Golden Occasions, LLC

The first Global Dialogue's success should be measured against three outcomes. First, the Dialogue should produce a co-chairs' summary that captures the actual diversity of submissions received — including perspectives from small organizations, the Global South, and stakeholders outside the conventional governance forums. Aggregation that flattens unusual perspectives into the dominant frames would represent an early structural failure. Operative paragraph 7 of A/RES/79/325 places significant weight on the co-chairs' summary as input to subsequent intergovernmental work; ensuring that summary preserves substantive diversity is foundational. Second, the Dialogue should establish working relationships between the Independent International Scientific Panel and the multistakeholder participants that allow scientific assessment and stakeholder perspective to inform each other iteratively. The Panel's annual report should not be received passively at the Dialogue; it should be engaged with substantively, with stakeholder responses feeding back into the Panel's subsequent work. Third, the Dialogue should clarify how its outputs interoperate with existing AI governance forums — the OECD, Council of Europe Framework Convention on AI, IGF, and standards bodies including ITU. Operative paragraph 8 envisions common understandings as input to the Global Digital Compact review; those understandings will only carry weight if the Dialogue's relationship to existing instruments is structurally clear rather than competitive or duplicative. Beyond these structural outcomes, success should be measured by whether the Dialogue surfaces concrete questions for the second session that the broader governance landscape cannot easily answer elsewhere. The Dialogue's distinctive value is its UN-anchored multistakeholder convening capacity. Success means using that capacity to advance questions that benefit specifically from that convening — not duplicating work that other forums are positioned to do faster. A successful first Dialogue is one whose second session has a sharper agenda than its first.

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

Our four selections reflect the thematic territory our written submission addresses directly and where we believe small-firm and organizational-scale perspectives warrant greater weight in the Dialogue's deliberations. Interoperability of governance approaches is our first priority because the multilateral governance conversation has thus far focused on harmonization between national regulatory regimes and between international forums. The interoperability problem at the organizational level - between firm-level governance and multilateral norms - is substantially underweighted, and it is the layer at which most economic activity actually happens. Without organizational-level interoperability, multilateral instruments rely on enforcement bandwidth that does not exist at scale. Protection and promotion of human rights is our second priority because human rights frameworks are most resilient where they are mirrored by structural protections within the systems that interact with them. Governance architectures that treat AI involvement substantively - rather than as a peripheral compliance matter - provide stronger structural conditions for human rights commitments to hold under operational pressure. This is not a substitute for international human rights law; it is a complement to it at the organizational layer. Transparency, accountability, and human oversight is our third priority because oversight requires a procedural substrate. Where AI involvement is substantive, oversight that depends on operator vigilance alone is fragile. Constitutional governance architecture provides the structural substrate that makes oversight technically meaningful rather than ceremonial. Safe, secure and trustworthy AI is our fourth priority because the documentation discipline our submission addresses is a precondition for the trust the theme names. Trustworthiness requires legibility, and legibility requires documentation. This priority is foundational to the others. These four themes are not independent; they reinforce each other. Treating them as a coherent cluster, rather than as separate workstreams, would strengthen the Dialogue's structural coherence.

Three cross-cutting issues warrant the Dialogue's attention beyond the seven enumerated themes. First, the organizational layer of AI governance is structurally underrepresented in current discourse. Existing thematic areas address national regulation, technical standards, and human rights frameworks at the multilateral level - but the layer at which most AI deployment is actually governed is the firm level, and there is no thematic area dedicated to how organizational governance interoperates with the others. This is the specific gap our written submission addresses, and it warrants recognition as cross-cutting rather than as a sub-element of interoperability. Second, the operating ethic governing human-AI collaboration in organizations whose work depends on substantive AI involvement is not addressed by any of the seven themes. Current governance discourse oscillates between AI-as-tool framings and AI-as-autonomous-agent framings; neither describes the operational reality of substantive AI-human collaboration. The Dialogue's second session, in our view, should consider whether a third frame - AI as partner under consensual governance - warrants serious examination as an operating ethic for organizations and, by extension, for the multilateral instruments that interoperate with them. Third, the capacity gap for small organizations and Global South participants in AI governance is real and structural. Existing capacity-building frameworks focus on technical literacy and regulatory familiarity; the procedural and documentation capacities that make organizational governance interoperable with multilateral norms receive less attention. Without addressing this gap, AI governance frameworks risk being effective only for large organizations with resources to participate in the conversation - leaving the substantial majority of AI deployment ungoverned by the instruments the Dialogue produces. These three cross-cutting issues are not independent of the seven themes; they are perspectives that, if held alongside the themes, sharpen the Dialogue's architectural coherence.

