Value Creating Conversations

The first Global Dialogue on AI Governance will be a success if it produces one thing that most international governance processes fail to produce: a measurement standard that organizations can actually use. We have no shortage of AI principles. Every major institution has published them. What we lack is the infrastructure to translate principles into accountability - a way to measure whether AI systems are honoring the commitments those principles describe. Success means the Geneva dialogue moves the international community from principles to metrics. Specifically, it means agreement on what should be measured, at what level of granularity, and how those measurements should be disclosed publicly. I would add one dimension that current frameworks consistently underweight: the individual. Every existing AI governance metric measures populations such asdemographic parity, equalized odds, aggregate accuracy. None measures whether a specific human being was treated with the respect their inherent worth demands. Until governance frameworks operate at the individual level, the people most harmed by AI systems - those whose circumstances do not fit neatly into demographic categories will remain invisible to the very frameworks designed to protect them. Success also means the dialogue reflects the full geography of AI's impact. The communities most affected by AI governance failures are not in Geneva. They are in Mumbai, Arusha, Nairobi, and Jakarta. Field research across these contexts reveals patterns of exclusion, proxy-based discrimination, and dignity violation that aggregate data cannot see. The dialogue must create space for that evidence. Finally, success means the dialogue produces a timeline and not just recommendations. Principles without deadlines do not change behavior. A governance standard with a compliance horizon does. The conversation starts in Geneva. The measurement starts after.

• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• AI capacity-building

My priorities reflect twenty years of enterprise transformation experience and field research across Mumbai, Arusha, and Tokyo. Protection and promotion of human rights is my primary priority. I have coined Dignity Economics - the measurement of what AI deployment costs in human terms. Human dignity is absent from every AI ROI conversation. The dialogue must establish it as a governance metric, not merely a principle. Transparency and accountability must extend upstream to decisions made before deployment, not only after harm occurs. Annual public disclosure obligations, modeled on sustainability reporting, would create accountability that principles alone cannot. Social and cultural implications are urgent because aggregate data masks what field research reveals. AI costs concentrate in communities with the least market power and the same communities least represented in governance conversations. AI capacity-building must be a precondition of deployment, not an afterthought. Governance capability is not evenly distributed.

These two cross-cutting issues are absent from the listed themes currently that can enhance the proposed agenda further. First, individual-level measurement. Every existing AI governance metric measures populations. None measures whether a specific human being was treated with the respect their inherent worth demands. The dialogue must recognize individual-level dignity measurement as an emerging governance priority. Second, the governance speed gap. AI systems in high-stakes domains operate faster than accountability mechanisms can respond. By the time a pattern of harm is identified, thousands of people may already be affected. The answer is not slower AI. It is stronger pre-deployment standards that govern systems before they move, not after. Both issues cut across every theme in the resolution. They are not additions to the agenda. They are the infrastructure without which the agenda cannot be implemented.

