Positive Human Diversity Ventures

Open and inclusive dialogue of AI risks, proposed mitigation, funding, governance and agreement. Working groups are formed to follow up with stakeholders, actors, government agencies, private and public organizations, local representatives, civil society ,etc.

• Interoperability of governance approaches
• Safe, secure and trustworthy AI

Interoperability in AI governance frameworks enables diverse, cross-border regulations to work together, reducing fragmentation and promoting trustworthy, safe, and efficient AI deployment. It involves aligning ethical principles, technical standards (e.g., ISO/IEC 42001, NIST AI RMF), and compliance procedures across jurisdictions. Key Aspects of AI Governance Interoperability: Three Pillars of Interoperability: Efforts focus on ethical (shared values), regulatory (compatible laws), and technical (shared standards) harmonization. Compliance Strategy: Instead of waiting for a single global rulebook, companies are adopting interoperable strategies that allow "portable evidence" (policies, audits, and risk assessments) to be used across multiple markets like the EU AI Act and NIST frameworks. Global Coordination: The United Nations and other international bodies are fostering collaboration to ensure AI governance is not only safe but also adaptable to different national contexts, bridging gaps between regions like the Global South and more developed digital economies. Key Frameworks: Interoperability frequently bridges major systems, including the OECD AI Principles, G7 initiatives, and the UN Global Digital Compact. Benefits: Enhanced interoperability aims to reduce legal uncertainty, encourage innovation, boost international competitiveness, and enhance public trust in AI.

Existential risks of artificial intelligence (AI) refer to the potential for advanced AI, particularly superintelligence, to cause human extinction or irreversible global catastrophe. Key risks include loss of control over superior AI, dangerous goal misalignment (the AI pursuing goals harmful to humans), and the weaponization of AI systems, with some studies suggesting a non-trivial risk of extinction.

AI governance in Asia faces significant challenges due to a fragmented regulatory landscape, balancing rapid economic innovation with safety, diverse technical capabilities, and ethical risks. Key hurdles include reconciling, a patchwork of national policies (e.g., China, Japan, ASEAN), addressing data privacy gaps, and managing talent shortages. Key challenges of AI governance in Asia include: Fragmented Regulatory Landscape: Asia lacks a unified AI regulatory framework. Countries have varying approaches—ranging from pro-innovation to restrictive—which creates regulatory arbitrage opportunities. Balancing Innovation and Safety: Governments are eager to leverage AI for economic growth (e.g., in ASEAN, projecting a US$950 billion GDP contribution by 2030), yet this goal can conflict with developing strict ethical and safety standards. Diverse Technical and Regulatory Capabilities: There is a wide gap in readiness. While Singapore sets regional standards through tools like AI Verify, others like Cambodia and Myanmar are still establishing foundational data protection laws. Data Governance and Privacy Gaps: Many regions struggle with data localization, cross-border transfers, and securing data infrastructure, complicating the creation of a seamless regional AI ecosystem. Emerging Technical Risks: Addressing algorithmic bias, accountability, and securing autonomous agentic systems is critical, particularly as organizations experiment with AI without mature governance structures. Geopolitical and Global Alignment: Asian nations face the challenge of aligning with, or navigating, global standards (like the EU AI Act) while maintaining local relevance, often risking being rule-takers rather than rule-makers.

The UN Global Dialogue on AI Governance, established in August 2025 by the UN General Assembly and operating under the Global Digital Compact, plays a crucial role in international AI cooperation by serving as a universal, inclusive platform designed to ensure AI governance reflects the priorities of all nations, not just technologically advanced ones. It serves as a complementary mechanism to existing efforts, aiming for interoperability of AI approaches across different national and regional regimes.

The AI Dialogue should act as a bridge between diverse regional and thematic approaches, ensuring they are complementary rather than conflicting. Global & Regional Frameworks: G7 Hiroshima AI Process (HAIP): Provides a framework for safe, secure, and trustworthy AI and code of conduct for organizations. EU AI Act: The world's first comprehensive, risk-based AI law. OECD AI Principles: Fosters international consensus on trustworthy AI. GPAI (Global Partnership on AI): Combines technical expertise with policy, now integrated with the OECD. ASEAN Responsible AI Roadmap (2025-2030): A regional framework for Southeast Asia. Council of Europe Framework Convention on AI: A legally binding treaty focusing on human rights. Multi-stakeholder & Technical Mechanisms: Internet Governance Forum (IGF): Serves as a model for multi-stakeholder participation. International Network for Advanced AI Measurement, Evaluation and Science: Focuses on technical evaluation infrastructure. Generation AI initiative (UNICEF): Focused on protecting children's rights. Partnership on AI (PAI): Brings together civil society, academia, and industry.

