UPF Barcelona School of Management

The Global Dialogue should not attempt the politically impossible: a unified global framework that asks nations at vastly different levels of AI maturity, geopolitical interest, and regulatory philosophy to govern the same way. That ambition will produce either lowest-common-denominator text or paralysis. The realistic agreement to pursue is a set of minimum shared floors: baseline commitments that every signatory can implement according to their own institutional capacity and political context, while preserving space for divergence above that floor. In practice, this means agreeing on three non-negotiable principles with flexible implementation. That no AI system deployed at societal scale, in public services, justice, health, or security, should operate without a documented accountability chain traceable to a human or institution. That access to foundational AI tools, open models, open data, basic computing infrastructure, must be treated as a global public good, with wealthier nations and major AI powers making concrete, time-bound contributions to close the divide. That governance frameworks must be designed to talk to each other, even when they disagree, through shared taxonomies, mutual recognition mechanisms, and interoperability standards that prevent the current fragmentation from hardening into permanent geopolitical blocs. These three floors are politically achievable precisely because they do not ask the US, China, the EU, or the African Union to abandon their positions. They ask only that each accepts a minimum shared reality: that ungoverned AI at scale is a collective risk, that AI exclusion is a collective failure, and that governance fragmentation benefits no one except those powerful enough to exploit the gaps. That is a deal that can actually be made.

AI capacity-building;Open-source software, open data and open AI models;Transparency, accountability, and human oversight;Social, economic, ethical, cultural, linguistic and technical implications of AI;

Not every seat at the table is yours to claim. The Global Dialogue on AI Governance will be crowded with governments, tech giants, and international bodies all jostling for influence. The smart move is not to show up everywhere. It is to show up where you actually have something real to say. So where does UPF Barcelona School of Management belong in this conversation? Start with the big messy human question. AI is not just a technology problem. It reshapes jobs, cultures, languages, and power. That is precisely the territory where a business and social sciences school thinks, researches, and teaches. We do not just study AI systems. We study what happens to people when those systems arrive. Then comes accountability. Europe is living through the world's most ambitious AI regulation experiment right now. The EU AI Act is not abstract for us. It is the operating environment of every company we train leaders for. We have skin in the game on transparency and human oversight, and our research on why human judgment cannot be automated away gives us something concrete to contribute. Third, capacity building. This one is personal. Teaching in Rwanda, Peru, Japan, and across Latin America is not a line on a CV. It is evidence that we understand what AI exclusion looks like on the ground, and what it actually takes to close that gap. That experience is rare in this room. Finally, open AI. Advocating for open models and open data is where European academia can play an honest broker role between powerful nations hoarding AI advantage and developing countries locked out of it. Four priorities. One coherent voice. That is how you make an impact rather than just make noise.

The four thematic pillars of the AI Dialogue are necessary. They are not sufficient. Two deeper disruptions are already underway that none of the listed themes adequately capture, and ignoring them will make whatever governance frameworks emerge feel obsolete before the ink dries. The first is the crisis of human identity. Yuval Noah Harari has pointed to something unsettling: for the first time in history, an external system can master human language with greater fluency, speed, and reach than any individual human being. Language has always been our most intimate technology. It is how we construct selfhood, culture, memory, and meaning. When AI colonises that space, the question is no longer just what humans can do that machines cannot. The question becomes who are we when the thing that made us distinctly human is no longer exclusively ours. Governance frameworks focused on transparency and oversight do not touch this. We need an entirely new conversation about cultivating a human identity that lives beyond language, one rooted in embodied experience, emotional depth, moral imagination, and relational presence. This is not philosophy for its own sake. It is the foundation without which no AI governance framework will have a willing human subject at its centre. The second is the figure of the AI immigrant. Billions of people are arriving in an AI-native world without a map, without preparation, and without institutional support. Like all immigrants, they face a choice between assimilation and exclusion. Unlike most immigration debates, nobody is advocating for them. Capacity-building frameworks talk about skills and infrastructure. They rarely talk about dignity, disorientation, or the psychological cost of forced technological adaptation.

