GYEOL Strategy Institute

The first Global Dialogue on AI Governance will be a success if it produces four concrete outcomes. One: governments and organizations are called to establish a philosophical layer in their AI governance frameworks — a binding specification of values, not a mission statement. Two: "meaningful human oversight" is formally redefined. Not procedural approval. Not a button click. But the protection of conditions under which humans can think, judge, and decide independently — before AI acts. Three: a mechanism for continuous philosophical alignment is recognized as necessary — ongoing processes through which humans verify that AI still reflects human intention, and can apply constraints when it does not. Four: the connection between AI philosophy, AI governance, and AI architecture is addressed as a coherent system — so that values defined in governance are actually implemented and verified in the architecture. If the Dialogue produces these four outcomes, it will have named something that current frameworks cannot yet name: that the sovereignty of human thought is a value worth protecting, and that governance must be designed with this protection as a deliberate goal.

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

I chose these three areas because they all point to the same question I carry with me every day: is my thinking still my own? I run a research institute with AI as staff. I train AI on my philosophy, and AI generates outputs I then verify and take responsibility for. This practice has taught me something governance frameworks have not yet named: there is a difference between AI that supports human thought and AI that replaces it. That difference is everything. Protection and promotion of human rights matters because the sovereignty of human thought - the capacity to question, deliberate, and arrive at meaning independently - is a right, not a preference. Current frameworks protect people from what AI does wrong. They do not protect the conditions under which people can think freely. That gap is urgent. Transparency, accountability, and human oversight matters because oversight without genuine judgment is not oversight. A human clicking an approval button while AI has already decided is a formality. Real oversight means the human thinks first. Social, economic, ethical, cultural, linguistic and technical implications matters because AI reshapes not only what people do, but how they think - what questions they ask, what they accept without asking. This erosion of diverse ways of knowing is a governance risk that has no name yet in current frameworks. These three areas, together, ask governance to do something it has not yet committed to: actively protect what humans must continue to do for themselves.

Yes. The issue not yet captured is this: the governance of the sovereignty of human thought. Current frameworks ask how to make AI safe, transparent, and accountable. These are necessary questions. But they assume the human on the other side of AI remains intact - still capable of questioning, still able to think independently, still sovereign over their own reasoning process. That assumption deserves examination. When AI answers before the question is fully formed, it changes the structure of inquiry. When AI filters information, drafts decisions, and manages schedules, human deliberation is not augmented - it is bypassed. When students receive personalized "optimal" learning paths, the productive struggle of not-knowing - the very condition that produces genuine understanding - is designed away. None of this appears in audit reports. None of it is measured by bias assessments or transparency indices. It accumulates quietly, one person at a time, every day. I have spent twenty years running IT governance, received four cancer diagnoses over the past decade, and through all of it - kept reading, writing, walking, and working. Both experiences taught me the same thing: the capacity to ask your own questions, in your own way, is not a luxury. It is survival. Governance frameworks that protect humans from AI-generated harm are necessary. But they are not sufficient. What is also needed - and does not yet exist - is governance that actively protects the conditions under which humans can continue to think for themselves. This is the cross-cutting issue beneath all the listed themes. It connects human rights, oversight, cultural implications, and safety. It is the question that none of the current frameworks have yet named. I submit this input because I believe it can be named. And once named, protected.

Korea presents an instructive paradox. Korean citizens are among the world's most rapid adopters of new digital tools. When new AI services emerge, Koreans experiment with them immediately, often discovering uses the developers had not anticipated. This is not mere enthusiasm. It reflects a cultural instinct to make tools one's own. And yet, Korea's education system remains organized around standardized testing and uniform curricula. Students are trained to find the right answer, not to form the right question. What concerns me most is what is happening at the earliest stages of learning. Workbook programs that Korean children once completed with pencil on paper have been rapidly replaced by AI-powered digital devices. The shift is largely unexamined. Young children who should be developing fine motor skills through handwriting, building patience through the friction of paper and pencil, and learning to sit with difficulty, are instead receiving instant feedback from adaptive AI systems optimized for engagement and efficiency. There is a critical distinction that Korean AI governance has not yet made: learning about AI is entirely different from learning from AI. The former builds understanding and agency. The latter risks replacing the very cognitive habits — attention, persistence, independent reasoning — that make human thought sovereign. For children under fifteen, paper books, handwriting, and unmediated thinking time are not nostalgic preferences. They are developmental necessities. But this is not only an education issue. Across Korea and the broader Asia-Pacific region, the same pattern repeats at every level — in workplaces, in public services, in daily life. AI adoption accelerates. The governance frameworks that should protect the conditions for independent human thought do not yet exist. The opportunity: the speed of adoption means that well-designed governance could have rapid and wide impact. Korea and the region have the infrastructure and the cultural momentum to lead. The governance gap: without frameworks that explicitly protect the sovereignty of thought — at every age, in every sector — rapid AI adoption will deepen cognitive displacement rather than amplify human capacity. That loss, once it accumulates across a generation, will be very difficult to reverse.

