Solopreneur

In my opinion, the first Global Dialogue on AI Governance would be a success if it produces practical, human-centred outcomes rather than only broad statements of principle. A successful outcome would include clear agreement that AI governance should focus on real-world harms: discrimination, surveillance abuse, fraud, deepfakes, cyber misuse, labour disruption, lack of transparency, and concentration of power. The Dialogue should avoid both uncritical hype and vague fear, and instead focus on measurable risks and workable safeguards. It would also be a success if the Dialogue recognises that AI governance must not entrench the power of a few large companies or wealthy states. Open-source AI, independent developers, researchers, small businesses, civil society, and creators should remain part of the AI ecosystem. Governance should distinguish between high-risk deployment and general research, experimentation, or open development. Another important outcome would be international support for transparency and accountability where AI affects people's rights, opportunities, or access to public services. People should know when AI is involved in important decisions, understand its role, and have a clear way to challenge harmful or unfair outcomes. The Dialogue should also make progress on common standards for synthetic media, provenance, and labelling, while protecting privacy and free expression. Finally, success would mean giving meaningful attention to smaller countries, developing nations, disabled people, local languages, and under-resourced communities. AI should expand human agency and opportunity, not deepen existing divides. In short, the first Dialogue would succeed if it creates a foundation for practical cooperation: shared norms, enforceable accountability, inclusive participation, and a clear commitment that AI should serve the public interest, not just institutional or corporate power.

• Transparency, accountability, and human oversight
• Protection and promotion of human rights
• Open-source software, open data and open AI models
• AI capacity-building

My selections reflect the areas where AI governance is most likely to affect ordinary people in practical ways. Transparency, accountability, and human oversight are essential because AI systems are increasingly being used in decisions that affect people's lives, including employment, education, public services, healthcare, finance, media, and access to information. People should know when AI is being used, understand its role, and have a clear way to challenge harmful, unfair, or incorrect outcomes. The protection and promotion of human rights should be central to AI governance. AI must not become a tool for discrimination, surveillance abuse, censorship, exclusion, or the concentration of power. Human dignity, privacy, due process, accessibility, and freedom of expression should remain core principles. I also selected open-source software, open data, and open AI models because open development is important for innovation, competition, research, transparency, and accountability. Governance should not accidentally create a system where only a small number of large companies or wealthy states can build and control advanced AI. Finally, AI capacity-building is urgent because the benefits and risks of AI are not evenly distributed. Smaller countries, developing nations, local communities, disabled people, educators, small businesses, and independent creators need access to knowledge, infrastructure, tools, and training. Without that, AI could deepen existing inequalities instead of expanding opportunity.

Yes. One important cross-cutting issue is the concentration of AI power. Compute access, advanced models, data, technical talent, and distribution channels are increasingly controlled by a small number of large companies and wealthy countries. This creates risks for competition, democratic accountability, national sovereignty, and public trust. AI governance should address not only safety, but also who has the power to build, deploy, audit, and benefit from these systems. Another emerging issue is synthetic media and information integrity. AI-generated images, video, audio, text, and deepfakes are becoming more realistic and easier to produce. This affects elections, journalism, scams, public trust, personal reputation, and social cohesion. International work on provenance, labelling, verification, platform responsibility, and public education will be important. A further issue is human dependency and deskilling. As AI systems become more capable, people may increasingly outsource judgement, creativity, research, communication, and decision-making to automated systems. This could weaken human agency if not handled carefully. AI should support human capability, not replace independent thinking. Finally, accessibility should be treated as a core issue, not a side concern. AI has major potential to assist disabled people through communication tools, mobility support, education, employment access, and independent living. But poorly designed AI systems can also exclude disabled users. Inclusive design and disability representation should be built into AI governance from the start.

The governance gaps in these areas affect New Zealand, the wider Pacific region, and independent technology users in several ways. The most significant challenge is that AI adoption is moving faster than public understanding, regulation, and institutional capability. AI systems are already influencing work, education, media, public services, cybersecurity, and access to information. Without clear transparency, accountability, and human oversight, people may not know when AI is affecting them or how to challenge unfair, inaccurate, or harmful outcomes. For smaller countries like New Zealand, another major challenge is dependency on large overseas technology companies. Most advanced AI infrastructure, models, cloud platforms, and compute resources are controlled offshore. This creates risks around sovereignty, data protection, pricing, local relevance, Māori data governance, Pacific language support, and long-term resilience. There are also risks around misinformation, synthetic media, scams, and election integrity. AI-generated content can spread quickly, and smaller media ecosystems may have fewer resources to detect and respond to it. At the same time, the opportunities are significant. AI can help New Zealand and Pacific communities improve productivity, education, healthcare access, disaster response, accessibility, small business capability, public service delivery, and local-language tools. For disabled people, AI also has real potential to improve communication, independence, employment access, and participation in society. Open-source AI is especially important because it can reduce dependency on a few dominant companies and allow local developers, researchers, educators, and small organisations to build tools suited to local needs. Overall, the key issue is balance. New Zealand and the Pacific need governance that protects human rights, privacy, accountability, and fairness, while still enabling innovation, capacity-building, open development, and practical public benefit.

