Arbor ESG Tech Innovation

Having just participated in the stakeholder consultation, I would judge the first Global Dialogue a success against four concrete outcomes. First, a working interface with the Independent International Scientific Panel. The Dialogue's authority depends on evidence, not aspiration. Success means establishing how the Panel's findings will systematically feed into the agenda — ideally with a first commissioned assessment on a specific issue (frontier risks, AI in critical infrastructure, or labour-market exposure) whose conclusions shape the May 2027 New York session. Second, a stocktaking of governance interoperability rather than another declaration. The OECD AI Principles, the G7 Hiroshima Process, the Council of Europe Framework Convention, and the EU AI Act already exist. The Dialogue's distinctive value is coordination — an honest map of where these regimes converge and diverge, so smaller economies are not forced to choose between competing great-power rulebooks. The MIT AI Risk Initiative's April 2026 mapping already shows that socioeconomic risks receive comparatively little attention across 1,000+ frameworks. Closing that gap would be a tangible deliverable. Third, real multistakeholder substance — not parallel monologues. Civil society, technical experts, industry, and the Global South must shape conclusions, not observe them. If Geneva produces structured, interactive sessions whose outputs visibly reflect non-state input, the Dialogue's legitimacy is established. If it does not, universal participation collapses into symbolic participation. Fourth, tangible movement on capacity-building. Concrete pledges to the Global Fund for AI capacity, compute credits, and fellowship programmes — so the roughly 118 countries currently outside any major AI governance initiative gain real footing rather than symbolic seats. The trap to avoid is confusing a successful event with successful governance. Photographs are easy; durable cooperation is hard. Geneva should be measured ten years from now — by whether universal participation translated into durable cooperation.

• Safe, secure and trustworthy AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI

1. Safe, Secure, and Trustworthy AI As AI capabilities advance and deployment broadens across critical sectors, ensuring that these systems remain safe, secure, and worthy of public trust is a shared interest of all Member States and stakeholders. The transboundary nature of frontier risks - including misuse, failures in critical infrastructure, and degradation of the information environment - means that gaps in any single jurisdiction can become global vulnerabilities. International dialogue offers a valuable opportunity to develop shared evaluation approaches, exchange information on emerging risks, and advance cooperation on incident response. Coordinated engagement under the auspices of the United Nations would meaningfully complement existing efforts and contribute to global stability and to the responsible advancement of the technology. 2. Societal, Economic, Ethical, Cultural, Linguistic, and Technical Impacts of AI The implications of artificial intelligence extend across the social, economic, ethical, cultural, linguistic, and technical dimensions of human life, with effects that vary considerably across regions and communities. Sustained attention to this thematic area would help ensure that the benefits of AI are broadly shared, and that governance approaches reflect the rich diversity of cultures, languages, and developmental contexts represented within the United Nations membership. Inclusive deliberation on these matters would also support the alignment of AI development with the 2030 Agenda for Sustainable Development, and help guard against the emergence of new structural divides across countries and communities. 3. Protection and Promotion of Human Rights International human rights law offers a universally recognised and well-established framework that can helpfully inform deliberations on AI governance throughout the technology's lifecycle. Building upon this shared foundation would allow the Dialogue to draw on existing consensus, complement the work of relevant United Nations human rights bodies, and offer a coherent reference point accessible to all Member States and stakeholders. Sustained attention to the protection and promotion of human rights would also reinforce the human-centred approach reflected in the Global Digital Compact, and provide an operational basis for accountability across the lifecycle of AI design, deployment, and use. 4. Interoperability of Governance Approaches The international AI governance landscape is increasingly rich in initiatives - including the OECD AI Principles, the G7 Hiroshima Process, the Council of Europe Framework Convention, the European Union AI Act, and a growing range of national and regional frameworks. Sustained attention to interoperability would help ensure that these efforts cohere rather than fragment, and that all Member States - particularly smaller economies and emerging markets - can navigate the landscape and align with common reference points. The Dialogue offers a timely and inclusive setting in which Member States and stakeholders may exchange experiences and explore practical pathways toward compatibility, mutual recognition, and shared standards.