Different stakeholders can contribute to the AI Dialogue by bridging gaps in knowledge, standards, and ethical application. Key contributions include providing evidence-based input, identifying risks, and fostering international cooperation through inclusive, multi-stakeholder participation. The dialogue should focus on concrete, action-oriented governance rather than high-level principles, with sessions held in both Geneva and New York.

As of early 2026, global AI governance discussions remain heavily dominated by high-income, AI-intensive nations (principally the US, Europe, and China). Despite the rising influence of the Global South, significant gaps exist in representation, with only roughly 7% of global AI policies originating from Africa and Latin America combined. To create a truly representative global AI governance framework, the following steps are needed: Strengthen Institutional Representation (South-South Cooperation): Scale up initiatives like the Africa Asia AI Policymaker Network to foster collaboration, share policy playbooks, and build collective bargaining power. Create dedicated funding and travel support for delegates from Less Developed Countries (LDCs), Landlocked Developing Countries (LLDCs), and Small Island Developing States (SIDS) to participate in UN-led dialogues. Institutionalize Public & Civil Society Participation: Establish a standing "citizens-track" or civil society forum within major AI governance summits (such as the UN Global Dialogue on AI Governance) to ensure continuous input, rather than one-off consultations. Directly involve local NGOs and community organizations in data curation and auditing to identify bias. Implement "Equity by Design": Mandate diverse teams in the development and auditing of AI systems to ensure marginalized perspectives are included from the outset. Shift from relying only on AI auditing to incorporating the perspectives of community advocates in determining whether a system should be deployed at all. Localize AI and Data Governance: Encourage the creation of community-driven datasets and support linguistic diversity in AI models to protect indigenous and minority languages. Protect local data sovereignty by ensuring "data sovereignty" and "community ownership" of data.

To foster meaningful and dynamic engagement during an AI Dialogue in 2026, the most effective formats move beyond passive listening, leveraging AI-powered personalization, real-time sentiment analysis, and interactive, hands-on experiences.

As AI adoption accelerates, effective governance has shifted from high-level policy documents to operational, technical, and risk-based solutions. In 2026, key strategies include automated compliance, "privacy-by-design" data practices, and proactive risk management frameworks. Here are examples of policies, practices, platforms, and approaches promoting effective AI governance: 1. Policies and Frameworks EU AI Act: The world's first comprehensive AI law, setting legally binding standards based on a risk-based framework (prohibited, high-risk, limited-risk, minimal-risk). It requires strict transparency for General-Purpose AI (GPAI) and high-risk applications, with implementation continuing through 2026. NIST AI Risk Management Framework (AI RMF): A widely adopted, voluntary U.S. framework focused on four functions: Govern, Map, Measure, and Manage. It helps organizations identify and mitigate risks, often used as a foundation for building internal AI policies. Singapore's Model AI Governance Framework: A pioneering framework (updated in 2026 for agentic AI) that provides practical, non-binding guidance on managing risks in AI deployment, focusing on human accountability and technical controls. ISO/IEC 42001: The first international, certifiable standard for AI Management Systems (AIMS), enabling organizations to demonstrate governance maturity through third-party certification. 2. Operational Practices AI Inventory and Categorization: Maintaining a comprehensive, continuously updated register of all AI systems (including "Shadow AI" or unapproved tools) to track purpose, owners, and data sources. Human-in-the-Loop (HITL) Oversight: Ensuring that high-stakes decisions (e.g., in hiring or loan approvals) require human review, preventing fully autonomous systems from making decisions that could lead to unfair outcomes. Red Teaming and Adversarial Testing: Proactively testing AI systems with malicious inputs to identify vulnerabilities, bias, or safety risks before deployment, such as the AI Guardian suite (Litmus) in Singapore. Ethical Review Boards: Establishing cross-functional committees (combining legal, technical, and ethical experts) to review AI projects, such as IBM's AI Ethics Board. 3. Approaches to Addressing Challenges Data-Centric Governance: Recognizing that AI failures often result from poor data, organizations are investing in data lineage, quality monitoring, and "privacy-by-design" (e.g., differential privacy). Centralized-Federated Model: A central team sets standards and frameworks, while domain-specific teams (e.g., in finance or HR) apply them, balancing consistency with agility. Transparency Reports: Publishing documentation (e.g., model cards) that explains how a model was trained, its limitations, and its intended use, building trust with stakeholders. Regulatory Sandboxes: Controlled environments where companies can test AI products under regulatory supervision before full deployment, allowed under frameworks like the EU AI Act.