In collaboration (labelled MxN) with Mission: Mārama, Nakedize is proud to be designing and evolving a Universal AI Declaration (UAID) for Aotearoa New Zealand.
We recently socialised the UAID at the Aotearoa AI Summit in Wellington and are very grateful for the support of the Social Licence Workstream of the AI Forum for making this happen, and for their ongoing collaboration.
Latest collaboration
We are now evolving the UAID with the the AI for Good Aotearoa community of practice and have a Lunch & Learn session online at noon on 28 October. You can register here.
At this session, we will be reviewing the eight Declaration commitments (which declarers can make or not make), and three default guardrails for meeting each commitment.
| Responsible AI Commitment | Commitment description |
|---|---|
| C1 Purpose | Our commitment to only use AI for the benefit of all |
| C2 Accessibility | Our commitment to ensure universal and equitable access to beneficial AI technology |
| C3 Boundaries | Our commitment to prohibit and prevent the misuse of AI |
| C4 Ecological responsibility | Our commitment to protecting nature and to resource stewardship within planetary boundaries |
| C5 Social responsibility | Our commitment to support wellbeing and thriving societies |
| C6 Ethical responsibility | Our commitment to preserving human judgement and accountability |
| C7 Cognition responsibility | Our commitment to preserving human capability and creativity |
| C8 Cultural responsibility | Our commitment to respect confidentiality, privacy, IP, culture and all forms of data sovereignty |
After the session, we will update a living document (Google Doc) containing all key content from the Declaration to incorporate feedback from the session. Then we’ll share the updated document with all participants so they are able to provide further feedback about the commitments and guardrails, as well as feedback on other key content (AI facts and oversight mechanisms).
What is a Universal AI Declaration?
- A public statement that’s completed online
Any entity (from a sole trader to a multinational) declares why and how it uses AI. - Sections for Understanding, Commitments and Guardrails, Governance
Declaration sections covering common facts the organisation understands about AI, what it commits to achieving and avoiding, and how it’s governing its AI use. - A transparent promise that renews and evolves every 12 months
Commitments are renewed annually, so the public sees progress rather than promises.
Declaration prototypes
Having widely socialised our v6 prototype, we are working on v7.
If you wish to review the latest prototype, please email simon@nakedize.com.
We have summarised feedback on v6 and plans and requests for help with v7 further below on this page.
Why is a declaration essential for responsible AI?
Trust doesn’t come from reports alone.
It comes from organisations saying out loud what they are doing and then being held accountable to their commitments.
Who is involved?
How does the declaration work?
The Declaration has three sections:
- Understanding – 32 AI facts in 5 groups
A widely-sourced, plain-language baseline of what AI is and its societal impacts – both helpful and harmful - Commitments and Guardrails – 8 commitments and 24 supporting guardrails
Responsible AI commitments that address what we know about AI (the facts), and that are each supported by three guardrails (rules for acting responsibly) - Governance – 5 governance mechanisms
The mechanisms include a publicly accessible AI policy
The Declaration is bound by a set of rules:
- One declaration per entity (regardless of entity size, thus ‘universal’)
- Renewed every 12 months
- AI policies must be publicly accessible
- The declaration is signed by a named, appropriately-authorised representative
What will change because of the UAID?
- During the first year engagement and dialogue begins so trust can grow.
- From the second year and beyond responsible AI performance improves and best practice spreads.
Who is behind the UAID?
MxN (comprising Mission Marama Ltd and Nakedize Ltd) is a collaboration for public-good sustainability projects related to information, AI, and social change.
This work is a charitable project being delivered as part of the Same Page Initiative. This initiative is dedicated to addressing issues that seem too big to solve by engineering the simplest possible (public movement) solutions through concise and simple information. Funding is still on the todo list, so please reach out if you can help on that front.
Juhi Shareef, Mission: Mārama
Juhi is an AI entrepreneur and experienced sustainability practitioner. Her impact company is currently building an ethical AI collaboration platform for mission-driven groups. Her previous role was Chief Responsibility Officer for Tourism Holdings Limited. In May, Juhi served with Emma MacDonald and Albert Bifet on the panel discussion Beyond the Tools: Sustainable AI in Practiceat the AI and Creativity Summit in Wellington.
juhi@marama.global | marama.global | 020 4128 5586
Simon Hertnon, Nakedize
Simon is an author, educator, and thought leader shining a light on the impacts of unneeded information. He has just launched the Helpful Information Framework to expand and enhance plain language practice. In partnership with Victoria University of Wellington, Simon has taught writing skills to thousands of knowledge workers. He is also the inaugural head judge of the New Zealand and Australia Plain Language Awards.
simon@nakedize.com | nakedize.com | 027 573 4464
What feedback have we received on v6 and what are our responses?
Here’s a rough list :
- Is there a verification or challenge mechanism? (Because if claims were later shown to inaccurate or unsupported, the initiative could ultimately undermine rather than strengthen trust, ie, how do we avoid AI-washing?)
We have many ideas, all based on a strategy to apply sunlight (transparency) by enlisting millions of sets of eyes, facilitating feedback loops, and sharing the oversight load across a community of experts seeking to support responsible AI. See the next section for specific developments. - Potential to split education (or preparation) elements into a separate resource
We’re working on this led by the requirement that the UAID must be self-sufficient (and not over-complicated and overwhelming) whilst also drawing on and linking to the most helpful supporting and educational material available. - More concrete definition of AI computation
We have already added the OECD definition: “An AI system is a machine-based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptiveness after deployment.” — OECD - Useful for commitments to be more specific, and helpful to know the guardrails before making the commitments
We have clarified the commitments by moving the guardrails forward from the Governance section to sit next to each commitment. - Human Rights Commission alignment
We identified this need and have already added alignment notes to 14 of the 32 facts. - A context for Declaration makers of jurisdiction-level governance rules and mechanisms
We also identified an opportunity to research and highlight governance requirements that sit at a country (or other jurisdictional) level and have captured key UNESCO ethics and governance information to highlight. We envisage a status (or report) card for national governance requirements for each country’s UAID.
What are we working on and evolving for subsequent versions?
Again, here’s a rough (non-exhaustive) list:
- We are concentrating on engagement and feedback loops because high participation breeds transparency, trust, and assurance. Thus we will:
- highlight that transparency is promoted by entities having a Staff concerns channel and Public concerns channel (two of the five oversight mechanisms)
- promote engagement and governance bright spots to inspire efficient, widespread improvement
- provide a platform or simple method for public feedback to supplement existing staff and/or public feedback channels, or to provide a consistent feedback mechanism for the public if the AI user doesn’t have feedback channels
- We are determined to avoid an exhaustive, effort-intensive oversight approach that drastically reduces engagement and participation (a la B-Corp). Instead, the UAID is a light-touch, fear-free pathway to both disclosure and engagement.
- We are exploring a peer review option (AI experts can be involved), so that entities completing the Declaration can request (pay for) peers to provide really valuable and up-to-date guidance on their draft Declaration.
- We will add persona-based (eg, sole trader, SME, corporation, public agency, academic institution) help and suggestions for completing the declaration, either within the form, as a separate page, or as a combination.
How do you contribute feedback or put your hand up for more involvement?
See the AI For Good Aotearoa Community of Practice notice at the top of this page.
Provide Juhi or Simon with your name and contact details, and we’ll let you know how. We’re seeking help with:
- Reviewing key content
- Accessibility design
- Fundraising
- Web App design and build
- Roadshow hosting and participation
- Promoting the UAID
More questions?
Contact Juhi or Simon (contact details above).