Spain, where I live 8 months a year, is in an unusually exposed position. AESIA has produced more than 20 guidance documents and is considered the most advanced national AI authority in the EU. That leadership creates a paradox: the expectations are high, the clock is running, and full implementation of the EU AI Act becomes mandatory in August 2026. For UPF Barcelona School of Management, this is not an abstract debate. Every executive programme and corporate partnership now sits inside a regulatory environment demanding documented accountability chains, transparency in AI-generated content, and demonstrable human oversight. The governance gap here is not about missing frameworks but about missing organisational capacity to implement them intelligently. Japan, where i live the rest of the year, has taken the opposite bet. Japan's AI Promotion Act views AI as a strategic asset and promotes industrial use while mitigating risks through transparency. International Bar Association Crucially, the Act contains no explicit penalties, with enforcement resting on a cooperative and reputational model rather than binding mandates. Future of Privacy Forum This innovation-first approach creates a different kind of gap: voluntary frameworks concentrate benefits among those already powerful enough to self-govern, leaving smaller actors without meaningful recourse. The governance gap in Japan is not overreach but underreach. The business school sector is living through an identity crisis disguised as an opportunity. Schools are now making deliberate decisions about curriculum design, faculty roles, infrastructure investment, and ethical responsibility. But most are producing AI strategies without AI governance cultures. For UPF, positioned between Spain's advanced regulatory environment and a global student body arriving from contexts as different as Japan and Peru, that gap is both a vulnerability and a singular opportunity to lead.

Architecture takes time. But you have to lay the foundations before you build. The risk facing the first Global Dialogue on AI Governance is the same risk that has haunted every major multilateral technology initiative before it: producing language that satisfies everyone in the room and changes nothing outside it. Avoiding that fate requires being honest about what international cooperation on AI governance can realistically do right now, and what it cannot. What it cannot do is harmonise. The distance between the EU's binding, rights-based regulatory architecture, Japan's innovation-first voluntary model, and the governance vacuums across much of the Global South is not a drafting problem. It is a structural expression of different political economies, different levels of AI maturity, and different calculations about who benefits from regulation and who bears its costs. No dialogue resolves that in one session. What the Dialogue can do is build the infrastructure for cooperation that does not yet exist. That means three things. First, it can establish a shared early warning system: a mechanism through which countries can flag AI-related harms crossing borders before they become crises, without requiring agreement on the underlying governance frameworks that produced them. Second, it can create a genuine knowledge commons. Not another repository of principles, but a living, multilaterally maintained body of evidence on what governance interventions are actually working, for whom, and under what conditions. Business schools, research universities, and civil society organisations have a specific role to play here that governments and tech companies cannot fill alone. Third, it can make capacity asymmetry visible and politically costly to ignore. The Dialogue succeeds if it becomes the forum where the AI divide stops being a footnote in development discussions and becomes a first-order governance failure demanding a first-order institutional response.

The landscape of AI governance initiatives is already crowded. The Dialogue's first obligation is intellectual honesty about what exists before claiming to add something new. The OECD AI Principles and the UNESCO Recommendation on the Ethics of AI represent the most broadly adopted normative foundations, with over 40 countries aligned around shared values on transparency, accountability, and human-centred design. The Hiroshima AI Process, launched under Japan's G7 presidency, produced the first internationally endorsed code of conduct for advanced AI developers, demonstrating that major AI powers can find common language when political will exists. The Council of Europe Framework Convention on AI and Human Rights, Democracy and the Rule of Law goes further, offering the first binding international instrument, albeit with limited signatories. The UN Secretary-General's Advisory Body on AI produced a governance blueprint that the Dialogue inherits directly. Regionally, the EU AI Act is reshaping global standards de facto, as any company operating internationally must now reckon with its requirements regardless of their home jurisdiction. The honest gap across all of these is the same: they were designed by and for countries that already have institutional capacity to engage with them. The Global South is consulted but rarely co-authors the frameworks it is then expected to implement. This is precisely where the Dialogue adds irreplaceable value. It is the only forum with universal membership, convened under UN auspices, where countries that have been rule-takers in AI governance can become rule-makers. Its added value is not another framework. It is legitimacy, inclusion, and the political weight that only a genuinely multilateral process can generate.

