UN CSTD Data Governance Working Group, AGW Legal & Advisory, IFIP AI Governance

For the first Global Dialogue on AI Governance to be considered a success, it must move beyond high level discussion and deliver tangible, operational outcomes that establish the foundation for a minimum global benchmark for AI governance. In a world where major economies are adopting fundamentally different regulatory strategies, the Dialogue's value lies in identifying shared boundaries, common principles, and interoperable mechanisms that all jurisdictions can align with, regardless of political or economic system. A successful outcome would include: 1. Clear articulation of baseline expectations for safe, secure and trustworthy AI, grounded in human rights, transparency, accountability and risk management. 2. Agreement on minimum governance functions every country should implement—such as safety evaluations, incident reporting, data governance safeguards and oversight mechanisms—even if regulatory forms differ. 3. Commitment to interoperable governance approaches, enabling systems built in one jurisdiction to operate safely in another, reducing fragmentation and compliance burdens. 4. A roadmap for capacity building and bridging AI divides, ensuring developing countries can meaningfully participate in global AI ecosystems. 5. A mandate to advance both soft law and hard law, recognising that voluntary measures alone are insufficient for high risk or cross border AI systems. 6. A continuity mechanism—such as a standing platform, working groups or annual review cycle—to ensure the Dialogue leads to implementation, not merely consultation. Ultimately, success means producing the first globally recognised benchmark for AI governance—practical, enforceable, and capable of guiding national, regional and sectoral frameworks toward greater coherence in a divided geopolitical landscape.

• Protection and promotion of human rights
• Safe, secure and trustworthy AI
• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI

Given the requirement to select only four priorities, my choices reflect areas where early global convergence is both possible and urgently needed. First, "Protection and promotion of human rights" provides the normative anchor for all other governance functions. It inherently encompasses the principles of transparency, accountability and human oversight, which are instrumental to ensuring that AI systems respect dignity, autonomy and fairness. Embedding human rights at the centre of the Dialogue ensures that technical and regulatory measures remain grounded in universally recognised obligations. Second, "Safe, secure and trustworthy AI" captures the operational safeguards required to prevent harm and ensure reliability. This theme also provides space to address the governance of increasingly autonomous and agentic AI systems, where safety and security must be ensured across dynamic, in flight data flows and complex system behaviours. Within this framing, interoperability of governance approaches becomes indispensable. In a fragmented geopolitical environment, interoperability enables systems developed in one jurisdiction to operate safely in another, supports cross border data flows, and facilitates shared safety evaluations. In practice, interoperability is a prerequisite for achieving safe, secure and trustworthy AI at global scale. Third, "AI capacity building" is essential for global equity and meaningful participation. It implicitly includes the importance of open source software, open data and open AI models as enablers of innovation, transparency and shared benefit. Without sustained capacity building, many countries will be unable to implement even the most basic governance functions, deepening global divides. Fourth, addressing the social, economic, ethical, cultural, linguistic and technical implications of AI is critical. AI will reshape established human structures and processes, including workforce dynamics, skills requirements, education and training systems, intellectual property concepts, and the distribution of tangible and intangible wealth. Understanding these implications is essential for ensuring that AI contributes to inclusive and sustainable development. Together, these four priorities create a coherent foundation for establishing a minimum global benchmark for AI governance, balancing normative commitments, technical safeguards, capacity needs and cross border coherence.

Several cross cutting and emerging issues require explicit attention, as they are not fully captured by the existing thematic clusters but are rapidly becoming central to effective AI governance. First, the governance of AI agents, agentic AI systems, and dynamic agentic data flows is an emerging challenge that cuts across all themes. Agentic systems increasingly retrieve, generate, and act on data autonomously, blurring the boundaries between training data, operational data, and agent generated data. These systems introduce new risks related to memory accumulation, state retention, provenance, and real time decision making. Current themes do not explicitly address these dynamics, leaving a foundational gap in the Dialogue's coverage. Second, the indispensable link between AI governance and data governance requires stronger recognition. Data governance is the bedrock of AI governance, yet the Dialogue's themes treat it only indirectly. In practice, AI governance cannot succeed without robust governance across the entire data lifecycle, including data quality, provenance, interoperability, rights management, and stewardship models. This is especially important for ensuring safe, trustworthy AI in cross border contexts. Third, the rights and governance authorities of Indigenous Peoples and other underrepresented communities remain insufficiently addressed. Their perspectives are not merely ethical considerations but governance requirements grounded in international law, including FPIC, collective rights, and the CARE Principles. These issues cut across human rights, capacity building, and socio cultural impacts but are not explicitly captured in the current themes. Fourth, the geopolitical implications of AI-including concentration of compute, data, and model power-require explicit attention. Without addressing structural asymmetries, global governance risks reinforcing existing divides and limiting equitable participation in AI development and benefit sharing. Addressing these cross cutting issues will strengthen the Dialogue's ability to deliver a coherent, future ready global governance framework.