The proposed thematic clusters are comprehensive, but several cross-cutting issues warrant explicit attention to ensure the Dialogue's outcomes are operationally meaningful. The AI-energy-climate nexus. The International Energy Agency projects that global data-centre electricity consumption will roughly double by 2030, with AI-focused facilities tripling in the same period. AI governance is increasingly inseparable from energy security, water use, and the credibility of net-zero commitments. This dimension cuts across capacity, sustainability, and trustworthiness, but is not yet anchored as a thematic priority. Capital markets and fiduciary accountability. AI now informs trillions in investment decisions, yet recent analyses indicate that only a small share of large public companies disclose any board-level oversight of AI. As AI-related shareholder resolutions multiply, the financial sector is becoming a de facto governance venue - without the international coordination such a role requires. The Dialogue could helpfully connect with the work of UN-supported initiatives such as the Principles for Responsible Investment. Frontier and agentic systems. Current frameworks tend to address "AI systems" generically. The MIT AI Risk Initiative's April 2026 mapping of over 1,000 governance documents found that frontier, foundation, and emerging multi-agent systems receive disproportionately little dedicated attention. The Dialogue's first session would benefit from acknowledging this gap. Independent evaluation infrastructure. Shared evaluation methods, incident-reporting channels, and red-team access for trusted third parties remain underdeveloped at the international level. Without such infrastructure, principles remain aspirational. Meaningful inclusion of underrepresented stakeholders. Indigenous communities, smaller language groups, persons with disabilities, and youth are frequently invoked but unevenly represented in deliberations. Concrete participation modalities - beyond statements of intent - would strengthen the legitimacy of outcomes. These issues are interconnected. Surfacing them explicitly would strengthen the Dialogue's coherence and its capacity to translate universal participation into durable cooperation.

• Speaking from the perspective of the sustainable investing and AI governance technology sector, governance gaps in three of the four thematic areas are already producing measurable downstream effects.Safe, secure, and trustworthy AI. Documented AI incidents reached 362 in 2025, nearly double the 2024 figure (Stanford 2026 AI Index). Yet there is no universally recognised incident-reporting channel, no mutually accepted evaluation methodology, and no shared red-line architecture across jurisdictions. For investors and asset managers, this translates into material due-diligence uncertainty — portfolio companies operating across multiple regulatory regimes face inconsistent expectations, and material AI-related risks are difficult to assess on a comparable basis. The Foundation Model Transparency Index has fallen from 58 to 40, with the most capable models disclosing the least. This opacity is not only a technical concern
• it is a fiduciary one.Societal, economic, ethical, cultural, linguistic, and technical impacts. The IMF's Mind the Gap working paper projects that AI's growth impact in advanced economies could be more than double that in low-income countries. UNDP's Asia-Pacific Next Great Divergence report finds AI usage near 5% in many low-income economies versus two in three in some high-income economies. For sustainable investors with mandates aligned to the 2030 Agenda for Sustainable Development, this divergence threatens decades of convergence gains and complicates ESG portfolio construction across emerging markets.Protection and promotion of human rights. International human rights law provides a stable normative anchor, yet its operational integration into AI procurement, deployment, and corporate governance remains uneven. PRI's March 2026 analysis cites ISS Governance research showing that only around 8% of more than 3,000 US-listed companies disclose any board-level oversight of AI. AI-related shareholder resolutions nearly tripled between 2023 and 2025, reflecting a market response to this oversight gap. Investors are escalating in the absence of harmonised guidance
• the cost of that absence is fragmented, sometimes contradictory, expectations placed on portfolio companies.Interoperability of governance approaches. The OECD AI Principles, the G7 Hiroshima Process, the Council of Europe Framework Convention, the EU AI Act, and a growing body of national legislation are creating overlapping but non-aligned obligations. The MIT AI Risk Initiative's April 2026 Mapping the AI Governance Landscape analysed over 1,000 governance documents and found that socioeconomic risks receive comparatively little attention. For multinationally exposed asset managers and the technology providers that serve them, the absence of a coherent reference architecture imposes real compliance and operational costs.These gaps are interconnected. Capital markets are increasingly playing a de facto governance role — but they are doing so in the absence of the international coordination such a role requires. The Dialogue's distinctive contribution would be to provide that coordinating layer.

The Dialogue's distinctive contribution lies not in producing additional principles, but in providing the coordinating architecture the existing landscape lacks. Several specific roles would advance international cooperation meaningfully.A coherent reference layer across existing initiatives. The OECD AI Principles, the G7 Hiroshima Process, the Council of Europe Framework Convention, the European Union AI Act, and a growing range of national frameworks already provide substantive guidance. The Dialogue can offer an honest mapping of where these regimes converge, diverge, and create compliance friction — giving smaller economies and emerging markets a navigable reference rather than forcing them to choose between competing rulebooks.A structured interface with the Independent International Scientific Panel. The Dialogue's authority depends on evidence rather than aspiration. Establishing how the Panel's findings systematically inform the Dialogue's agenda — including through commissioned assessments on specific issues such as frontier risks, AI in critical infrastructure, or AI-driven energy demand — would distinguish the Dialogue from purely declaratory forums.A genuine multistakeholder venue, not a parallel-monologue venue. Civil society, technical experts, industry, and underrepresented stakeholders should shape conclusions rather than observe them. Structured, interactive sessions whose outputs visibly reflect non-state input would establish the Dialogue's legitimacy. The April 2026 Draft Note's clustering is a strong baseline; sustained attention to participation modalities will determine whether universal participation translates into substantive participation.A platform for cross-cutting issues without natural homes. The AI–energy–climate nexus, capital markets and fiduciary accountability, frontier and agentic systems, and independent evaluation infrastructure all sit across — rather than within — existing initiatives. The Dialogue is uniquely positioned to convene them.Ultimately, the Dialogue's success will depend on whether it is treated as a coordinating instrument rather than a competing one. That is the role it can credibly fill.

