IBM

IBM welcomes the opportunity to contribute to the Global Dialogue on AI Governance convened by the United Nations. As a global technology company with several decades of experience in AI research, development, and deployment, IBM is committed to advancing AI that is inclusive, trustworthy, and beneficial to society. We believe that two of the proposed thematic clusters are especially important for the success of the Global Dialogue on AI Governance: • Bridging AI divides: capacity-building, access, and digital foundations • Safe, secure, and trustworthy AI: responsible and interoperable approaches We believe these areas are mutually reinforcing and foundational to ensuring that AI contributes to sustainable and equitable global development. AI presents transformative opportunities, but without targeted action, it risks exacerbating existing global inequalities. Addressing disparities in access to skills, infrastructure, and resources is critical. We encourage the UN system to support frameworks that promote global recognition of AI-related credentials and facilitate cross-border collaboration in workforce development. IBM also encourages the Global Dialogue to promote trusted open ecosystems that balance accessibility with safeguards for security, privacy, and intellectual property. We also believe that the Global Dialogue should support the development of globally aligned lifecycle governance approaches that integrate technical, organizational, and operational safeguards. Fragmentation in regulatory and policy approaches can create barriers to innovation and deployment. Interoperability is key. The Global Dialogue can play a critical convening role in fostering coherence across national and regional AI governance frameworks. Not all AI systems pose the same level of risk. Governance frameworks should be risk-based, focusing on high-impact and high-risk use cases; context-sensitive, accounting for sectoral and cultural differences; and flexible and adaptive, keeping pace with technological evolution. The Global Dialogue could play an important role in supporting supports policies that enable innovation while ensuring appropriate safeguards where risks are greatest.

• Interoperability of governance approaches
• Transparency, accountability, and human oversight
• Open-source software, open data and open AI models
• Safe, secure and trustworthy AI

Addressing disparities in access to skills, infrastructure, and resources is critical. IBM emphasizes scalable, inclusive approaches to digital and AI literacy. Initiatives such as IBM SkillsBuild provide free access to digital, technical, and professional skills training to underserved communities worldwide. We encourage the UN system to support frameworks that promote global recognition of AI-related credentials and facilitate cross-border collaboration in the workforce. Bridging AI divides requires investment in high-performance computing (HPC) infrastructure, reliable connectivity and cloud access, and data ecosystems that are inclusive and representative. IBM supports initiatives that expand equitable access to compute resources, including through cloud-based platforms and hybrid models that allow countries to leverage global infrastructure while maintaining sovereignty over data and systems. Open-source software and models lower barriers to entry, responsibly-governed open datasets enable local innovation, and collaborative research ecosystems foster knowledge transfer. IBM encourages the UN to promote trusted open ecosystems that balance accessibility with safeguards for security, privacy, and intellectual property. Trustworthy AI is essential to realizing the benefits of AI at scale. IBM has implemented internal governance structures, processes, and tools to ensure responsible AI development and deployment. We encourage the development of globally aligned lifecycle governance approaches that integrate technical, organizational, and operational safeguards. The UN can refer to existing open-source tools, such as Singapore's AI Verify. IBM recommends alignment on core principles and definitions in regulatory and policy approaches across jurisdictions, mutual recognition of standards and assurance mechanisms, and international collaboration on technical standards, including through multi-stakeholder bodies. The UN should build on existing efforts, including the Hiroshima AI Process, Singapore's Model AI Governance Frameworks, and the OECD. Governance frameworks should be risk-based, context-sensitive, and flexible and adaptive. IBM supports policies that enable innovation while ensuring appropriate safeguards where risks are greatest.

Effective AI governance requires collaboration among governments, industry, academia, and civil society. IBM encourages the UN to continue fostering inclusive and multi-stakeholder dialogue mechanisms. IBM supports recognizing the increasing infrastructure and energy demands of AI and fostering multi-stakeholder collaboration to advance principles for efficient, resilient, and sustainable AI system design. IBM supports the recognition of voluntary AI Commons initiatives, such as impact platforms, trusted AI frameworks, and science networks that enable the sharing of benchmarks, best practices, and scalable use cases across countries to advance trusted and inclusive global AI governance. In addition to technical skills, there is a need to build capacity in regulatory and policy expertise, standards development and implementation, and AI governance. Targeted support for developing countries is essential to ensure equitable participation in global AI governance. AI has the potential to accelerate progress toward the UN Sustainable Development Goals, but only if it is developed and deployed in a way that is inclusive, trustworthy, and aligned with shared global values. IBM stands ready to collaborate with the United Nations and other stakeholders to expand access to AI skills and infrastructure, advance responsible, open, and trustworthy AI practices, and promote interoperable and inclusive governance frameworks. We look forward to contributing to the ongoing Global Dialogue for AI Governance.

