori views of AI for ethics, bias and data
This is significant as it not only writes the standard prose that we often see in some government documents that only recognise Iwi (tribes) who are a minority community voice. This statement recognises that Māori who are individuals who identify as having Māori descent and make up a myriad of different community collectives, who make up about 20% of the population and a significant part of the economy and international trade have a voice.
Under the current government, there is a new directive to not use the Māori language first for names (or at all) etc, despite the Māori language being an official language of Aotearoa New Zealand. The wording could raise some concerns as people refer to Treaty of Waitangi as English version which is different than the Māori version. But in the legislation titled Treaty of Waitangi Act 1975 Section 2 Interpretation, Treaty means the Treaty of Waitangi as set out in English and in Maori in Schedule 1.
The emphasis of ethics and bias is also important, as these are internationally recognised human rights issues that must be addressed as good AI governance. Of note, the largest parts of society who are impacted by not including ethics and bias are People of Colour, Indigenous Peoples, females and LGBTQ+.
In another positive manner, this statement appears to overlook the advise given to the Minister of Digitising Government, from a minority Māori tribal group.
“I have met with recently met with the Data Iwi Leaders’ Group (DILG) and was pleased at the ready
uptake of AI by the DILG. There seems to be a common acceptance that the danger for Māori in AI is to be left outside of its ability to bring about substantial improvements in the provision of services for Māori.”Source: Minister Hon Judith Colins. Cabinet paper: Approach to work on Artificial Intelligence. Released 25 July 2024 pg 4&5. Author: Office of the Minister of Science, Innovation and Technology.
The above advise appears at odds with their parent representative National Iwi Chairs Forum who are in partnership with NetSafe to address the over representation of Māori with online safety issues. It is also at odds with international best practices, and with a number of research reports including NetSafe, the AI Forum, DataCom, InternetNZ and academic research that proves that negative engagements with online technologies results in disengagement.
Potential barriers or enablers for AI adoption
Coordinated approach to assessing policies acting as a barrier to or enabler of AI adoption
The previous commentary is great example of how the AI Framework has already been proactive and stopped potential bias and harms to 20% of the population. If rolled out correctly, Aotearoa New Zealand could be international leaders in addressing human rights and bias issues.
4. Word Programme
This intention of the section to to support agencies to embrace AI responsibly. The guidance of the 6 pillars are strong.
I think that any government claiming social licence is dangerous for the general population, but in consideration of the 5 other pillars, namely with Guardrails, the risk balance should be acceptable. For Māori, the Treaty of Waitangi clause is an added layer of protection.
6 key pillars
- Governance; Ensure human accountability for inclusive implementation of data and AI use
- Innovation; Develop pathways for safe AI innovation for government agencies
- Guardrails; Support safe and trustworthy use of AI and its underpinning data
- Social Licence; Understand how to build public trust and worker engagement in government use of AI
- Capability; Build internal and external AI capability and safety by design
- Global Voice; Build our global reputation as trusted partner and Public Service AI enabler
5. Outcome Statement
The Public Service models best practice in AI use, enabling and contributing to the wider community and economy in alignment with New Zealand’s position as a trusted global partner.
This will require that civil society, Māori and the AI industry monitor how the best practices are implemented and to work with the public sector to ensure all voices are being heard, not influential lobbyists with their own mandates that represent national communities.
Conclusion
Aotearoa New Zealand is not rushing forward to regulate AI as other developed countries have. But are taking a more cautious approach that I believe is the best approach at this stage. The human rights based approach, caution and making the AI Framework opt in, are all beneficial for all of Aotearoa New Zeeland and will allow us to be seen as international best practice.
As AI develops and we see the monopoly of American AI tech loose its dominance over the world, Aotearoa New Zealand will need to consider AI regulation to protect our sovereign country from being held to ransom by hardware manufacturers, so we can develop our own AI, tech oligarchs and to ensure we as a country continue to participle in international trade. The recent DeepSeek AI has shown the world that AI can be produced at minimal costs that once prohibited AI developments in other countries and has shown Aotearoa New Zealand could create its own sovereign AI and expect good societal and economic returns from doing so.