Universidad de Los Andes

To consider the first Global Dialogue on AI Governance a success, it must move beyond high-level principles to establish a practical, inclusive, and interoperable framework. Success should be measured by three primary outcomes:1. Institutional Interoperability & Fragmentation Reduction. The Dialogue is successful if it creates a "Common Regulatory Language." With the EU AI Act in enforcement and various national frameworks emerging, the UN must serve as the bridge to prevent a fractured global market. A successful outcome would be a roadmap for mutual recognition of safety standards, ensuring that a "safe" AI system in one jurisdiction meets baseline global expectations.2. Bridging the "Governance Gap" for the Global South. True success requires that AI governance is not a "club of wealthy nations." We must see concrete commitments to a Global AI Capacity Development Network. This includes not just sharing best practices, but providing developing nations with the technical and legal resources to participate in the AI economy without being mere "data exporters" or passive consumers of foreign technology.3. Scientific Independence and Real-World Accountability. The Dialogue must empower the Independent International Scientific Panel on AI to provide unbiased, evidence-based risk assessments. A successful session will produce a mechanism for "voluntary incident reporting," where states and companies share significant AI failures transparently. This shifts the culture from defensive secrecy to collective learning.Ultimately, the Dialogue succeeds if it establishes a permanent, multistakeholder feedback loop—one where civil society and technical experts have a seat at the table alongside governments to ensure AI remains anchored in human rights and the Sustainable Development Goals.

• Safe, secure and trustworthy AI
• Open-source software, open data and open AI models
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI

To ensure AI serves as a global public good, our entity prioritizes a governance model that balances technical robustness with deep social responsibility. We have identified the following four areas for urgent action: 1. Transparency, Accountability, and Human Oversight (Priority 6) Governance is hollow without a "look under the hood." We advocate for mandatory disclosure of AI training data origins and decision-making logic, particularly in high-stakes sectors like healthcare and finance. Accountability must move from abstract theory to legal reality through "human-in-the-loop" mandates that ensure a human remains the final arbiter of automated decisions. 2. Open-Source Software, Open Data, and Open AI Models (Priority 7) Innovation must not be gated behind proprietary silos. Promoting open-source models is the most effective way to democratize AI. By supporting open data and models, we lower the barrier to entry for smaller players and allow the global technical community to conduct independent safety audits, which proprietary "black box" systems currently prevent. 3. Safe, Secure, and Trustworthy AI (Priority 1) Trust is the currency of AI adoption. Our focus is on the creation of global red-teaming standards and "safety-by-design" protocols. This ensures that systems are resilient against adversarial attacks and that their reliability is verified before they reach the public square. 4. Social, Economic, Ethical, Cultural, and Technical Implications (Priority 3) AI does not exist in a vacuum. We prioritize the mitigation of algorithmic bias that threatens linguistic and cultural diversity. Our engagement focuses on the socio-economic impact of automation, ensuring that AI-driven efficiency does not come at the cost of labor rights or cultural erasure. By focusing on these four pillars, the Global Dialogue can foster an ecosystem that is technically sound, publicly verifiable, and culturally inclusive.

To truly future-proof AI governance, the Dialogue must look beyond the screen. We identify Environmental Sustainability, Cognitive Liberty (Neuro-rights), and AI-Biotech Biosecurity as three critical 'blind spots' in the current thematic framework. Addressing these ensures that AI does not solve digital problems while creating physical or biological crises.

The primary governance gap affecting our region is the "Asymmetry of Oversight." While we are rapid adopters of AI, the lack of localized regulatory frameworks for Transparency and Accountability means we often import "black-box" systems that are not calibrated for our specific socioeconomic contexts. Significant Challenges: Algorithmic Colonialism & Bias: Without local oversight, imported AI models often carry Western-centric biases. This directly impacts Social and Cultural Implications (Priority 3), as automated systems in credit scoring or hiring may inadvertently discriminate against local demographic nuances or linguistic dialects not represented in original training sets. Security Vulnerabilities: The gap in Safe and Trustworthy AI creates a "testing ground" effect. Lack of stringent local red-teaming standards makes our digital infrastructure more susceptible to AI-driven cyber-attacks and misinformation campaigns that threaten social cohesion. Significant Opportunities: The Open-Source Leapfrog: The shift toward Open-Source and Open Data presents a historic opportunity. By adopting open models, our local developers can bypass expensive proprietary "tolls." This allows for the creation of "Sovereign AI"—tools trained on local data that solve regional problems, such as precision agriculture or local language processing, without yielding data autonomy to foreign entities. Regulatory Innovation: The current governance vacuum allows us to build "Sandboxes" that are more agile than those in heavily regulated markets. We have the opportunity to design an accountability framework that prioritizes human oversight while remaining lean enough to foster a thriving startup ecosystem. By bridging these gaps, we transition from being passive consumers of global AI to active architects of a localized, ethical digital future.

