UNESCO AIEB/W4EAI; ISO SC42 on AI via Standard Council of Canada; Center of AI and Digital Policy (CAIDP)

From my perspective, the first Global Dialogue on AI Governance would be successful if it helps the UN take realistic first steps toward AI governance that is both inclusive and better coordinated globally. First, it should set a clear direction on inclusion. Many people who are deeply affected by AI are still largely absent from global conversations, including communities with limited or unstable Internet access, people who do not speak English or other UN working languages, and citizens and workers in vulnerable positions. Their data and labour are often part of digital and AI systems, but their voices rarely appear in conferences, reports or media. A meaningful outcome would be for the Dialogue to identify concrete ways for the UN system to reach these groups – for example through regional consultations, partnerships with local organisations, and support for national and community-level discussions in their local languages – and to make a clear commitment to follow through. Second, it should set a path toward a shared global baseline. There is already a great deal of work on AI principles, regulations and standards across international organisations, states, standards development organisations, companies and civil-society groups. These efforts are valuable but often fragmented. The Dialogue could add real value by mapping existing initiatives and areas of convergence, highlighting key gaps, and launching a modest set of follow-up processes to move toward a shared global reference point for safe, rights-based and inclusive AI governance, which different regions can adapt to their own contexts and capacities. If the first Dialogue can clarify these directions and agree on a few concrete next steps, it will already be an important achievement and a useful foundation for future work.

• Safe, secure and trustworthy AI
• Interoperability of governance approaches
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

I chose these four areas because they reflect some of the most fundamental and interconnected problems people face with AI in everyday life. For safe, secure and trustworthy AI, AI is now embedded in essential services such as finance, employment, health, education and public administration. Yet when we use online services or social media, we often have no real choice about how our data is used. There is often no meaningful opt-out, and if we do not accept data-sharing with third parties or "partners", we cannot use the service. This makes "consent" feel more like a requirement than a free decision, which undermines trust. For the protection and promotion of human rights, once people share their data, they usually cannot see or control what happens next with the collected data. They do not know how far their data travels or how it is combined, and there is a real risk of misuse, including discrimination and unfair targeting. This raises serious concerns for privacy and other fundamental rights. Transparency, accountability and human oversight are often missing in AI systems. Many AI-supported decisions are hard to understand or challenge, and it is not always clear who is responsible when harm occurs. In practice, the rules for data use are frequently shaped more by the de facto practices of large technology companies - which run big ecosystems of devices, apps and services - than by clear public rules. Finally, the interoperability of governance approaches is critical because AI services and data flows cross borders, while laws, principles and standards are still fragmented. We need approaches that can work together across countries and sectors, while still allowing for local differences and capacities.

Yes. One important cross-cutting issue is the need to govern how personal data moves and is used in digital and AI supply chains. Many current discussions focus on individual AI systems or high-level principles. However, serious problems arise because personal data passes through long chains of organisations - from initial collection and aggregation, through storage, analysis and model training, to deployment in downstream services. These chains typically involve multiple public and private actors, including contractors and sub-contractors in different jurisdictions. When data moves in this way, responsibilities between actors can become unclear, regulators find it hard to trace data flows, and people whose data and labour are used have very little visibility or control. This affects several of the Dialogue's themes at once: safety and trust, human rights, transparency and accountability, and the interoperability of governance approaches. In my view, digital and AI supply chain management for personal data and human rights protection should be recognised as such a cross-cutting issue. Addressing it will require coordination among the following: • public policymaking and regulation, including data protection and AI laws, and public procurement; • international technical standards and certification, for example ISO/IEC 27001 and 27701, ISO 31700 and ISO/IEC 42001; and • inclusive participation and outreach to underrepresented communities, so that people most affected by AI - but usually left out of global debates - can help shape how these policies and standards are designed and implemented.

In Canada, AI-related governance has been progressing through the Privacy Act for federal public-sector institutions and The Personal Information Protection and Electronic Documents Act (PIPEDA) for private-sector organisations, alongside emerging federal privacy and AI legislation. However, these laws and frameworks do not yet clearly define how personal data should be governed at downstream levels of digital and AI supply chains, including where contractors, subcontractors, vendors and partners may access, process or reuse personal data. This challenge is also global. EU residents benefit from the most integrated protection framework through the GDPR and EU AI Act, including clearer accountability across controller–processor chains and AI value chains. By contrast, many people outside the EU live in countries where data-protection laws are newer, weaker, unevenly enforced or absent. UNCTAD's Global Cyberlaw Tracker, accessed in 2026, shows that 79% of countries have data protection and privacy legislation, but adoption remains uneven: 65% in Asia-Pacific, 57% among least developed countries and 51% among small island developing states. Legal adoption also does not necessarily mean effective enforcement, accessible remedies or real control over data used in AI systems. International standards can help fill part of this gap, but their adoption and function are uneven. ISO/IEC 27001 is a mature, widely used information-security management standard. ISO/IEC 27701 provides a privacy information management framework and was revised in 2025 as a stand-alone management-system standard. ISO 31700 offers high-level privacy-by-design requirements for consumer goods and services, while ISO/IEC 42001 provides an AI management-system framework whose certification practice is still developing. These standards are useful tools, but not universal safeguards, especially in jurisdictions with weaker laws or limited enforcement capacity. The opportunity is to connect law, procurement, standards and capacity-building so protections extend through global supply chains to the people behind AI systems.

