AGID

In my view, the first Global Dialogue on AI Governance would be successful if it delivers concrete, inclusive and implementation-oriented outcomes that help connect the growing diversity of AI governance approaches across levels. A first key outcome would be the development of a shared understanding of existing frameworks and initiatives, identifying areas of convergence and divergence across global, regional and national efforts. Given the richness of the current ecosystem—from the UNESCO Recommendation and OECD Principles to the EU AI Act and national strategies—greater coherence and mutual understanding are essential to avoid fragmentation and support interoperability. Second, success would require a strong focus on moving from principles to practice. The Dialogue should facilitate the exchange of concrete tools, methodologies and use cases, particularly supporting public administrations in translating governance principles into operational approaches. In this context, national experiences—such as the AI guidelines on development and procurement adopted in Italy—can provide practical examples of how principles such as transparency, accountability and human oversight can be embedded into real decision-making processes. Third, it should reinforce multistakeholder and inclusive participation, ensuring meaningful engagement from governments, private sector, academia, civil society and the technical community, including perspectives from developing countries. Fourth, the Dialogue should strengthen linkages with existing processes, notably the Internet Governance Forum and its National and Regional Initiatives, to ensure continuity, avoid duplication and enhance impact. Finally, an important outcome would be the definition of clear follow-up mechanisms, including priority areas for cooperation and ongoing platforms for knowledge exchange. Overall, the success of the first Dialogue will depend on its ability to connect principles with implementation and global discussions with national realities, laying the foundation for a more coherent and action-oriented AI governance ecosystem.

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

From my perspective, these four priorities reflect key areas where urgent action is needed to ensure that AI governance frameworks can be effectively translated into practice. Safe, secure and trustworthy AI and transparency, accountability, and human oversight are foundational to building public trust, particularly in the deployment of AI systems in the public sector. Public administrations have a responsibility to ensure that AI systems are reliable, explainable and aligned with fundamental rights. In this context, the experience of Italy-through the AI guidelines on development and procurement developed by AgID-demonstrates how these principles can be operationalized, embedding requirements on transparency, risk management and human oversight into decision-making processes. Interoperability of governance approaches is critical in a fragmented global landscape, where multiple regulatory and policy frameworks are emerging. Greater coherence across jurisdictions is essential to support cross-border cooperation and enable more consistent governance models. Finally, open-source software, open data and open AI models play a key enabling role in fostering transparency, innovation and inclusiveness. Promoting openness-where appropriate-can support accountability, facilitate knowledge sharing and reduce barriers to participation. In this context, strengthening capacity-building efforts remains essential to ensure that public institutions and stakeholders are equipped to effectively design, deploy and govern AI systems. Taken together, these priorities highlight the need to connect governance principles with implementation, while enabling a more open, coherent and trustworthy AI ecosystem.

In addition to the thematic areas already identified, several cross-cutting and emerging issues could further strengthen the Global Dialogue on AI Governance. First, there is a clear need to focus on the operationalization of governance frameworks, bridging the gap between high-level principles and real-world implementation. As highlighted by ongoing national and international experiences, translating principles into practical tools, processes and institutional practices remains a key challenge across contexts. Second, the environmental sustainability of AI systems should be recognized as a cross-cutting priority. This includes considerations related to energy consumption, infrastructure and lifecycle impacts, and their alignment with broader sustainable development objectives. Integrating sustainability into governance approaches will be essential to ensure that AI contributes positively to long-term societal goals. Third, the role of data governance and interoperability as foundational enablers cuts across multiple themes. Ensuring data quality, accessibility, security and interoperability is critical for the development of trustworthy AI systems and for enabling cooperation across jurisdictions. Fourth, the importance of capacity-building and inclusive participation emerges as a transversal dimension. Beyond technical skills, strengthening institutional and governance capacities is essential to enable meaningful engagement and effective implementation, particularly in contexts with different levels of readiness. Finally, the need to connect global discussions with national and local implementation remains a key cross-cutting issue. Strengthening linkages with multistakeholder platforms, such as the Internet Governance Forum and its National and Regional Initiatives, can support more effective two-way exchanges between global processes and local realities. Addressing these cross-cutting dimensions would contribute to a more coherent, implementation-oriented and sustainable global AI governance ecosystem.