In the technology sector, governance gaps in AI are creating both significant challenges and important opportunities. A primary challenge is fragmentation across regulatory and policy frameworks. Divergent national and regional approaches to AI governance increase complexity for organizations operating globally, raising compliance costs and slowing the deployment of beneficial AI systems. The lack of interoperability across frameworks also limits the scalability of trustworthy AI practices. A second challenge is the uneven distribution of capabilities and infrastructure. Gaps in access to high-performance computing, quality data, and AI skills risk deepening global inequalities and limiting the ability of many countries and organizations to participate meaningfully in the AI economy. A third challenge relates to operationalizing trustworthy AI. While high-level principles such as fairness, transparency, and accountability are widely endorsed, there is still a need for practical and standardized methods to implement and measure the operationalization of these principles consistently across contexts. These gaps create also important opportunities. First, there is an opportunity to advance interoperable and risk-based governance approaches that align across jurisdictions while remaining adaptable to local contexts. International dialogue can help reduce fragmentation and support mutual recognition of standards and assurance mechanisms. Second, there is an opportunity to scale inclusive capacity-building efforts, including a global recognition of AI skills and an expanded access to digital infrastructure, enabling broader participation in AI development and use. Finally, there is an opportunity to strengthen trusted open ecosystems, including open-source models, tools, and benchmarks, which can accelerate innovation while embedding safeguards for security, privacy, and accountability. Addressing these challenges collaboratively can enable AI to deliver more equitable and trustworthy outcomes globally.

The Global AI Dialogue can play a critical role in advancing international cooperation by acting as a practical, multi-stakeholder platform to promote alignment, inclusion, and implementation of AI governance approaches. First, the Dialogue can help address fragmentation by fostering interoperability across governance frameworks. By promoting alignment on core principles, shared definitions, and risk-based, lifecycle-oriented approaches, it can reduce unnecessary divergence and support mutual recognition of standards and assurance mechanisms. This is essential to enable innovation while ensuring consistent safeguards across jurisdictions. Second, the Dialogue can facilitate practical exchange of tools, frameworks, and implementation approaches. Many organizations have developed operational methods for governing AI systems, including technical tools, risk management processes, and internal governance structures. Sharing and adapting these resources can accelerate the translation of high-level principles into real-world practice. Third, the Dialogue can play a central role in bridging AI divides through coordinated capacity-building efforts. This includes promoting global recognition of AI-related skills and credentials, expanding access to digital infrastructure and compute resources, and supporting the development of local AI ecosystems. Targeted support for developing countries is essential to ensure equitable participation in both AI innovation and governance. Fourth, the Dialogue can advance trusted open ecosystems, encouraging collaboration around open-source software, open data, and open AI models, while ensuring appropriate safeguards for security, privacy, and intellectual property. Finally, the Dialogue can support the development of globally aligned lifecycle governance approaches that integrate technical, organizational, and operational safeguards, helping organizations operationalize trustworthy AI consistently across contexts. Through these roles, the Global AI Dialogue can strengthen international cooperation and enable more inclusive, interoperable, and trustworthy AI governance.

The Global AI Dialogue should build on a range of existing international, multi-stakeholder, and technical initiatives that have already advanced AI governance, while helping to connect and scale their impact. Key initiatives include the OECD AI Principles, which provide a widely endorsed foundation for trustworthy AI; the Hiroshima AI Process, which has advanced discussions among major economies on generative AI governance; and national and regional frameworks such as Singapore Model AI Governance Framework and technical assurance tools like AI Verify. In addition, standards development organizations such as ISO and IEEE play a critical role in translating principles into technical specifications. The Dialogue should also connect with open and collaborative ecosystems, including open-source communities and AI Commons initiatives that enable the sharing of datasets, models, benchmarks, and best practices across borders. While these efforts are valuable, they remain fragmented across geographies, sectors, and levels of implementation. The added value of the Global AI Dialogue lies in its ability to act as a neutral convening platform that connects these initiatives, promotes coherence, and identifies pathways toward interoperability. Specifically, the Dialogue can: • Facilitate alignment and mutual recognition across frameworks and standards • Promote practical interoperability by mapping how different governance approaches relate and can be operationalized together • Support capacity-building by making existing tools, frameworks, and best practices more accessible, particularly for developing countries • Encourage the development of trusted open ecosystems, balancing openness with safeguards for security, privacy, and intellectual property By building bridges across existing efforts and focusing on implementation, the Global AI Dialogue can amplify impact and accelerate the development of inclusive, trustworthy, and interoperable AI governance globally.

Effective participation in the Global AI Dialogue requires meaningful contributions from governments, industry, academia, and civil society, each bringing complementary expertise. • Governments can provide policy direction, identify national priorities, and support alignment across regulatory frameworks. • Industry can contribute practical experience in developing and deploying AI systems, including governance tools, risk management processes, and operational best practices. • Academia and the technical community can advance research, evaluation methodologies, and independent analysis of emerging risks and capabilities. • Civil society can represent societal perspectives, including impacts on human rights, equity, and inclusion, and help ensure accountability. To maximize impact, the Dialogue should be structured as a continuous, action-oriented process, rather than a one-time event. First, it should combine high-level plenary discussions with thematic working groups focused on priority areas such as interoperability, capacity-building, and trustworthy AI. These working groups should produce practical outputs, including guidance, toolkits, and mappings of existing frameworks. Second, the Dialogue should incorporate regular multi-stakeholder consultations and open calls for input, ensuring broad geographic and sectoral representation, including from developing countries. Third, it should emphasize implementation and knowledge-sharing, including repositories of governance tools, case studies, and best practices that stakeholders can adapt to their contexts. Fourth, the Dialogue should support pilot collaborations or "sandboxes", where stakeholders can test interoperable governance approaches and technical solutions in real-world settings. Finally, the Dialogue should include mechanisms for continuity and accountability, such as periodic progress reviews and measurable outcomes. This structure would enable the Global AI Dialogue to move from principles to practice, fostering inclusive, interoperable, and trustworthy AI governance.