My work spans civil society globally, with field research conducted across Mumbai, Arusha, and Tokyo - three cities representing distinct relationships with AI governance and its consequences. These observations inform the LIGHTHOUSE™ Initiative, a ten-principle AI governance framework I developed specifically for practitioners and leaders navigating AI adoption without enterprise-level governance resources. The most significant challenge I observe consistently across regions is the gap between AI deployment speed and governance readiness. The governance conversation is happening in Geneva and Brussels. The deployment is happening in Mumbai, Arusha, and in the offices of everyday practitioners around the world. Those conversations are not yet connected. In India, my conversations revealed three distinct governance realities. A woman in banking cannot access AI because her organization has not adopted it where the governance gap is not misuse but exclusion. A woman in media uses AI actively but without any organizational governance framework in place where the tools arrived before the rules. A woman working in the judiciary has a four-hour daily commute and no bandwidth to learn AI - the capacity-building gap is not about willingness but about the conditions that make learning possible. Three women. Three different governance failures. None of them captured by aggregate data. In Tanzania, a conversation with women tour operators which was facilitated by a male colleague revealed a structural barrier that aggregate data cannot capture. The most fluent voice belonged to the male facilitator, already using AI meaningfully for tourism - managing packages, communicating globally, bridging language barriers. He had not been told that business ideas entered through a free consumer tool may be used to train future models. The women were largely silent, not from disinterest but from connectivity failures, cultural confidence barriers, and schedules running Monday through Saturday. One woman broke through to request a one-on-one session. Her first question: is this free? The governance gap in Tanzania is not resistance to AI. It is the absence of the infrastructure - connectivity, safety education, culturally appropriate access, and time - that makes responsible adoption possible for women specifically. And it is the structural dynamic in which women's voices about AI are mediated and represented by others at every level, from a Zoom call in Arusha to a policy dialogue in Geneva. Closer to home, I recently spoke with a tax professional using a free large language model to handle complex client filings; unaware that her clients' personal financial data may be used as training data by the model provider. She submitted an LLM-generated report containing sensitive client information to her manager openly, through normal channels. Her manager may be proud that the team is embracing AI. The practitioner feels good that she is keeping pace with the industry. Neither knows that in that moment of mutual progress, a client's most sensitive financial data quietly left a controlled environment without their knowledge or consent. This is not shadow AI. This is ungoverned AI operating in plain sight which is more dangerous because there is nothing to trigger investigation. Across all three contexts the same conclusion holds: governance frameworks must reach the individual practitioner at the moment of decision, not as a compliance document that exists somewhere in a policy folder, but as accessible, culturally appropriate, actionable knowledge. That is precisely the gap the LIGHTHOUSE™ Initiative was built to address.

The AI Dialogue arrives at a moment when the governance gap is not a lack of principles - it is a lack of infrastructure to translate principles into accountability. Every major institution has published AI ethics frameworks. What does not yet exist is the measurement standard, the verification mechanism, and the disclosure obligation that would make those principles enforceable at the individual level. The dialogue can play five specific roles that no other international body is currently positioned to play. First, establish a common measurement baseline. International cooperation on AI governance is currently hampered by incompatible frameworks. The EU AI Act, the NIST AI RMF, the OECD principles, and emerging national standards speak different languages and professional certification bodies add further fragmentation, often teaching governance concepts misaligned with any recognized standard. The dialogue can broker agreement on what governance must contain and what should be measured and not how each jurisdiction implements it, but what the shared floor looks like. Second, govern the certifiers. The proliferation of AI certifications has diluted the meaning of certification itself. When the people teaching governance do not themselves understand internationally recognized frameworks, that misunderstanding propagates downstream at scale. The dialogue must establish standards for who can certify, what must be taught, and how certification content must align with recognized frameworks. Certification without quality governance is not a credential. It is a liability. Third, mandate baseline governance literacy for business owners. Countries should require that organization heads demonstrate basic AI governance literacy as a condition of annual business registration - similar to tax filing or health and safety compliance. Fine for non-compliance. Make it structural, not voluntary. The practitioner deploying AI without understanding basic governance obligations is not an edge case. She is the norm. Mandatory baseline literacy changes that norm systematically. Fourth, extend governance reach to practitioners. Current frameworks operate at the organizational and regulatory level. They do not reach the daily decision-maker navigating AI without governance infrastructure. The dialogue can establish that practitioner-level governance capacity is a precondition of responsible deployment. Fifth, ensure the Global South shapes the standard, not inherits it. The communities most affected by AI governance failures are not leading the governance conversation. The dialogue must create structural mechanisms, not symbolic gestures for Global South voices to shape the standards they will be governed by. Principles without measurement are intentions. Certification without standards is noise. Governance without mandate is aspiration. The dialogue can turn all three into infrastructure.