The Dialogue should build upon and connect with initiatives across four governance layers, while contributing what none of them alone can offer. Intergovernmental and regional frameworks. The OECD AI Principles, the G7 Hiroshima Process, the Council of Europe Framework Convention on AI, the European Union AI Act, the African Union Continental AI Strategy, and the Bletchley–Seoul–Paris AI Safety Summit series each provide substantive substance. The Dialogue's added value is universal participation: extending the conversation to the roughly 118 countries currently outside any major AI governance initiative. UN-system mechanisms. The Independent International Scientific Panel on AI, the Global Digital Compact, the WSIS+20 process, UNESCO's Recommendation on the Ethics of AI, the Office of the UN High Commissioner for Human Rights' B-Tech Project, and the ITU's AI for Good platform all carry directly relevant mandates. The Dialogue can serve as the integrating layer across these efforts, ensuring coherence rather than duplication. Expert and standards bodies. ISO/IEC JTC 1/SC 42, the IEEE Global Initiative on Ethics of Autonomous and Intelligent Systems, NIST's AI Risk Management Framework, and the AI Safety Institute network (UK, US, Japan, Singapore, EU, and others) provide the technical infrastructure that political dialogue depends on. The Dialogue can help translate technical consensus into intergovernmental reference points. Stakeholder and capital-markets initiatives. The UN-supported Principles for Responsible Investment, the Global Partnership on AI, the Partnership on AI, and civil society networks contributing through RightsCon and the Internet Governance Forum bring perspectives — particularly from finance, civil society, and the technical community — that intergovernmental forums alone cannot capture. The Dialogue's distinctive added value is convening authority across all four layers simultaneously. No other forum can credibly do so under universal membership. That convening role, executed well, is the cooperation infrastructure the field currently lacks.

Effective multistakeholder participation requires both differentiated contribution pathways and a structure designed to integrate them.Stakeholder contributions.Member States can anchor the Dialogue's legitimacy by articulating national priorities, sharing domestic governance experience, and committing to implementation pathways beyond the room.Civil society organisations — including human rights, digital rights, and consumer protection groups — can surface lived impacts, hold all participants accountable to rights-based commitments, and represent communities not directly seated at the table.The technical and scientific community can translate evidence into policy-ready language, support the work of the Independent International Scientific Panel, and advise on evaluation methodologies, incident reporting, and red-team practice.Industry and the private sector can contribute deployment realities, technical roadmaps, and operational constraints — alongside transparency on incidents, risks, and governance practices.The financial sector and capital-markets actors, including UN-supported initiatives such as the Principles for Responsible Investment, can connect AI governance to fiduciary duty, investor stewardship, and the climate-transition agenda.Underrepresented stakeholders — Indigenous communities, smaller language groups, persons with disabilities, youth, and Global South participants — should be resourced rather than only invited.Format and structure recommendations. Reduce plenary time; expand interactive sessions. Plenaries establish positions; thematic working sessions produce substance. The April 2026 Draft Note's clustering is a strong baseline; the balance should tilt further toward structured interaction. Cross-stakeholder composition within sessions. Avoid replicating sectoral silos. Each thematic session should mix governments, civil society, technical experts, and industry by design. Visible synthesis of non-state input. Outputs must demonstrably reflect multistakeholder contribution to establish legitimacy. Pre-Dialogue and intersessional engagement. Continuous engagement — through regional consultations, written submissions, and online platforms — would extend participation to those unable to travel to Geneva. Capacity and travel support for Global South and civil society participants is essential.