Despite increasing global attention to AI governance, several important voices remain underrepresented. First, stakeholders from developing countries, particularly in the Global South, are often underrepresented due to gaps in technical capacity, infrastructure, and access to policy forums. This limits the diversity of perspectives shaping global norms and risks reinforcing existing inequalities. Second, small and medium-sized enterprises (SMEs) and startups are less visible in governance discussions, despite being key drivers of innovation and adoption. Their operational constraints and practical insights are essential to designing feasible and scalable governance approaches. Third, workers and affected communities, including those impacted by AI-driven decisions in areas such as employment, healthcare, and public services, are often not directly included. Their lived experiences are critical to understanding real-world impacts and risks. Fourth, linguistic and cultural minorities remain underrepresented, particularly in discussions on data, model development, and evaluation, where dominant languages and contexts shape outcomes. Finally, interdisciplinary perspectives, including from social sciences and humanities, are still not fully integrated into technical and policy discussions. To address these gaps, the Global AI Dialogue can: • Expand access and participation, including through funding mechanisms, regional hubs, and hybrid participation formats • Invest in capacity-building, particularly in AI governance, policy, and technical skills • Support multilingual engagement and locally grounded consultations • Create structured channels for input from SMEs, workers, and civil society organizations • Integrate interdisciplinary expertise into working groups and outputs Broadening participation is essential to ensuring that AI governance frameworks are inclusive, context-sensitive, and globally legitimate.

To foster meaningful and dynamic engagement, the Global AI Dialogue should complement traditional discussions with interactive, implementation-oriented formats that enable collaboration across stakeholders. First, the Dialogue could include thematic "implementation labs" or policy sprints, where governments, industry, academia, and civil society work together on concrete challenges, such as interoperability of governance frameworks or operationalizing transparency and accountability. These sessions should produce tangible outputs, such as draft guidance, mappings, or prototype tools. Second, regulatory and technical sandboxes could allow participants to test governance approaches, standards, and tools in real-world or simulated environments. This would help identify gaps, improve interoperability, and build shared understanding of what works in practice. Third, the Dialogue could host global repositories and live demonstrations of governance tools, frameworks, and open-source solutions. Interactive showcases would enable participants to explore how responsible AI practices are implemented across different contexts and sectors. Fourth, regional and multilingual dialogue tracks could ensure broader participation and bring locally grounded perspectives into global discussions. These could be connected through a central platform that synthesizes insights and feeds them into global outcomes. Fifth, the use of structured multi-stakeholder deliberation formats, such as facilitated roundtables or consensus-building exercises, could help bridge differing perspectives and move toward shared positions on complex issues. Finally, continuous digital engagement platforms could support ongoing collaboration beyond formal sessions, enabling iterative input, knowledge-sharing, and co-creation of outputs. These formats would help the Global AI Dialogue move beyond high-level exchange toward practical collaboration, accelerating the development of inclusive, interoperable, and trustworthy AI governance.

A number of existing policies, practices, and platforms provide concrete approaches to operationalizing effective AI governance. At the policy level, the OECD AI Principles and the EU AI Act exemplify risk-based and human-centric approaches, establishing differentiated obligations based on the level of risk and emphasizing accountability, transparency, and oversight. At the organizational level, companies have developed end-to-end AI governance frameworks that integrate principles into practice across the AI lifecycle. These include internal review processes, risk assessment methodologies, and technical tooling to address issues such as bias, explainability, robustness, and monitoring after deployment. Embedding governance into development workflows is critical to ensuring consistent and scalable implementation. Technical platforms and tools also play a key role. For example, AI Verify provides a testing and validation framework to assess AI systems against governance principles, while open-source toolkits enable developers to detect bias, improve explainability, and document model behavior. These tools help translate high-level requirements into measurable and auditable practices. In addition, open and collaborative ecosystems are essential. Open-source software, shared benchmarks, and common evaluation methodologies enable broader participation and foster transparency, while supporting innovation. When combined with appropriate safeguards for privacy, security, and intellectual property, these ecosystems can scale responsible AI practices globally. Finally, capacity-building initiatives, such as IBM SkillsBuild, demonstrate how expanding access to AI and digital skills can support both adoption and governance, particularly in underserved communities. Together, these approaches highlight the importance of combining risk-based policies, lifecycle governance practices, technical tools, open ecosystems, and capacity-building to achieve effective and scalable AI governance.