To maximize international cooperation, the Global Dialogue on AI Governance must transition from a "talking shop" to a "Global Clearinghouse" for AI safety and equity. Based on your focus on openness and accountability, the Dialogue can serve three pivotal roles:1. Harmonizing the "Regulatory Patchwork"The greatest risk to international cooperation is fragmented national laws that stifle innovation. The Dialogue can establish Common Technical Standards and a Mutual Recognition Framework. By creating a unified "baseline" for what constitutes Safe and Trustworthy AI (Priority 1), the UN can ensure that a startup in one country can scale globally without navigating 193 different sets of compliance rules.2. Standardizing Transparency and Openness (Priority 6 & 7)The Dialogue is the ideal platform to negotiate a "Global Open-Source Accord." This would encourage nations to share non-sensitive, high-quality public datasets and open-source models. Such cooperation prevents a "Digital Iron Curtain" and ensures that the technical implications of AI are audited by a global community rather than a handful of private corporations.3. Establishing an "Early Warning System"Cooperation is most urgent when things go wrong. The Dialogue can formalize an International AI Incident Database. Similar to aviation safety reporting, this would allow nations to share data on AI failures or "near-misses" without fear of immediate litigation. This collective learning turns localized technical failures into global governance safeguards.ConclusionUltimately, the Dialogue's role is to ensure that AI governance is interoperable, inclusive, and evidence-based. It must bridge the gap between the "Global North's" focus on risk and the "Global South's" focus on development, proving that safety and progress.

To ensure the Global Dialogue on AI Governance is effective, it must avoid reinventing the wheel and instead act as the "connective tissue" between a fragmented landscape of existing initiatives. Strategic Foundations to Build Upon The Dialogue should integrate and scale the work of: The OECD AI Principles & Policy Observatory: Leveraging their established definitions and intergovernmental standards to ensure regulatory interoperability. ITU's "AI for Good" & Global Standards: Utilizing the ITU's technical expertise and its massive network of 50+ UN agencies to ground policy in engineering reality. UNESCO's Recommendation on the Ethics of AI: Building on the first global standard focused specifically on human rights and ethical safeguards. The Global Partnership on AI (GPAI): Connecting with its project-oriented research to turn high-level UN diplomacy into applied technical solutions. The Unique Added Value of the AI Dialogue While the initiatives above are excellent, they are often limited by membership (OECD/GPAI) or specific mandates (UNESCO/ITU). The UN Global Dialogue brings three unique "force multipliers": Universal Legitimacy: As the only forum involving all 193 Member States, it is the only body that can claim a truly global mandate, ensuring the "Global South" is an architect—not just a consumer—of AI rules. Scientific Integration: By hosting the Independent International Scientific Panel on AI, the Dialogue anchors political debate in peer-reviewed evidence, effectively creating an "IPCC for AI" to separate hype from systemic risk. Cross-Pillar Coordination: The Dialogue serves as the "Glue" between the Global Digital Compact, the Sustainable Development Goals (SDGs), and international human rights law, ensuring AI development does not happen in a silo away from climate and equity targets. In short, the Dialogue's added value is inclusion at scale—turning a fragmented "patchwork" of rules into a coherent global framework.

1. The "Open-Source" Consultation Model The Dialogue should implement a Permanent Digital Repository for written inputs, much like a public "GitHub for Policy." This allows academia and technical experts to submit "Pull Requests" on draft frameworks, ensuring transparency and version control in the governance process. 2. Multi-Stakeholder "Working Nodes" Rather than just plenary sessions, the structure should feature Thematic Working Groups composed of balanced quotas: 25% Government, 25% Civil Society, 25% Industry, and 25% Academia. These "Nodes" should focus on technical interoperability and human rights impact assessments. 3. "Red-Teaming" Policy Labs Civil society and technical experts should be invited to participate in Policy Stress Tests. These interactive workshops would simulate real-world AI failures (e.g., a cross-border deepfake crisis) to evaluate if proposed UN frameworks are practically enforceable or merely theoretical. 4. Regional Hubs for the Global South To lower the barrier to entry, the Dialogue should hold Regional Satellite Summits. This ensures that stakeholders from resource-constrained environments can contribute without the prohibitive costs of international travel.