The AI Dialogue can play an important coordinating role by bringing together existing AI governance tools and the actors who use them, helping countries turn principles, assessments and standards into practical cooperation and harmonization. First, the Dialogue can help translate readiness assessments into cooperation priorities. Tools such as UNESCO's Readiness Assessment Methodology (RAM) already identify gaps relevant to digital and AI data supply chains, including data protection, data sharing, procurement, accountability and applied standards. Rather than repeating these findings, the Dialogue could focus on addressing weaknesses identified through such assessments, especially where personal data moves across organisations, sectors, contractors and borders. Second, the Dialogue can support practical harmonization through international standards. Standards matter because they can make governance requirements more concrete, measurable, comparable and auditable. They can help organisations operationalise trustworthiness, transparency and accountability, and where certification is available, independent assessment can strengthen public confidence. The Dialogue could give states, standards development organisations, UN entities and civil society a space to consider how standards can strengthen safeguards along digital and AI supply chains while remaining adaptable to different legal systems and capacities. Third, the Dialogue can connect these tools with law, policy and public-sector procurement rules that extend safeguards to AI vendors, contractors and partners. By bringing together countries with different experience levels, it can support mutual learning on aligning legal obligations, technical standards and institutional practice without imposing a single model. This would help move international cooperation from abstract principles toward concrete mechanisms for protecting personal data and human rights.

The AI Dialogue should build on existing initiatives rather than create entirely new structures. Normative and assessment frameworks include the UNESCO Recommendation on the Ethics of AI and RAM, which help countries assess AI governance systems and identify gaps. National and regional laws, including the GDPR and EU AI Act, and evolving frameworks in countries such as Canada, including the Privacy Act, The Personal Information Protection and Electronic Documents Act (PIPEDA), and emerging federal privacy and AI legislation, provide concrete regulatory models and lessons. International standards and certification mechanisms are also important. ISO/IEC 27001 and 27701, ISO 31700 on consumer privacy by design, and ISO/IEC 42001 on AI management systems provide practical tools for information security, privacy and AI governance. The Seoul Statement on Artificial Intelligence, issued by ISO, IEC and ITU at the 2025 International AI Standards Summit in Seoul, further shows the role of international standards in supporting a safe, inclusive, open, sustainable, fair and secure digital future. The Dialogue should also connect with broader standards development organisations, including IEEE, where relevant to socio-technical and ethical AI standards. On outreach and inclusion, UNDP, UNHCR, ITU, UNESCO, civil-society organisations, educational institutions and expert networks such as the Center for AI and Digital Policy (CAIDP) can help reach underrepresented communities and support AI literacy. The AI Dialogue could add value by mapping and connecting these initiatives, encouraging partnerships between UN entities, governments, standards development organisations, CSOs and educational institutions, and using this ecosystem to develop practical models of digital and AI supply-chain governance that protect personal data and human rights.

The AI Dialogue should be structured as a practical collaboration process, with each stakeholder group contributing to specific workstreams and outputs. UN and UN entities: The UN can convene actors, ensure regional balance and connect the Dialogue with existing UN work on AI, human rights, development and digital inclusion. Relevant UN entities can help align the Dialogue with readiness assessments, capacity-building, connectivity programmes and multilingual outreach. Governments and public authorities: Governments can identify where laws, oversight mechanisms and public-sector procurement rules do or do not extend safeguards to AI vendors, contractors and partners. They can also share enforcement experience and capacity-building needs. Standards development organisations: SDOs can explain how international standards can make AI governance requirements more concrete, measurable, auditable and interoperable. They can support workstreams on standards, certification and practical safeguards for digital and AI supply chains. Private sector: Companies can explain how AI systems, data flows and vendor relationships operate in practice, including what makes responsible governance difficult across complex supply chains. Users and data subjects: Users should have structured channels to contribute practical experience. Their consent is often formal rather than meaningful: access to public or private services frequently depends on accepting lengthy terms and privacy notices, with limited opt-out options and little visibility over downstream data use. Civil society, workers and education: CSOs, consumer groups, worker organisations and educational institutions can translate user and community experiences into policy-relevant evidence, support AI and data literacy, and connect local concerns with global discussions. Structurally, the Dialogue could combine plenaries with thematic working groups on digital and AI supply-chain governance, procurement safeguards, standards and certification, and user/data subject rights. Each group should produce practical outputs, such as model clauses, checklist-style guidance, consultation summaries or capacity-building priorities, with regional consultations and reporting back between annual meetings.