Governance gaps and ongoing developments in the selected thematic areas are already having a significant impact on public administrations in Italy and, more broadly, across Europe. One of the main challenges concerns the translation of governance principles into operational practices. While comprehensive regulatory frameworks—such as the EU AI Act—provide an important foundation, public institutions are still adapting their internal processes to effectively implement requirements related to risk management, transparency, accountability and human oversight. This is particularly evident in areas such as AI procurement and system lifecycle management, where practical tools and methodologies are still evolving. A second challenge relates to the fragmentation of governance approaches at the global level. Differences in regulatory models, standards and levels of maturity can create barriers to interoperability and complicate cross-border cooperation, especially when AI systems and data flows operate across jurisdictions. In addition, the increasing availability of open data and open-source AI components presents both opportunities and risks. While openness can foster innovation, transparency and accessibility, it also requires robust governance to ensure quality, security and responsible use. At the same time, growing attention to the environmental sustainability of AI systems is introducing new considerations for public administrations, including energy consumption, infrastructure requirements and lifecycle impacts, which need to be integrated into governance and procurement processes. An additional cross-cutting challenge—and opportunity—relates to the need to strengthen AI literacy and public awareness across society. Beyond institutional and technical capacity, promoting a broader understanding of AI among citizens and stakeholders is essential to foster trust and enable informed participation. These developments also offer significant opportunities. In Italy, the work carried out by AgID particularly through the development of national AI guidelines on adoption, development and procurement is contributing to embedding governance principles into concrete decision-making processes within public administrations. Overall, these dynamics create an opportunity to enhance institutional capacity, foster policy learning and promote more coherent, interoperable and sustainable AI governance approaches.

The Global Dialogue on AI Governance can play a pivotal role in advancing international cooperation by acting as a bridge across diverse governance approaches, stakeholders and levels of implementation. First, it can support greater coherence and interoperability across existing frameworks and initiatives. In a landscape characterized by a wide range of global, regional and national efforts, the Dialogue can facilitate mutual understanding, identify areas of convergence and promote alignment, while avoiding duplication and building on existing processes. Second, the Dialogue can contribute to translating high-level principles into operational practices. By providing a space to exchange concrete tools, methodologies and use cases, it can support policy learning and help stakeholders—particularly public institutions—implement governance frameworks more effectively. Third, it can strengthen multistakeholder and multilevel cooperation, ensuring meaningful participation from governments, the private sector, civil society, academia and the technical community. In this context, strengthening cooperation with established platforms such as the Internet Governance Forum and its National and Regional Initiatives would be particularly valuable. These platforms can support structured, continuous and inclusive dialogue, and facilitate a two-way exchange between global discussions and national and regional implementation experiences. Fourth, it can support capacity-building and inclusive participation, helping to address differences in levels of readiness and enabling broader engagement in AI governance processes. Finally, by identifying priority areas for cooperation and fostering ongoing exchanges, the Dialogue can ensure continuity and sustained collaboration, contributing to a more coherent, inclusive and implementation-oriented global AI governance ecosystem.

The Global Dialogue on AI Governance should build upon and connect with the wide range of existing international, regional and national initiatives that are already shaping the AI governance landscape. At the global level, this includes frameworks such as the UNESCO Recommendation on the Ethics of AI, the OECD AI Principles, as well as ongoing work by UN entities, including UNDP and ITU. In parallel, important policy directions are emerging from fora such as the G7 and the G20, while at the regional level, initiatives such as the European Union's AI Act provide comprehensive regulatory approaches. Equally important are multistakeholder platforms and mechanisms, in particular the Internet Governance Forum and its National and Regional Initiatives. These provide established spaces for inclusive dialogue, policy exchange and bottom-up engagement, and can serve as valuable connectors between global discussions and national and local implementation. In addition, existing national strategies, guidelines and operational frameworks—including experiences related to AI development, procurement and deployment—represent an important source of practical knowledge that can inform global discussions. The added value of the Global Dialogue lies not in creating new frameworks, but in its ability to connect, align and operationalize existing efforts. In particular, it can: foster coherence and interoperability across governance approaches; support the translation of principles into practice through the exchange of tools and use cases; facilitate policy learning across countries with different levels of readiness; strengthen synergies between multistakeholder processes and institutional frameworks. By acting as a convening and connecting platform, the Dialogue can enhance the overall effectiveness, inclusiveness and impact of the global AI governance ecosystem.

Different stakeholders can contribute to the AI Dialogue by bringing complementary perspectives, expertise and experiences, which are essential to shaping effective and inclusive AI governance. Governments can contribute by sharing policy approaches, regulatory experiences and implementation challenges, particularly in areas such as public sector deployment, procurement and oversight. The private sector can provide insights on technological developments, innovation dynamics and practical constraints. Civil society plays a key role in highlighting societal impacts, fundamental rights and inclusion, while academia and the technical community can contribute evidence-based analysis, methodologies and standards. Local communities and users should also be enabled to share their experiences, ensuring that governance discussions are grounded in real-world impacts. To ensure meaningful contributions, the format and structure of the Dialogue will be critical. First, the Dialogue could benefit from a structured yet flexible architecture, combining plenary discussions with smaller, thematic and interactive sessions. Organizing discussions around a limited number of thematic clusters—covering both governance principles and implementation challenges—could support focus and coherence. Second, interactive and case-based formats should be prioritized. Dedicated sessions for sharing practical experiences, tools and use cases can help move from abstract discussions to concrete learning. Third, the Dialogue should ensure strong linkages with existing multistakeholder processes, such as the Internet Governance Forum and its National and Regional Initiatives, enabling a two-way exchange between global discussions and national and regional realities. Finally, establishing clear mechanisms for follow-up and continuity—such as communities of practice or regular exchanges—would help sustain engagement and ensure that stakeholder contributions translate into tangible outcomes. A well-designed, inclusive and implementation-oriented structure will be key to maximizing the value of stakeholder participation.