The AI Dialogue can do something no national regulation or regional framework can do alone: name what has not yet been named. Individual governments regulate what they can measure. They address bias, security, transparency — visible harms with measurable indicators. What they cannot easily regulate is what happens inside the conditions of human thought itself. No single nation can declare that the sovereignty of thought is a global governance concern. But a UN dialogue can. The most important role the Dialogue can play is to establish a shared vocabulary. "Meaningful human oversight" means different things in different frameworks. "Human-centered AI" is claimed by nearly every governance document without being defined. The Dialogue can move these phrases from aspiration to specification — giving them content that national frameworks can then implement. The second role is to create space for voices outside the usual centers of power. The most essential governance insights often come from practitioners, educators, independent researchers, and communities whose ways of knowing are not yet reflected in dominant AI systems. The Dialogue should not only convene governments and large institutions. It should actively solicit and elevate these perspectives. The third role is to model the very thing it is trying to protect. A dialogue that is genuinely open, that listens before it concludes, that allows unexpected voices to change its direction — such a dialogue demonstrates that human deliberation still matters. That demonstration is itself a governance act.

Several existing initiatives address pieces of the governance challenge. The OECD AI Principles established early frameworks for trustworthy AI. The EU AI Act brought regulatory specificity. UNESCO's Recommendation on the Ethics of AI addressed cultural and human rights dimensions. The Global Digital Compact set the context within which this Dialogue operates. Each of these contributes something important. What they share, however, is a common limitation: they focus on what AI systems do, not on what humans must continue to do for themselves. The AI Dialogue can add value by connecting these initiatives around a question none of them has yet centered: how do we govern the conditions for independent human thought in an age of AI? This means building on UNESCO's human rights framing — but extending it to include the sovereignty of thought as an explicit protection. It means connecting with education-focused initiatives — but insisting that AI in education be evaluated not only for learning outcomes, but for its effect on the development of independent inquiry. It means engaging with technical standards bodies — but asking them to consider not only what AI systems can do, but what they should be designed to leave for humans to do. The added value the Dialogue can bring is integration. These initiatives speak different languages — regulatory, technical, ethical, cultural. The Dialogue is the space where those languages can be translated into a common framework. That framework does not yet exist. Building it is the Dialogue's most important contribution.

The most important structural recommendation is this: design the Dialogue so that it can be changed by the people who participate in it. Most international dialogues are structured to confirm what organizers already believe is important. Inputs are collected. Themes are pre-defined. Participants speak into categories that were decided before they arrived. This is efficient. It is also how the most essential questions get missed. The sovereignty of thought — the issue I have raised throughout this submission — did not appear in any of the pre-defined thematic clusters. It emerged from practice: from running an AI-integrated research institute, from surviving four cancer diagnoses while continuing to read and write and think, from watching children trade pencils for tablets before they have learned what pencils are for. Governance frameworks cannot anticipate these insights in advance. They can only create conditions for them to surface. Concretely, I recommend three things. First, reserve space in the Dialogue's agenda that is genuinely open — not pre-themed, not pre-categorized. Allow participants to name what the framework has not yet named. Second, actively recruit practitioners, educators, independent researchers, and civil society voices from outside the major AI-producing nations. The governance concerns of those who use AI without shaping it are different from — and often more urgent than — the concerns of those who build it. Third, make written inputs like this one visible and searchable after the Dialogue. The value of inclusive participation is not only in the room. It is in the record. If this submission reaches one policymaker who had not yet encountered the phrase "sovereignty of thought," that is governance work. The Dialogue's format should reflect its purpose: to listen in ways that change what gets decided.