The governance gaps most visible from our vantage point — a small AI-supported governance research practice in the United States, operating in the small-organization segment of the AI deployment landscape — concern the absence of accessible, interoperable frameworks for organizational AI governance at scale below large enterprise. Three challenges stand out. The first is fragmentation of available guidance. Small organizations adopting AI substantively encounter a landscape of national regulatory developments, sector-specific compliance frameworks, technical standards, and aspirational principles — none of which are designed to interoperate at the organizational scale where small firms actually operate. The result is governance fragmentation that produces either over-investment in compliance theater or under-investment in actual governance discipline. Neither outcome serves trustworthy AI deployment. The second is the absence of a structural framework for substantive AI involvement. Existing governance frameworks generally assume AI as a deployed technology — an object subject to regulation. They are less well-suited to organizations whose work depends on substantive AI collaboration, where the governance instrument needs to address how human and AI participants interact within the work, not only how AI outputs are bounded. The third is the documentation capacity gap. Multilateral instruments, national regulatory frameworks, and sector standards all rely on organizational documentation as the substrate for compliance, audit, and interoperability. Small organizations adopting AI lack accessible guidance on what governance documentation should look like, how it should be ratified, and how it interoperates with the broader instruments. The opportunity these gaps reveal is real: the multilateral governance conversation can address them through capacity-building targeted at the organizational layer — the layer where most economic activity happens and where governance instruments either become operationally meaningful or remain abstract.

The Dialogue's distinctive value lies in three roles that other forums are less well-positioned to fill. The first is convening across stakeholder categories that rarely engage substantively in the same room. National governments, international organizations, civil society, technical communities, academia, and the private sector all participate in AI governance conversations — but most existing forums weight one or two categories heavily and treat the others as peripheral. The Dialogue's UN-anchored multistakeholder mandate, established under operative paragraph 4 of A/RES/79/325, positions it to host genuine cross-stakeholder engagement rather than category-dominant convening. This matters because AI governance failures often occur at the seams between stakeholder categories rather than within any one of them. The second is integrating scientific assessment with governance deliberation iteratively. The Independent International Scientific Panel produces evidence-based assessments; the Dialogue receives those assessments and considers their implications for governance. The interaction between the two — assessment informing deliberation, deliberation surfacing questions for subsequent assessment — is the Dialogue's structural opportunity. Most existing forums separate these functions; the Dialogue joins them. The third is establishing common understandings as input to the Global Digital Compact review. Operative paragraph 8 of A/RES/79/325 envisions the second Dialogue's intergovernmental consultations producing common understandings of priority areas. This is the Dialogue's most consequential potential output. Common understandings are the substrate for subsequent multilateral work; their absence has been a structural weakness in AI governance discourse to date. The Dialogue should resist two failure modes: becoming another forum that produces principles without implementation pathways, and becoming a procedural venue whose outputs duplicate work other forums are positioned to do faster. Its distinctive value is in roles that genuinely require UN-anchored multistakeholder convening — and those are the roles it should prioritize.