The AI Dialogue can play an important role by creating a trusted international space where governments, civil society, researchers, industry, smaller countries, and affected communities can discuss AI governance openly and practically. Its most useful role would be to help align countries around shared principles and practical standards without forcing a rigid one-size-fits-all model. AI governance needs international cooperation because AI systems, data flows, platforms, cyber risks, synthetic media, and supply chains all cross borders. No country can manage these issues alone. The Dialogue can help identify common expectations for transparency, accountability, human oversight, privacy, safety testing, incident reporting, and redress when AI causes harm. It can also support cooperation on synthetic media, deepfakes, scams, cybersecurity, and election integrity, where international coordination is especially important. Another valuable role is capacity-building. Smaller countries and developing regions need access to expertise, tools, infrastructure, training, and policy support so they are not only rule-takers in a system shaped by larger powers and major technology companies. The Dialogue should help ensure that AI governance includes the needs of small states, the Pacific, local languages, disabled people, and under-resourced communities. The Dialogue can also encourage interoperability between national governance approaches. Countries do not need identical laws, but they do need compatible frameworks so that rights, safety, innovation, and accountability are not weakened by regulatory gaps. In short, the AI Dialogue should act as a bridge: between countries, between public and private actors, and between innovation and public-interest safeguards. Its success should be measured by whether it helps produce practical cooperation, not just broad declarations.

The AI Dialogue should build on existing international work rather than duplicate it. Relevant initiatives include the UNESCO Recommendation on the Ethics of Artificial Intelligence, the OECD AI Principles, the G7 Hiroshima AI Process, the Global Partnership on AI, the Council of Europe Framework Convention on AI and human rights, the work of the International Telecommunication Union, and AI-related standards work by bodies such as ISO and IEC. It should also connect with regional and national efforts, including the European Union AI Act, ASEAN guidance on AI governance, and emerging AI strategies from smaller states and developing regions. The added value of the AI Dialogue is that the United Nations can provide a broader and more inclusive forum than many existing mechanisms. Some AI governance discussions are led mainly by wealthy states, major technology companies, or technical standards bodies. The UN can help bring in smaller countries, developing nations, civil society, disabled people, independent researchers, open-source communities, educators, workers, and affected communities. The Dialogue can also help connect technical, human rights, economic, cultural, linguistic, and development perspectives in one place. That is important because AI governance is not only about safety or innovation. It is also about power, access, rights, accountability, and public trust. Its strongest contribution would be coordination: mapping what already exists, identifying gaps, encouraging interoperability, sharing best practices, and helping countries build capacity. It should avoid creating another layer of vague principles. Instead, it should help turn existing work into practical cooperation, especially around transparency, accountability, human oversight, open AI, synthetic media, accessibility, and support for under-resourced communities.

Different stakeholders should contribute according to their practical experience, not just their institutional status. Governments can share policy approaches, legal frameworks, public-sector use cases, and capacity-building needs. Technology companies can provide technical expertise, risk information, model documentation, safety practices, and infrastructure insights. Civil society can highlight impacts on rights, privacy, equality, accessibility, and public trust. Researchers can contribute evidence, evaluation methods, and independent analysis. Open-source communities and independent developers can explain how innovation actually happens outside large companies. Workers, educators, creators, disabled people, youth, and affected communities should be included because they experience the real-world effects of AI directly. The Dialogue should be structured to avoid becoming a stage for only governments and major corporations. It should include open written submissions, public livestreamed sessions, regional consultations, expert panels, technical workshops, and smaller working groups focused on specific issues such as transparency, synthetic media, open AI, human rights, accessibility, and capacity-building. There should also be accessible ways for individuals and smaller organisations to participate. This means plain-language materials, remote participation, multilingual access, disability accommodations, transparent agendas, published summaries, and clear explanations of how public input is used. The format should combine high-level political discussion with practical technical and social detail. Broad principles are useful, but the Dialogue should also examine real examples: AI in public services, education, employment, healthcare, policing, media, cybersecurity, and elections. Finally, the Dialogue should have continuity between meetings. It should publish outcomes, track recommendations, identify unresolved issues, and maintain working channels between annual sessions. Inclusive participation only matters if it influences the final agenda and outputs.