Despite increasing attention to inclusiveness, several voices and perspectives remain underrepresented in global discussions on AI governance. First, stakeholders from developing countries and least developed contexts often face structural barriers to participation, including limited institutional capacity, access to expertise and financial constraints. Second, local communities and end-users, including citizens more broadly, remain insufficiently represented. In particular, groups such as women, youth and older persons are often excluded not only due to structural barriers, but also because of limited access to the knowledge and skills needed to understand and meaningfully engage with AI-related issues. This can prevent them from effectively contributing to discussions that directly affect their lives. Third, small and medium-sized enterprises (SMEs) and public sector practitioners are often underrepresented compared to large technology actors, despite their key role in the practical deployment and use of AI systems. To address these gaps, it is essential to combine inclusive participation mechanisms with stronger efforts to promote AI literacy and public awareness. Enhancing understanding of AI—its opportunities, risks and implications—across society is a prerequisite for meaningful and informed participation. In addition, ensuring accessible and inclusive formats, including hybrid participation, multilingual engagement and targeted outreach, can help lower barriers to entry. Finally, leveraging existing multistakeholder platforms, such as the Internet Governance Forum and its National and Regional Initiatives, can provide structured channels to bring diverse and locally grounded perspectives into global discussions. A more inclusive approach—supported by stronger AI literacy—will be essential to ensure that AI governance processes are both representative and effective.

To foster meaningful and dynamic engagement, the AI Dialogue could benefit from adopting innovative, interactive and implementation-oriented formats that go beyond traditional panel discussions. First, case-based and scenario-driven sessions could play a central role. Presenting concrete use cases—such as AI deployment in public services, healthcare or education—can help ground discussions in real-world challenges and enable participants to explore practical solutions collaboratively. Second, the Dialogue could introduce multistakeholder "labs" or collaborative working sessions, where participants from different backgrounds jointly explore specific governance challenges and co-develop approaches, tools or recommendations. These formats can help move from dialogue to co-creation and experimentation. Third, small-group, facilitated discussions (e.g. roundtables or breakout sessions) can encourage more inclusive participation, allowing a wider range of voices to contribute meaningfully, including those who may be less comfortable speaking in larger plenary settings. Fourth, interactive digital tools—such as real-time polling, collaborative platforms or moderated online inputs—can enhance engagement, especially in hybrid settings, and allow participants to actively shape discussions. Fifth, peer-learning and experience-sharing sessions could provide space for governments, practitioners and other stakeholders to exchange concrete practices, including challenges encountered and lessons learned. Finally, ensuring strong linkages with ongoing multistakeholder processes, such as the Internet Governance Forum and its National and Regional Initiatives, could enable continuous engagement beyond the Dialogue itself, creating a more dynamic and sustained participation ecosystem. A combination of these formats can help ensure that the Dialogue is not only inclusive, but also practical, interactive and outcome-oriented.

Several existing policies, practices and platforms provide valuable examples of how AI governance can be effectively operationalized and can offer concrete solutions to current challenges. At the policy level, frameworks such as the UNESCO Recommendation on the Ethics of AI and the OECD AI Principles have played an important role in establishing common values and guiding principles for trustworthy, human-centric AI. At the regional level, the European Union's AI Act represents a comprehensive regulatory approach based on a risk-based framework. However, particularly relevant are implementation-oriented initiatives that translate these principles into practice. In this regard, the AI Hub for Sustainable Development at UNDP represents a promising example of a multistakeholder platform that combines capacity-building, infrastructure and concrete projects to support the responsible use of AI, particularly in developing contexts. Such initiatives demonstrate how governance can be embedded into real-world applications while contributing to sustainable development objectives. At the national level, experiences such as the AI guidelines on development and procurement developed by AgID in Italy provide further examples of how governance principles-such as transparency, accountability and human oversight-can be operationalized through practical tools and decision-making processes within public administrations. In addition, multistakeholder platforms such as the Internet Governance Forum and its National and Regional Initiatives offer effective models for inclusive and participatory governance, supporting continuous dialogue and knowledge exchange. Together, these examples highlight that effective AI governance requires not only principles and regulation, but also practical initiatives, operational tools and sustained collaboration across stakeholders and levels.