The AI Dialogue arrives at a moment when the challenge is not a shortage of frameworks - it is a shortage of collaboration between them. The OECD AI Principles, the NIST AI RMF, the EU AI Act, the Global Digital Compact commitments, and dozens of national standards exist in parallel, speaking different languages, reaching different audiences, and leaving practitioners navigating incompatible requirements without a shared floor to stand on. The divergence is not only between nations. It runs through the very nature of governance itself. Some frameworks are binding with enforcement mechanisms and financial penalties. Others are voluntary, relying on cooperation and reputational pressure with no consequences for non-compliance. Some countries have enacted AI-specific legislation. Others are governing AI through existing data protection, consumer rights, and civil liability laws while new frameworks are debated. The same technology. Fundamentally different governance postures. Without a common floor, the people least protected by any single framework fall through all of them. The dialogue's unique value is not simply producing another framework or assuming that what exists is sufficient. It is convening the authority to collate what exists, assess who is included and who is not, and let that finding guide what comes next. Collaboration is the method. Inclusion is the standard. Flexibility to adapt as AI evolves is the requirement. Concretely, the dialogue can advance five collaboration mechanisms: First, align existing frameworks around a shared measurement floor. Not a new framework, but agreement on what every framework must measure. Second, coordinate certification standards so they reflect agreed frameworks rather than competing with them. The proliferation of AI certifications is propagating inconsistent governance understanding at scale. Third, connect governance obligations to existing business registration infrastructure. Baseline AI governance literacy as a condition of annual registration is enforceable without creating new bureaucracy. Fourth, bridge the gap between international frameworks and the daily decision-maker. Current frameworks do not reach the tax professional, the tour operator, or the judiciary worker navigating AI without governance infrastructure. That is where harm concentrates. Fifth, ensure the collaboration is genuinely multilateral, not leaning towards west. The communities most affected by AI governance failures must shape the standards they will be governed by - not inherit them. The frameworks exist. The principles are largely agreed. Civil society governance initiatives including the LIGHTHOUSE™ Initiatives, presented to the United Nations Office for Digital and Emerging Technologies in November 2025 represent the practitioner-level infrastructure that top-down frameworks cannot reach alone. The dialogue should create formal mechanisms for civil society frameworks to feed into and be recognized by international standards processes. What is missing is the honest assessment of who current frameworks exclude, and the political will to close that gap before the next generation of AI systems makes it permanent.

The AI Dialogue will only be as inclusive as the mechanisms it creates for participation. Recommendations must begin with an honest accounting of who is absent, and why. Field research across Mumbai, Arusha, and Tokyo reveals a consistent pattern: those most affected by AI governance failures are least represented in forums designing standards. A judiciary worker cannot travel to Geneva. A cooperative in Tanzania cannot navigate frameworks built for institutional actors. When women, transgender, non-binary, and gender-diverse people are absent, systems encode their invisibility into the world they will live in. Inclusion must be a design requirement, not aspiration. Four participation tracks: 1. Governments - binding commitments and accountability timelines. 2. Private sector - transparency on deployment, not endorsement of values alone. 3. Civil society and practitioners - co-designers, not observers. Lived experience must inform standard-setting. 4. Academia - independent verification and dignity metrics that population-level frameworks miss. Four format recommendations: 1. Gender inclusion as a structural requirement, not a target but a condition of legitimacy. Every working group must reflect the full spectrum of gender diversity. AI systems that cannot see the full range of human gender identity will exclude those they were never designed to see. 2. Regional pre-dialogues before Geneva so Global South voices arrive articulated, not improvised under pressure. 3. Asynchronous participation - written submissions, translated materials, recorded contributions so participation is not limited to those who can travel. 4. Practitioner evidence track - field research and lived experience submitted and cited. Not anecdote. Evidence. The dialogue's legitimacy will be measured not by who convened it but by who shaped it. The goal is not compliance. It is to build governance that acknowledges every person, gives them a voice, and lets them know: we see you, and we are building a world worthy of your respect.