The Dialogue's credibility depends entirely on who is actually in the room and whether their presence shapes outcomes or merely decorates them. Governments set the political floor through negotiated commitments, but they should not monopolise the agenda. Civil society and academic institutions must contribute as co-authors of evidence, not as invited commentators after positions have already hardened. The private sector, including major AI developers, belongs at the table with disclosure obligations attached, not open microphones. On format, three structural recommendations matter. The Dialogue should open with a dedicated listening phase in which developing countries and underrepresented communities present documented AI impacts before any governance proposals are tabled. Evidence before solutions. Thematic tracks should run in parallel but converge on shared floor commitments, preventing the most contested issues from blocking progress on areas of genuine consensus. Finally, the Dialogue must have a follow-up mechanism with named accountability. A conversation without consequences is a conference. The difference between the two is whether anyone checks back.

The most underrepresented voice in global AI governance is the one that represents the majority of humanity: ordinary people who neither build AI systems nor negotiate international frameworks, but live entirely inside their consequences. But there is a sharper paradox worth naming directly. The UN process is well-spirited, inclusive by design, and largely powerless. The technology companies shaping AI at civilisational scale were not elected, are not accountable to any multilateral body, and make decisions affecting billions without asking anyone. The real representation gap is not a missing seat at the table. It is that the people with the most power to shape AI are structurally outside the room where governance happens, and currently have every incentive to stay there.

Every major multilateral governance forum in history has made the same mistake: it organises itself around the people who are comfortable in multilateral governance forums. The result is technically competent, politically cautious, and almost entirely disconnected from the lived reality it claims to address. The AI Dialogue needs a different architecture. Not panels. Not position papers read aloud to people checking their phones. Something that actually forces uncomfortable encounters between people who would never otherwise share a room or a problem. The format that could genuinely change something is what might be called a Reverse Commission. Instead of governments and institutions producing recommendations that filter down to communities, you invert the entire logic. You start with randomly selected citizens (may be also a "teen" board of people under twenty years old) from less than twenty countries across different income levels, regions, and AI exposure contexts. They are given three months, supported by philosophers, technologists, and social scientists, to define the AI governance questions that actually matter to their lives. Not the questions experts assume matter. The actual questions. Those questions then become the mandatory opening agenda of every formal session. Companies, politicians, regulators, and academics must respond to that citizen-generated agenda before tabling their own. The burden of relevance shifts. Philosophers sit not as keynote speakers but as embedded challengers inside negotiating rooms, with the explicit mandate to slow down consensus when it is being manufactured rather than earned. Their role is not decoration but friction, the productive kind that prevents bad agreements from being dressed up as good ones.

Each of the following works because it relocates power before relocating responsibility. Taiwan's digital minister Audrey Tang pioneered vTaiwan, a platform where algorithms map areas of citizen consensus on tech policy before politicians ever vote, effectively letting disagreement become data rather than deadlock. Chile embedded AI ethics review directly into its constitutional drafting process in 2021, making algorithmic accountability a foundational right rather than a regulatory afterthought, the first country to attempt governance at that constitutional depth. The Maori Data Sovereignty Network in New Zealand asserts that indigenous communities own not just their personal data but the collective intelligence derived from it, creating a governance model built on ancestral epistemology rather than individual rights frameworks. Kenya's Algorithmic Accountability Bill proposes mandatory community impact hearings before any AI system is deployed in public services, giving affected populations veto power before deployment rather than complaint mechanisms after harm.