Across Australia and the broader Asia–Pacific region, governance gaps in the areas of human rights, safety, interoperability and capacity building are already producing both significant challenges and important opportunities. First, the rapid deployment of AI systems—particularly generative and emerging agentic AI—has outpaced existing human rights and accountability frameworks. This creates heightened risks of discrimination, opacity, data misuse and erosion of public trust. Many organisations lack the tools to assess rights impacts or ensure meaningful human oversight, especially as AI systems become more autonomous and capable of retrieving, transforming and acting on data in flight. Second, gaps in safety and security governance are becoming more visible. Australia and the region face increasing exposure to unsafe model behaviours, cyber enabled misuse, and vulnerabilities arising from weak data governance foundations. Agentic AI systems amplify these risks by operating across dynamic, continuously evolving data flows that current regulatory structures were not designed to manage. Third, the absence of interoperability across governance approaches is creating fragmentation. Divergent regulatory models affect cross border data flows, trade, compliance burdens and the ability of smaller economies to participate meaningfully in global AI ecosystems. Without interoperability, safe and trustworthy AI cannot scale internationally. Fourth, capacity building gaps remain a major barrier. Many countries in the region lack the institutional, technical and regulatory capacity to implement even baseline governance functions. Yet this also presents a strong opportunity: open source tools, open data, shared standards and regional cooperation can accelerate capability uplift and support more inclusive participation. Overall, these governance gaps risk deepening digital and AI divides. Yet they also present an opportunity for the Global Dialogue to establish a minimum global benchmark that supports safety, rights protection, interoperability and equitable capacity building across all regions.

The AI Dialogue can play a pivotal role in advancing international cooperation by serving as the first truly global, inclusive, and multi stakeholder platform dedicated to building coherence across diverse AI governance approaches. Because AI cuts across technical, social, economic, cultural, linguistic and geopolitical domains, the Dialogue must embrace the full spectrum of communities and stakeholders—governments, industry, academia, civil society, technical bodies, indigenous communities and under represented regions—to ensure that governance outcomes are legitimate, equitable and globally relevant. First, the Dialogue can create a shared space for convergence. Major economies are adopting fundamentally different regulatory strategies, and this fragmentation risks deepening global divides. The Dialogue can help identify common principles, shared boundaries, and interoperable governance mechanisms that allow systems developed in one jurisdiction to operate safely in another. Second, the Dialogue can support capacity building and knowledge exchange. Many countries lack the institutional, technical and regulatory capacity to implement baseline AI governance functions. By facilitating access to expertise, open tools, standards and best practices, the Dialogue can help ensure that all countries—not only technologically advanced ones—can participate meaningfully in global AI ecosystems. Third, the Dialogue can strengthen cooperation on safety, security and risk management. As AI systems become more autonomous and agentic, risks increasingly cross borders. The Dialogue can promote shared safety evaluations, incident reporting mechanisms, and coordinated responses to emerging risks. Fourth, the Dialogue can embed human rights and inclusive development at the centre of global governance efforts. By bringing together diverse voices—including those from the Global South, Indigenous communities, and civil society—the Dialogue can ensure that AI governance reflects global values and priorities, not only those of major powers. Ultimately, the AI Dialogue can become the foundation for a minimum global benchmark for AI governance, enabling cooperation in a geopolitically divided world.