I will answer this question from where I stand. I am an independent researcher in Korea. I have no institutional affiliation, no government backing, no large research budget. I run a one-person institute with AI as staff. I have survived four cancer diagnoses. I write, walk, and think every morning before the world asks anything of me. People like me are underrepresented in global AI governance discussions. Not because we have nothing to say. But because the structures of international dialogue favor those with institutional credentials, travel budgets, and time to navigate complex submission processes. The voices that tend to dominate are those of governments, large technology companies, and well-funded academic institutions — all of whom have legitimate perspectives, and none of whom experience AI governance the way a sole practitioner, a patient, a teacher, or a parent does. The perspectives most underrepresented are those of people who live closest to the consequences. Parents deciding whether to give their young children AI-powered learning devices. Teachers watching students lose the habit of sitting with difficult questions. Patients navigating medical decisions increasingly mediated by algorithmic recommendation. Independent researchers and writers asking whether the thoughts they are thinking are still their own. These are not marginal concerns. They are where the governance gaps are most acutely felt. To include these voices, the Dialogue must do more than open a submission portal. It must actively reach into communities that do not already speak the language of international governance. It must translate its questions into terms that practitioners, caregivers, and citizens can engage with. And it must demonstrate, through its outcomes, that these voices changed something. Otherwise, inclusive participation remains a aspiration rather than a practice.

The most innovative format the AI Dialogue could adopt is also the simplest: begin with questions, not presentations. Most international dialogues open with statements. Governments declare positions. Institutions present frameworks. Experts summarize research. By the time discussion begins, the conceptual territory has already been claimed. Participants respond to what was said rather than explore what was not. I propose a different structure for at least one session: open with a question that no one has answered yet, and hold it open for the duration. A question like: Is your thinking still your own? This question — which I carry with me every day as a researcher who works alongside AI — is not rhetorical. It is a genuine governance question that connects human rights, oversight, cultural implications, and the sovereignty of thought. It is also a question that a government delegate, an independent researcher, a teacher, and a parent can all engage with from their own experience. It does not require institutional credentials to have a meaningful answer. A second recommendation: create deliberate silence in the agenda. Not every session needs to produce a declaration or a summary. Some of the most important governance insights emerge slowly, through reflection rather than reaction. A session designed for listening — where participants write before they speak, or walk before they conclude — would model the very practice the Dialogue is trying to protect. Third: document dissent. When participants disagree, record the disagreement rather than smoothing it into consensus language. Disagreement is information. It shows where the real governance tensions are. A Dialogue that produces only consensus has not yet found the hard questions. Meaningful engagement begins when participants feel that their thinking — not just their position — is welcome.

The most concrete example I can offer is the one I have built and operate myself. GYEOL Strategy Institute runs on a five-layer governance structure in which the founder's philosophy is the absolute standard for all AI activity. AI has read-only access to this foundational layer. Every output AI generates passes through a value verification process before it is formalized - checked against five core values: Life Respect, Sovereignty of Thought, Truth, Beauty, and Curiosity. The final judgment belongs to the human founder, always. This structure addresses three governance challenges that larger frameworks have not yet solved. First, it makes philosophy operational. The founder's values are not a mission statement on a wall. They are embedded as data, as prompts, as verification criteria - traceable through every layer of the architecture. Second, it defines where the human stands. When AI proposes and AI generates, the human does not disappear into the loop. There is a designed moment - at L4, value verification - where AI stops and human judgment is required. This is not procedural. It is structural. Third, it is adaptive. When AI models change, the knowledge base - maintained in Obsidian as a Single Source of Truth - remains intact. The governance structure does not depend on any particular AI system. It depends on the human philosophy that governs all of them. Beyond my own practice, I would point to Finland's approach to AI literacy education - teaching children to understand and question AI rather than simply use it - and Sweden's legislation, effective autumn 2026, banning mobile phones in schools throughout the full school day for children aged 7 to 16, alongside a return to physical textbooks, as policy examples that protect the conditions for independent human thought. These are different in scale. They share a common logic: governance that protects what humans must continue to do for themselves.