AI governance is currently suffering from a "participation divide" that mirrors the digital divide. To ensure the Internet remains a force for good, we must bring the following voices to the center of the table: The Underrepresented Voices The Global South's Technical Community: While "policy" is discussed globally, the network engineers and developers from emerging economies are often sidelined. Without their input, governance frameworks risk being technically infeasible or harmful to local internet infrastructure. Indigenous and Linguistic Minorities: AI models are predominantly trained on "High-Resource" languages. The communities whose languages and traditional knowledge are being "data-scraped" without consent or benefit-sharing are virtually invisible in high-level governance. Small-to-Medium Enterprises (SMEs) and Independent Developers: Current discussions are dominated by "Big Tech" and large states. The "permissionless innovation" that built the Internet is at risk if regulatory burdens are designed only for those with massive legal departments. How to Build an Inclusive Table The Multi-stakeholder Model: We must move beyond bilateral government agreements. True inclusion requires a "bottom-up" approach where civil society and the technical community have equal footing with states—not just as observers, but as co-authors of policy. Collaborative Capacity Building: Inclusion isn't just about an invitation; it's about the resources to participate. We need "Governance Fellowships" and regional technical hubs that empower local experts to analyze how global AI rules impact their local networks. Linguistic Inclusion: The Dialogue itself must be accessible. This means providing multilingual resources and asynchronous contribution platforms to overcome time-zone and language barriers. By adopting a multistakeholder, consensus-driven approach, we ensure that AI governance strengthens the open Internet rather than fragmenting it.

the AI Dialogue should adopt formats that prioritize asynchronous collaboration and interactive problem-solving. 1. The "Policy Hackathon" Instead of speeches, the Dialogue should host 48-hour sprints where multi-stakeholder teams (diplomats, coders, and activists) co-draft technical annexes or "Model Laws." This shifts the energy from debate to tangible co-creation, using version-control platforms like GitHub to track contributions transparently. 2. "Reverse Pitching" Sessions Flip the script: have governments "pitch" their regulatory challenges to a panel of civil society and technical experts. This forces policymakers to articulate specific pain points, allowing experts to provide targeted, real-world feedback rather than generic high-level advice. 3. Radical Inclusivity via "Digital Twins" For those unable to travel, create a Persistent Virtual Forum—a digital twin of the Dialogue. This allows for real-time, multilingual commenting and "up-voting" on proposals, ensuring that a developer in Lagos or a researcher in La Paz has the same procedural weight as a delegate in New York. 4. Foresight "War Gaming" Use role-playing scenarios to simulate AI-driven crises. These simulations reveal unforeseen policy gaps and force participants to negotiate under pressure, fostering deeper empathy and more resilient international cooperation.

The Dialogue should integrate and scale the work of: The DiploFoundation's AI Apprenticeship: This "learning-by-doing" model is a gold standard for capacity building. It bridges the gap between diplomatic theory and technical practice, empowering officials to build and audit AI agents themselves. It provides a blueprint for how the UN can foster "sovereign AI" capabilities in the Global South. The OECD AI Principles & Policy Observatory: Leveraging their established definitions to ensure global regulatory interoperability. ITU's "AI for Good" & Global Standards: Utilizing their technical network to ground policy in engineering reality. UNESCO's Recommendation on the Ethics of AI: Building on the first global framework centered on human rights and cultural diversity. The Unique Added Value of the AI Dialogue While these initiatives are excellent, they are often siloed. The UN Global Dialogue brings three unique "force multipliers": Universal Legitimacy: As the only forum involving all 193 Member States, it ensures that emerging economies are architects-not just consumers-of AI rules. Scientific Integration: By hosting the Independent International Scientific Panel on AI, the Dialogue anchors political debate in peer-reviewed evidence (an "IPCC for AI"). Cross-Pillar Coordination: It serves as the "Glue" between the Global Digital Compact, the SDGs, and international law, ensuring AI development supports-rather than undermines-climate and equity targets.