Several groups are underrepresented in global AI governance because they face barriers to access, expression, safety or visibility. 1. AI literacy and access barriers: People in developing countries and rural, remote or low-connectivity areas often lack the infrastructure needed to use AI or understand how it affects them. As a result, they may have few opportunities to express concerns or provide feedback on AI governance. 2. Language, education and expression barriers: Communities that do not speak UN official languages, Indigenous and minority-language groups, and people with limited access to education or literacy support are often excluded from formal consultations. Even when AI affects their data, work or public services, they may lack the language, format or forum needed to articulate their views. 3. Care, dependency and high-vulnerability contexts: This includes people in conflict zones, refugees and internally displaced persons, hospitalised or medically vulnerable people, persons with disabilities requiring third-party support, older persons with dementia, children and minors, women and caregivers constrained by unpaid care work, and children in developing countries who support family livelihoods. Such unpaid, household and informal family work is often poorly captured in official statistics, making these groups less visible in AI governance debates. 4. Invisible contributors and affected workers: Data labellers, content moderators, gig workers, migrant workers and other workers managed or evaluated through AI systems are often missing from policy debates, even though their labour supports AI systems or is directly shaped by them. These groups could be included through multilingual and accessible consultations, partnerships with local CSOs, refugee-support organisations, disability-rights groups, health and education institutions, worker organisations and community media. The Dialogue should also use low-bandwidth channels, locally facilitated meetings, citizen and worker panels, and feedback loops that formally summarise these perspectives into Dialogue outputs.

The AI Dialogue should use a two-way engagement model that connects top-down governance alignment with bottom-up community input. At present, many relevant actors work separately. The UN could add value by convening them, setting common directions and building a platform that links their work. 1. Top-down alignment labs: The Dialogue could create AI and data supply-chain alignment labs. These would bring together governments, regulators, UN entities, standards development organisations, companies, policy-focused CSOs and expert networks, including CAIDP, to compare existing tools, including readiness assessments, privacy and AI laws, standards, certification schemes and public-sector procurement rules. Their task would be to identify overlaps, inconsistencies and gaps in protecting personal data and human rights across digital and AI supply chains, especially where data moves through vendors, contractors, partners and borders. 2. Bottom-up listening and AI literacy network: The Dialogue could support a community listening and AI literacy network. UNESCO, UNDP, UNHCR, UN Women, UNICEF, WHO, national educational institutions, local CSOs, consumer-protection groups, community organisations, worker organisations and community media could work with local schools, refugee-support groups, women's organisations, health institutions and disability-rights groups. They could run multilingual and accessible AI literacy sessions, explain how data and AI affect daily life, and collect concerns from groups usually absent from global debates. 3. Feedback platform: The UN could develop a feedback platform connecting these two tracks. Regional AI governance clinics, citizen and worker panels, low-bandwidth surveys, radio dialogues, school workshops and sessions in refugee, healthcare and community-support settings could produce short, standardised summaries of concerns and recommendations. These inputs should then feed into Dialogue working groups on standards, procurement, data subject rights and supply-chain governance. This would make engagement continuous rather than one-off and would connect expert frameworks with lived experience.

The AI Dialogue should use a two-way engagement model that connects top-down governance alignment with bottom-up community input. At present, many relevant actors work separately. The UN could add value by convening them, setting common directions and building a platform that links their work. 1. Top-down alignment labs: The Dialogue could create AI and data supply-chain alignment labs. These would bring together governments, regulators, UN entities, standards development organisations, companies, policy-focused CSOs and expert networks, including CAIDP, to compare existing tools, including readiness assessments, privacy and AI laws, standards, certification schemes and public-sector procurement rules. Their task would be to identify overlaps, inconsistencies and gaps in protecting personal data and human rights across digital and AI supply chains, especially where data moves through vendors, contractors, partners and borders. 2. Bottom-up listening and AI literacy network: The Dialogue could support a community listening and AI literacy network. UNESCO, UNDP, UNHCR, UN Women, UNICEF, WHO, national educational institutions, local CSOs, consumer-protection groups, community organisations, worker organisations and community media could work with local schools, refugee-support groups, women's organisations, health institutions and disability-rights groups. They could run multilingual and accessible AI literacy sessions, explain how data and AI affect daily life, and collect concerns from groups usually absent from global debates. 3. Feedback platform: The UN could develop a feedback platform connecting these two tracks. Regional AI governance clinics, citizen and worker panels, low-bandwidth surveys, radio dialogues, school workshops and sessions in refugee, healthcare and community-support settings could produce short, standardised summaries of concerns and recommendations. These inputs should then feed into Dialogue working groups on standards, procurement, data subject rights and supply-chain governance. This would make engagement continuous rather than one-off and would connect expert frameworks with lived experience.