The Dialogue should build upon and connect with several existing instruments and forums rather than duplicate their work. Within the UN system, the UNESCO Recommendation on the Ethics of AI (2021) provides a comprehensive normative framework already adopted by 193 member states; the Dialogue should treat it as foundational rather than supplementary. ITU's technical standards portfolio (over 100 published AI-related standards as of 2024) addresses interoperability at the technical layer; the Dialogue's interoperability work at the governance layer should integrate with rather than parallel ITU's. The Inter-Agency Working Group on Artificial Intelligence, co-led by ITU and UNESCO, has already mapped the institutional landscape in CEB/2024/1/Add.1; the Dialogue should treat that mapping as its starting point. OHCHR's work on AI and human rights provides the rights framework that should anchor the Dialogue's deliberations on thematic area (e). Outside the UN system, the OECD AI Principles, the Council of Europe Framework Convention on Artificial Intelligence (in force as of September 2024), the Internet Governance Forum, and standards bodies including ISO, IEC, and IEEE all contribute distinct functions. The Dialogue's added value is not in replicating these — they are well-positioned in their respective domains — but in providing a UN-anchored convening space where their work can be integrated into a coherent multilateral conversation. The Dialogue's distinctive contribution is the integration itself. Existing instruments and forums each address pieces of the governance landscape; none of them is positioned to convene the whole picture under UN authority with the Charter and international human rights law as anchoring instruments. That convening role — combined with the Independent International Scientific Panel's assessment function — is the Dialogue's structural opportunity. Building upon existing work while occupying the integrative role it is uniquely positioned to fill is the right approach.

Different stakeholder categories contribute most effectively when the Dialogue's structure matches the function each category is positioned to perform. Member states are best positioned to contribute on national experience with AI governance instruments — what has been tried, what has worked, what has produced unintended consequences. The Dialogue should structure space for this contribution that goes beyond formal statements: comparative national-experience panels, structured peer-to-peer exchanges, and Member State submissions to the Dialogue's accumulating knowledge base. International organizations are positioned to contribute on existing instruments and institutional capacity. The Inter-Agency Working Group on Artificial Intelligence's mapping in CEB/2024/1/Add.1 should be a living input, updated annually and engaged with substantively rather than received once. Civil society and human rights organizations are positioned to contribute on the lived effects of AI deployment on affected communities — the data without which governance abstractions cannot be calibrated. The Dialogue should structure direct engagement formats for this contribution rather than relegating civil society to observer status. Technical communities and standards bodies are positioned to contribute on what is technically possible, what is technically required, and where the gap between governance aspiration and technical reality currently lies. The Dialogue's relationship with ITU, ISO, IEC, IEEE, and adjacent bodies should be structured for substantive technical input. Academia and research institutions are positioned to contribute long-horizon analysis and frameworks for thinking about emerging issues that operational stakeholders may be too close to see clearly. Private sector and small organizations are positioned to contribute on operational reality — what AI governance looks like at the layer where most deployment happens. Format recommendation: structured thematic working groups with clear stakeholder-category roles, feeding into plenary deliberation that integrates rather than aggregates. The Dialogue's value is in integration; the format should support it.

Several categories of perspective remain structurally underrepresented in global AI governance discussions. Global South participation, while formally encouraged, often faces structural barriers that go beyond travel costs — including time-zone asymmetries in virtual participation, language access gaps in working documents, and the convention of treating Global North governance frameworks as the default reference points to which Global South perspectives respond. Genuine inclusion requires structural reorientation, not only logistical support. Operative paragraph 11 of A/RES/79/325 addresses travel support; the deeper inclusion work goes beyond it. Affected communities — those whose lives are most directly shaped by AI deployment in employment, social services, criminal justice, healthcare, and education — are routinely discussed in AI governance conversations but rarely included as principal participants. The Dialogue should structure direct engagement with affected communities rather than treating their experience as data filtered through advocacy organizations. Small organizations and individual practitioners working on AI governance outside large institutional structures contribute perspectives that institutional submissions cannot. The Dialogue's openness to individual and small-organization submissions, established in the current call for written inputs, is a structural strength that should be preserved and extended into participation formats beyond written submission. Women, gender-diverse persons, Indigenous communities, persons with disabilities, and youth — communities whose underrepresentation is well-documented across multilateral governance generally and AI governance specifically — require dedicated participation pathways that go beyond inclusion as discussion topics. Workers in the AI value chain — including those performing data labeling, content moderation, and other often-invisible labor — represent a category of participant whose voice is structurally absent from governance conversations conducted by organizations whose work depends on them. Inclusion mechanisms: dedicated participation pathways, language access investment, structured engagement with affected communities, and explicit attention to the difference between formal participation availability and substantive participation reality.