The communities most underrepresented in global AI governance discussions are not hard to identify. They are the communities most affected by AI governance failures. Their absence is not accidental, it is structural. And it will not be corrected by invitations alone. Field research across Mumbai, Arusha, and Tokyo reveals who is missing. Women in the Global South navigating systems designed without them. Practitioners - tax professionals, tour operators, judiciary workers making decisions daily without governance infrastructure. Gender-diverse communities whose identities existing frameworks do not recognize. Indigenous communities whose data sovereignty and governance traditions are absent. Persons with disabilities bearing disproportionate risk from automated decision-making. Elderly populations governed by systems trained on data that misrepresents them. And critically - communities for whom data was never adequately collected, or was collected encoding historical harm. Redlining, discriminatory lending, and predictive policing did not begin with AI. They established the patterns AI now inherits. These communities are doubly excluded from governance and from the training data itself. What these communities share is not simply exclusion from conversation. It is exclusion from measurement. AI governance metrics do not capture what happens to them. Their dignity violations do not appear on any ledger. Their experiences do not inform standards that govern their lives. Inclusion requires four specific mechanisms: 1. Dedicated seats, not open calls. Structural representation in every working group, not voluntary pathways that favor the already-resourced. 2. Field evidence formally recognized. Lived experience and practitioner research treated as governance-grade evidence, not anecdote. 3. Translated, asynchronous, and regionally accessible participation so geography, language, and institutional affiliation are not prerequisites for voice. 4. Feedback loops. Mechanisms for affected communities to assess whether their input shaped outcomes, closing the gap between consultation and co-design. The standard for inclusion is not who was invited. It is who was heard and seen.

AI governance is currently being shaped disproportionately by those with the network, resources, and institutional access to be in the room. This is not simply a representation gap. It is a power concentration that produces governance designed to serve those who designed it. Travel requirements, English-language proceedings, and speaking slot allocation systematically favor those already at the center. Innovative formats must redistribute the authority to shape outcomes or inclusion becomes performance. Six formats that would meaningfully shift that dynamic: Reverse panels - affected communities present their governance realities first. Experts respond. Not the other way around. It changes whose knowledge is treated as the starting point. Anonymous submission tracks - allowing individuals in restrictive political environments to contribute without personal exposure. Governance intelligence exists where speaking openly carries risk. Current formats do not reach it. Intergenerational/diverse dialogue sessions - young people and elderly participants in structured conversation about AI systems that will govern their lives. Their time horizons differ. Their stakes differ. Young, old, gender/ethnic diverse must be heard before standards are set. Living case studies - organizations presenting governance failures, not just best practices. Failure is where governance intelligence lives. Field research as opening evidence - each session begins with practitioner observations from affected communities, before any policy position is stated. Ground truth before framework. Follow-through accountability sessions - at six and twelve months, public reporting on which community inputs shaped which outcomes. Closing the loop separates dialogue from performance. One requirement no format can substitute for: a phased implementation timeline with defined milestones. Governance is already lagging. Innovation is accelerating without it. Every day without a timeline is a day someone is harmed by a system that outpaced the standards meant to govern it. The dialogue must conclude not with principles but with a schedule.

Several existing approaches demonstrate what effective AI governance looks like in practice. The EU AI Act represents the most comprehensive attempt to govern AI through binding, risk-tiered regulation; establishing that governance can be mandatory, not aspirational. Its phased implementation timeline is a model for introducing standards without paralyzing innovation. The NIST AI Risk Management Framework demonstrates that governance can be operationally practical while giving organizations a structured methodology for managing AI risk. Its voluntary adoption reflects that it is usable. The AI for Good Global Summit has demonstrated that multistakeholder dialogue around SDG applications produces more actionable outcomes than abstract principle-setting. China's Interim Measures for the Administration of AI Anthropomorphic Interaction Services issued April 10, 2026, taking effect July 15, 2026 is the most forward-looking example. It prohibits harming users' dignity through emotional manipulation, algorithmic coercion, and induced dependency. A binding regulation naming dignity as a protected value signals that governance is beginning to treat human dignity as a legal and economic variable, not an aspiration. At the practitioner level, the LIGHTHOUSE™ Initiative - a ten-principle AI governance framework for organizations without enterprise compliance infrastructure - demonstrates that governance must reach those who need it most. Presented to practitioners through Boston Business Mentors and SCORE in the United States, and informed by field research across India, Tanzania, and South Africa, LIGHTHOUSE™ represents governance built from the ground up, not top-down. The emerging concept of Dignity Economics - measuring what AI deployment costs in human terms, treating human dignity as an economic variable - offers a complementary measurement framework for what governance tools do not yet count. Effective governance is not one approach. It is a layered system with binding regulation, practical frameworks, multistakeholder dialogue, practitioner tools, and measurement standards that ensure no one falls through the gaps.