As the Global Dialogue seeks to add value to the international governance landscape, it must build upon the substantial work already undertaken across the UN system and beyond. The most important of these is the UN CSTD Working Group on Data Governance, whose globally negotiated framework provides a coherent foundation for data stewardship, equity, benefit sharing, interoperability, and trusted cross border data flows. Because data is the fuel for generative and agentic AI systems, data governance is the bedrock of AI governance, the Dialogue should explicitly integrate this work. AI systems—particularly generative and agentic systems—depend entirely on the quality, provenance and governance of the data they ingest, retrieve, or generate. Without strong data governance, AI governance cannot succeed. At present, however, data governance and AI governance communities often operate in parallel. This fragmentation obscures a critical reality: data governance shapes AI, AI reshapes data governance, and both co evolve in a continuous feedback loop. The Dialogue can add significant value by bridging these communities and ensuring that AI governance frameworks incorporate governance across the entire data lifecycle. Second, the Dialogue should connect with existing global and regional normative instruments—including the EU AI Act, OECD AI Principles, UNESCO's Recommendation on the Ethics of AI and the Council of Europe's Framework Convention on AI. These instruments offer mature, interoperable elements that can be aligned rather than duplicated. Similarly, the work of technical standards bodies such as ISO/IEC JTC 1/SC 42 and ITU T provides operational tools for translating principles into practice. Third, the Dialogue must address emerging gaps not yet covered by existing frameworks—particularly the governance of agentic AI systems. These systems autonomously discover, retrieve and act on data, creating new categories of agent generated and agent retained data that existing frameworks do not fully address. This requires new approaches to runtime governance, memory governance and real time provenance and rights checks. Ultimately, the added value of the Global Dialogue lies in its ability to integrate these disparate initiatives into a unified, globally inclusive governance architecture that recognises data governance as foundational to safe, secure and trustworthy.

To ensure legitimacy, effectiveness and global relevance, the AI Dialogue must be designed as a genuinely inclusive, multi stakeholder process that enables meaningful participation from governments, industry, academia, civil society, technical communities, Indigenous Peoples, youth and under represented regions. First, participation should be enabled through diverse modalities, including written submissions, in person participation and virtual representation. Virtual formats are essential for ensuring equitable access for stakeholders who may lack the resources to attend in person, particularly from developing countries and small island states. Second, the Dialogue should adopt a structured, layered engagement model. This could include: • Open consultations for broad input on principles, risks and priorities • Thematic working groups aligned with the Dialogue's clusters (e.g., safety, human rights, interoperability, capacity‑building) • Technical roundtables involving standards bodies, researchers and practitioners • Regional dialogues to surface context‑specific needs and perspectives • Youth and civil society forums to ensure representation of diverse voices Third, the Dialogue should incorporate mechanisms for continuous participation, not only one off events. This may include online platforms for iterative drafting, transparent publication of submissions and opportunities to comment on evolving proposals. Fourth, the Dialogue should ensure that stakeholders from the Global South and marginalised communities have the resources and support needed to participate meaningfully, including translation, travel support, digital access and capacity building. Finally, the Dialogue should be structured to produce tangible outputs, such as draft benchmarks, interoperable governance elements and shared safety practices. Stakeholders should be invited not only to comment but to co create these outputs. By adopting these inclusive and flexible formats, the AI Dialogue can become a truly global platform that reflects the diversity of perspectives needed to shape effective and equitable AI governance.