Three formats warrant the Dialogue's consideration. Working groups with persistent membership across sessions. The Dialogue currently structures around annual sessions, with the second to alternate between Geneva and New York. Substantive work on complex governance questions benefits from continuity that single-session deliberations cannot provide. Working groups operating between sessions — drawing on the Independent International Scientific Panel's assessments, Member State and stakeholder inputs, and accumulated Dialogue outputs — would allow depth that a session-only structure cannot achieve. Structured engagement between the Independent Scientific Panel and stakeholders. Operative paragraph 1(d) of A/RES/79/325 envisions the Panel providing updates "up to twice a year" through interactive dialogue with the General Assembly. The Panel's interaction with Dialogue stakeholders is less specified. Formats that allow stakeholder questions to inform the Panel's assessment work, and Panel findings to inform stakeholder deliberation, would strengthen both functions. Case-based deliberation rather than principle-based deliberation alone. Multilateral governance discussions often default to principle articulation — important but insufficient. Dialogue formats that work through specific governance challenges in detail (a specific sector's compliance problem, a specific jurisdictional interoperability issue, a specific deployed system's oversight gap) produce learning that principle-level discussion cannot. Case-based deliberation also builds shared vocabulary across stakeholder categories that abstract discussion does not. A note on virtual and hybrid participation: the Dialogue should structure for hybrid participation as a first-class format rather than as accommodation. Geneva and New York convening is appropriate for the principal sessions; the Dialogue's broader engagement should be designed for participants who cannot travel, with attention to time-zone equity, interpretation availability, and substantive participation parity rather than observer status. The principle underlying all three recommendations: format follows function. The Dialogue's distinctive functions should drive its structural choices.

Several existing approaches warrant consideration as the Dialogue advances its work. The UNESCO Recommendation on the Ethics of AI (2021) and its accompanying Readiness Assessment Methodology demonstrate that comprehensive ethical frameworks can be adopted through multistakeholder process and operationalized through assessment tools that meet member states where they are. The methodology's deployment across more than 50 countries provides a body of comparative experience that subsequent governance work should build on rather than duplicate. ITU's standards development process, particularly its focus group mechanism for pre-standards work in emerging technology areas, demonstrates that agile, multistakeholder technical norm-setting is achievable within UN system structures. The ability to develop specifications rapidly while maintaining consensus discipline is directly relevant to AI governance, where the pace of technological change strains conventional standard-setting timelines. Sector-specific governance approaches, including WHO's regulatory considerations on AI for health and the UNICRI/INTERPOL framework on responsible limits for facial recognition in law enforcement, demonstrate that governance precision improves substantially when frameworks are tailored to specific use contexts rather than developed at maximum generality. Anticipatory governance practices developed by United Nations agencies in adjacent domains - WFP's work on anticipating food crises, UNHCR's predictive analytics for displacement, ECLAC's foresight methodologies - demonstrate institutional capacity for governance that addresses emerging risks before they materialize fully. These practices warrant adaptation for AI governance contexts. Multistakeholder convening platforms, including the AI for Good Global Summit and the Internet Governance Forum, demonstrate that productive engagement across stakeholder categories is achievable when convening structures are designed for it rather than retrofitted. The pattern across these examples: governance approaches succeed when they combine clear normative anchoring, multistakeholder development process, sector-specific tailoring where appropriate, and operational tools that translate principle into practice.