The Global South beyond major economies. Voices from least developed countries, small island developing States, and landlocked developing countries are frequently invoked but unevenly resourced to participate. Inclusion mechanism: dedicated travel and connectivity funding, regional pre-Dialogue consultations, and capacity-building support drawn from the proposed Global Fund for AI capacity development.Indigenous peoples and traditional knowledge holders. AI training data and governance frameworks rarely reflect Indigenous epistemologies, languages, or data sovereignty principles. Inclusion mechanism: formal consultation channels aligned with the UN Declaration on the Rights of Indigenous Peoples, and dedicated session time on Indigenous data sovereignty.Speakers of low-resource languages. Today's leading models are disproportionately trained on dominant languages, embedding systemic bias against the linguistic majority of humanity. Inclusion mechanism: multilingual deliberation infrastructure, support for evaluation in low-resource languages, and engagement with regional language consortia.Persons with disabilities. AI systems are increasingly mediating access to essential services — yet accessibility considerations rarely shape design, evaluation, or procurement standards. Inclusion mechanism: sustained participation by disability-rights organisations, alignment with the Convention on the Rights of Persons with Disabilities, and accessibility audits of the Dialogue itself.Youth and future generations. Those who will inherit the systems being built today have limited formal standing in governance deliberations. Inclusion mechanism: dedicated youth delegate channels, intergenerational sessions, and partnership with UN youth bodies.Workers and labour representatives. Those most directly exposed to AI-driven labour-market change — including informal-sector workers — are rarely seated alongside the firms deploying these systems. Inclusion mechanism: engagement with the International Labour Organization, trade unions, and worker-led civil society.Genuine inclusion requires resourced, structural participation — not symbolic invitation. The Dialogue's legitimacy will depend on closing this distance.

Several engagement formats would foster substantive, dynamic participation beyond the limits of conventional plenary deliberation.1. Cross-stakeholder thematic working sessions. Each thematic cluster should be deliberated in mixed-composition working groups — Member States, civil society, technical experts, industry, and capital-markets actors seated together by design rather than in sequence. This breaks the parallel-monologue pattern that limits most multilateral forums.2. Evidence briefings from the Independent International Scientific Panel. Short, structured briefings opening each thematic session would anchor deliberation in shared evidence rather than competing assertions. The Panel's first commissioned assessments — on frontier risks, AI in critical infrastructure, or AI-driven energy demand — could frame the agenda directly.3. Scenario-based deliberation. Concrete scenarios — a cross-border AI incident, a frontier model release, an algorithmic decision affecting essential services — surface where existing frameworks converge and diverge in ways principles-based discussion cannot. This format has been used effectively in other multilateral processes, including the climate and biosecurity domains.4. Hybrid and asynchronous participation channels. Continuous online platforms, regional pre-Dialogue consultations, and structured written contributions extend participation to those unable to travel to Geneva. Hybrid is a substantive design choice, not a logistical convenience.5. Stakeholder-led satellite sessions. Civil society, the technical community, and the financial sector should be empowered to convene their own sessions on the margins, with outputs systematically fed into the main deliberation. This preserves multistakeholder authorship rather than treating non-state actors as observers.6. Visible synthesis and closing accountability. A structured synthesis at the close of each session — capturing convergence, divergence, and outstanding questions — would make multistakeholder input visible in the record and create continuity into the May 2027 New York session.Format is not procedural detail. It is the substantive choice that determines whether universal participation produces durable cooperation.

The international AI governance landscape already contains operational examples worth building upon. Four categories illustrate the range of credible approaches.Policy and regulatory frameworks. The European Union AI Act offers the most developed risk-based regulatory architecture, with tiered obligations scaling to system risk. The Council of Europe Framework Convention on AI provides the first binding international treaty grounded in human rights, democracy, and the rule of law. Singapore's AI Verify offers a voluntary testing framework that bridges principles and technical evaluation. Together, these illustrate that proportionate, rights-based regulation and innovation are mutually reinforcing rather than opposed.Technical and evaluation infrastructure. The NIST AI Risk Management Framework, ISO/IEC 42001 (the first management-system standard for AI), and the AI Safety Institute network - including the United Kingdom, United States, Japan, Singapore, the European Union, and others - provide the evaluation methodologies, red-team capabilities, and incident-response practices that political dialogue depends on. Their continued cross-jurisdictional coordination is essential.Multistakeholder platforms. The OECD AI Policy Observatory, the Global Partnership on AI, the ITU's AI for Good platform, and UNESCO's Recommendation on the Ethics of AI demonstrate sustained convening across governments, technical communities, and civil society. The Office of the UN High Commissioner for Human Rights' B-Tech Project translates the UN Guiding Principles on Business and Human Rights into operational guidance for technology companies.Capital-markets and investor mechanisms. The UN-supported Principles for Responsible Investment, the OECD due diligence guidance for responsible AI (2026), and the World Benchmarking Alliance's Ethical AI benchmark connect AI governance to investor stewardship and corporate accountability - increasingly material as AI shapes trillions in investment decisions.These examples demonstrate that governance instruments exist; the gap is coordination, mutual recognition, and consistent application across jurisdictions.