Indigenous communities and stakeholders remain significantly underrepresented in global discussions on AI governance, despite being among the groups most affected by data practices, digital transformation and emerging AI systems. Indigenous Peoples are not merely stakeholders but rights holders with distinct governance authorities, cultural protocols and collective rights over data, knowledge and identity. Yet global AI governance processes rarely reflect these realities. This underrepresentation has concrete consequences. AI systems trained on ungoverned or misappropriated Indigenous data can perpetuate cultural harm, misrepresentation and extractive dynamics. The absence of Indigenous governance perspectives also means that global frameworks often overlook principles such as Free, Prior and Informed Consent (FPIC), collective rights, the CARE Principles for Indigenous Data Governance, and community defined benefit sharing. As a result, AI governance risks reinforcing historical inequities rather than addressing them. To meaningfully include Indigenous communities, the AI Dialogue should adopt the recommendations articulated in the Sarawak MultiMedia Authority Submission to the UN CSTD Data Governance Working Group, including: • Recognition of Indigenous Peoples as governance authorities, not just stakeholders, in AI and data governance processes. • Dedicated consultation pathways that respect Indigenous decision‑making structures, cultural protocols and community‑defined governance mechanisms. • Inclusion of Indigenous experts, institutions and community representatives in thematic working groups, drafting processes and advisory bodies. • Integration of Indigenous data governance frameworks—including FPIC, CARE Principles and community‑governed data infrastructures—into AI governance standards and benchmarks. • Resourcing and capacity‑building to ensure Indigenous communities can participate meaningfully, including translation, travel support and digital access. By embedding Indigenous perspectives and governance frameworks into the Global Dialogue, the process can become more equitable, culturally grounded and globally legitimate—ensuring that AI systems respect the rights, identities and knowledge systems of Indigenous Peoples worldwide.

To foster meaningful and dynamic engagement, the AI Dialogue should adopt innovative formats that enable broad, equitable and culturally grounded participation—particularly from communities historically excluded from global digital governance processes. First, hybrid and multi modal participation formats should be standard. This includes in person sessions, virtual participation, asynchronous written inputs and low bandwidth options to ensure accessibility for stakeholders in remote or resource constrained regions. Virtual "open rooms" and rotating time zone friendly sessions can broaden global reach. Second, the Dialogue should incorporate culturally grounded engagement models, such as community roundtables, Elders' circles and Indigenous led dialogues. These formats respect traditional decision making structures and ensure that Indigenous knowledge systems and governance principles—such as stewardship, collective rights and FPIC—are meaningfully integrated. Third, thematic co-creation labs can bring together governments, technical experts, civil society, Indigenous communities and industry to collaboratively draft governance elements, benchmarks and safeguards. These labs should be problem focused, iterative and designed to produce tangible outputs. Fourth, regional and localised dialogues can surface context specific needs, linguistic diversity and culturally relevant governance considerations. These should feed directly into the global process. Finally, transparent, continuous engagement platforms—including open drafting portals, public comment cycles and community feedback loops—can ensure that participation is sustained throughout the governance process. By adopting these innovative formats, the AI Dialogue can become a genuinely inclusive, dynamic and globally representative process that reflects diverse knowledge systems and governance traditions.

Several existing initiatives offer practical, scalable, and globally relevant approaches to strengthening AI governance-particularly when viewed through the lens of data governance as the foundational layer. First, the UN CSTD Working Group on Data Governance provides one of the most mature, globally negotiated frameworks for governing data across its lifecycle. Its emphasis on stewardship, equity, benefit sharing, interoperability, and trusted cross border data flows is directly applicable to AI governance, especially as generative and agentic AI systems increasingly rely on dynamic, in flight data retrieval and agent generated data. Second, the EU AI Act, OECD AI Principles and UNESCO's Recommendation on the Ethics of AI offer globally recognised normative frameworks. They provide actionable guidance on transparency, accountability, human oversight, and risk management, and have been adopted or adapted by dozens of countries. Their strength lies in their interoperability and their ability to serve as bridges between diverse regulatory models. Third, technical standards bodies-including ISO/IEC JTC 1/SC 42 and ITU T Study Groups-are producing concrete, implementable standards that translate principles into practice. ISO/IEC 38505 (data governance), ISO/IEC 42001 (AI management systems), and ISO/IEC 5259 (data quality for analytics and ML) provide operational tools for organisations to implement lifecycle governance, data quality controls, and risk management processes aligned with global expectations. Fourth, emerging practices in agentic AI governance-such as memory governance, real time provenance checks, dynamic data flow controls, and runtime rights management-demonstrate how governance must evolve to address new system architectures. These approaches offer practical mechanisms for managing risks associated with autonomous, self retrieving, and self modifying AI systems. Together, these initiatives show that effective AI governance is achievable when built on strong data governance foundations, interoperable standards, and inclusive, rights based approaches. They provide a solid base upon which the Global Dialogue can build a coherent, future ready governance architecture.