Association for Computing Machinery (ACM) Technology Policy Council (TPC)

1. A Shared Foundational Principles Charter Governments, big tech, civil society, and academia should co-author a living "Common Ground Charter" — a concise set of non-negotiable AI principles (safety, transparency, human dignity, equitable access) that all parties can endorse regardless of political or economic differences. Unlike static treaties, this charter should be version-controlled and publicly reviewable, modeled on open-source governance. 2. A Multi-Stakeholder Inclusion Mechanism To bring major AI capacity builders — including large technology companies, frontier model developers, regional AI hubs, and civil society researchers into sustained engagement, the Dialogue should establish a formal "AI Governance Innovation Track." This would offer participating organizations co-design privileges on non-binding standards, recognition as Dialogue Partners, and access to a shared early-warning network for emerging AI risks. Crucially, representation from the Global South must be structurally guaranteed, not optional.

• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Interoperability of governance approaches
• AI capacity-building

We selected these four priorities based on a simple logic: start from what we can actually do today, and build upward. 1. AI Capacity-Building Before any governance can work, people must understand what they are governing. Most governments and communities still lack the basic technical literacy and infrastructure to participate meaningfully in AI decision-making. This is the first floor. Without it, nothing else stands. 2. Social, Economic, Ethical, Cultural and Linguistic Implications Once we begin building, we must understand the real-world consequences of our actions - on jobs, on culture, on language, on vulnerable communities. Governance without this understanding is blind. We need to see the full human picture before drawing any rules. 3. Interoperability of Governance Approaches Different nations are already moving in different directions. If we cannot find compatibility between our approaches now, fragmentation will become permanent and irreversible. This is the moment to establish shared reference points - not one single law, but a common language between systems. 4. Transparency, Accountability and Human Oversight When we act, we need to know where the line is. Where is the boundary? What is the bottom limit we must never cross? Transparency and accountability define that border - so that within it, we are free to build good things confidently. Why not others as top priorities? Not because they do not matter - it matters enormously. But right now, there are fundamental technical problems we simply cannot solve yet. Declaring safety a governance priority before we have the capacity, the shared understanding, and the accountability structures in place produces only empty promises. We must build the foundation first.

AI Sovereignty One critical issue not adequately captured by the listed themes is AI sovereignty - the right of nations, communities, and peoples to exercise meaningful control over the AI systems that shape their societies. AI sovereignty operates on three levels, each urgently requiring attention: 1. Data Sovereignty The raw material of AI is data - and today, the vast majority of the world's data is processed, stored, and monetized by a handful of corporations headquartered in two or three countries. Nations, particularly in the Global South, are effectively exporting their most valuable resource without compensation or control. Governance frameworks must establish the right of peoples to own, govern, and benefit from data generated within their borders and communities. 2. Computational Sovereignty Access to the computing infrastructure required to train and run advanced AI models is deeply concentrated. Countries without domestic chip manufacturing, cloud infrastructure, or high-performance computing capacity are structurally dependent on foreign systems for their most critical decisions - in healthcare, defense, education, and public services. This is a new form of technological dependency that the Dialogue must name and address. 3. Algorithmic Sovereignty When a government deploys an AI system built, trained, and updated by a foreign entity, who is truly governing? Nations must retain the right to audit, modify, or reject AI systems operating within their jurisdiction - regardless of where those systems were developed. AI sovereignty is not about isolationism. It is about ensuring that every nation has a genuine seat at the table - not just in dialogue, but in the actual architecture of the AI systems that govern daily life. Without sovereignty, governance is participation without power.

International AI governance today resembles a thousand puzzle pieces scattered across the table — national strategies, regional frameworks, corporate standards, civil society demands, and community needs, each shaped differently, each incomplete alone. The AI Dialogue's most essential role is to be the puzzle assembler. 1. Building a Shared Grammar of Governance The Dialogue's most immediate contribution is creating a common language. Nations cannot cooperate on what they cannot mutually describe. The Dialogue should produce shared definitions, taxonomies, and risk classifications that allow different governance systems to recognize and communicate with each other — not replacing national frameworks, but connecting them. 2. From Dialogue to Mechanism Words without structure dissolve. The Dialogue must graduate from conversation to institution — establishing standing working groups, clear reporting cycles, and a permanent secretariat function that maintains momentum between annual sessions. International cooperation on AI cannot survive on two meetings per year alone. 3. Bridging the Power Asymmetry The Dialogue has a unique legitimacy that no corporate forum or regional body possesses — universal membership under the UN. This gives it the moral authority to ensure that smaller nations, indigenous communities, and underrepresented regions are not merely consulted but genuinely shape outcomes. Cooperation that only serves powerful actors is not cooperation — it is coordination of dominance. 4. Early Warning and Rapid Response Perhaps the most innovative role the Dialogue could play is to establish a global AI incident-sharing network — a cooperative mechanism in which nations share information about AI failures, harms, and near-misses in real time, enabling collective learning before crises escalate. The Dialogue's greatest contribution would be to prove that international cooperation on AI is not only necessary but also possible

Association for Computing Machinery Tech Policy Council

1. Thematic Working Groups Establish permanent working groups aligned to the four core priorities — capacity-building, societal implications, interoperability, and transparency. Each group meets between annual sessions, produces draft recommendations, and reports directly to the plenary. Working groups should be co-chaired by both a government representative and a non-government stakeholder. 2. Regional Committees Create structured regional committees — Africa, Asia-Pacific, Latin America, Small Island States, Europe, and North America — that aggregate regional priorities, surface locally specific challenges, and feed consolidated positions into the global Dialogue. This ensures Geneva hears many voices, not just the loudest ones. 3. A Multi-Stakeholder Advisory Council A standing council of rotating representatives from industry, civil society, academia, and indigenous communities that holds the Dialogue accountable to its own inclusivity commitments between sessions. 4. Open Submission and Digital Participation Not everyone can travel to Geneva. A permanent digital participation channel — with real translation support across major world languages — ensures the puzzle has no missing pieces by geography or resources.

Traditional conference formats — keynotes, panels, statements — produce passive audiences and active speakers. The AI Dialogue needs formats that turn every participant into a puzzle assembler, not just a witness. 1. Live Consensus Mapping Deploy a real-time digital platform during sessions where participants continuously tag areas of agreement, disagreement, and uncertainty as discussions unfold. AI-assisted analysis surfaces emerging consensus visually on a shared screen — making common ground visible the moment it appears, not months later in a published report. 2. Reverse Panels Instead of experts presenting to audiences, community representatives, small nation delegates, and civil society voices present their realities to expert panels — who must then respond with concrete governance implications. This structurally inverts the traditional power dynamic. 3. Red Team Challenges Dedicated sessions where stakeholders are assigned to stress-test proposed principles and frameworks — actively trying to break them, expose loopholes, or identify unintended consequences. The best governance is tested before deployment, not after. 4. Cross-Regional Negotiation Simulations Structured role-playing exercises where participants represent regions or stakeholder groups different from their own. This builds genuine empathy, reveals blind spots, and often produces unexpected breakthrough ideas that formal negotiation never surfaces. 5. The Unfinished Document Session Present a deliberately incomplete draft governance framework in the room. Give every stakeholder group thirty minutes to fill different sections simultaneously. Then assemble the pieces publicly. The gaps and contradictions that emerge become the most honest map of where real work is still needed. 6. Persistent Digital Twin of the Dialogue Between Geneva and New York sessions, maintain a living online space — open to all — where proposals evolve, comments accumulate, and consensus is tracked continuously. The Dialogue should never fully close between meetings. Engagement is not attendance. Engagement is co-ownership of the outcome.