Several voices are underrepresented in global AI governance discussions. Smaller countries, developing nations, and Pacific communities are often included less than major powers and large markets, even though they may be highly affected by AI systems built elsewhere. They should be included through regional consultations, funded participation, capacity-building programmes, and direct representation in working groups. Disabled people are also underrepresented. AI has major implications for accessibility, communication, mobility, employment, education, and independent living, but disabled people are often discussed as beneficiaries rather than included as experts. The Dialogue should require accessible participation, disability representation, and consultation with disabled-led organisations. Indigenous communities, including Māori and Pacific peoples, need stronger representation, especially on data governance, cultural knowledge, language preservation, consent, and sovereignty. Their perspectives should be included through culturally appropriate engagement and respect for Indigenous data governance frameworks. Open-source developers, independent researchers, small businesses, educators, creators, and workers are also often overshadowed by large technology companies and governments. They should be included through open submissions, technical workshops, public calls for evidence, and seats in advisory groups. Young people and students should also have a voice, because they will live with the long-term consequences of AI governance decisions. To include these groups meaningfully, the Dialogue should provide remote participation, multilingual materials, plain-language summaries, accessibility accommodations, travel support where possible, transparent agendas, and clear feedback loops showing how input influenced outcomes. Inclusion should not be symbolic. Underrepresented communities should help shape the questions, priorities, and recommendations, not merely respond to decisions already made.

The AI Dialogue should use formats that encourage practical discussion, not only formal speeches. One useful format would be problem-focused working sessions. Instead of broad panels, participants could work through specific scenarios such as AI in public services, synthetic media during elections, algorithmic hiring, AI-assisted education, healthcare triage, cyber misuse, or accessibility tools for disabled people. This would make the discussion more grounded and useful. Another valuable format would be multi-stakeholder roundtables with balanced representation. Each table should include governments, industry, researchers, civil society, open-source developers, smaller countries, and affected communities. This would reduce the risk of one group dominating the conversation. The Dialogue could also use "citizen evidence sessions" where ordinary people, workers, educators, disabled people, creators, and small businesses describe how AI is already affecting them. These lived-experience sessions should sit alongside technical and policy discussions. Interactive digital participation would also help. The Dialogue should allow remote input, live polling, multilingual Q&A, public comment threads, and transparent summaries of key themes. This is especially important for people who cannot travel. Technical demonstrations could be useful too. Showing real examples of deepfakes, AI bias, accessibility tools, model transparency, open-source systems, and governance audits would make abstract risks and opportunities easier to understand. Finally, the Dialogue should include follow-up mechanisms after the event: public draft recommendations, comment periods, online working groups, and progress tracking. Meaningful engagement should not end when the meeting ends. The best format would combine high-level diplomacy with practical workshops, public participation, technical evidence, and direct input from affected communities.

Effective AI governance should combine rights protection, technical standards, transparency, and practical accountability. Useful examples include risk-based governance approaches, where higher-risk AI systems face stronger requirements than low-risk or experimental uses. This helps avoid over-regulating harmless innovation while still applying strict safeguards to AI used in areas such as healthcare, employment, finance, education, policing, immigration, and public services. Algorithmic impact assessments are another good practice. Before deploying AI in sensitive settings, organisations should assess risks to privacy, fairness, accessibility, security, human rights, and public trust. These assessments should not be box-ticking exercises. They should include meaningful consultation, especially with affected communities. Independent audits and evaluation are also important. High-impact AI systems should be tested for bias, reliability, security, explainability, and failure modes. Results should be documented clearly enough for regulators, users, and affected people to understand. Public-sector AI procurement rules are a practical tool. Governments should require transparency, human oversight, data protection, appeal mechanisms, and clear accountability before buying or deploying AI systems. For synthetic media, content provenance and labelling standards can help address deepfakes, fraud, and misinformation. These should be developed internationally and implemented in ways that protect privacy and free expression. Open-source AI and open standards can also promote accountability, competition, research, and local innovation. They allow more people to inspect, adapt, and improve AI systems rather than relying only on closed corporate platforms. Finally, accessibility-by-design should be treated as a core governance practice. AI systems should be tested with disabled users and built to support inclusion from the start, not added later as an afterthought.